Flexible Information in the Social Sounds of Humpback Whales

Flexible Information in the Social Sounds of

Humpback Whales

Dana Anne Cusano

Bachelor of Arts (cum laude), Master of Research

0000-0002-4186-4206

A thesis submitted for the degree of Doctor of Philosophy at The

University of Queensland in 2020

School of Veterinary Science

i Abstract

Animals living in a highly social environment typically have frequent and diverse interactions. To facilitate these relationships, social animals often have complex communication systems consisting of both between- and within-call variation. Such variability may manifest in a diverse number of call types (between-call variation) as well as the potential for conveying information on the signaller’s internal motivational state or arousal (within-call variation). These aspects may be particularly important for social species or during complex social interactions, a concept known as the ‘social complexity hypothesis for communicative complexity’. However, not all species appear to conform to these trends. The humpback whale (Megaptera novaeangliae), like other baleen whales, has a supposedly simple social system characterised by small, temporary, and unstable groups. Despite this, humpback whales have one of the most complex communication systems of any non-human animal, especially during breeding-related social interactions. Although these interactions are undoubtedly mediated using acoustic signals, how potential information is conveyed (e.g. through call types and/or through changes in the structure of the calls) is poorly understood.

This thesis examines the potential information in the social calls of humpback whales, with a particular emphasis on within-call structural variation (e.g. changes in frequency, duration, or bandwidth). As humpback whales are thought to have a simple social system, this thesis also aims to determine the potential link between the complex communication system of this species and social interactions during behaviours associated with breeding. To do this, data were collected during interactions between males and females on the breeding grounds in the Great Barrier Reef and during their annual southward migration back towards the feeding grounds in Antarctica. The variety of these breeding interactions ranges from simple social interactions between a single female and her calf, to moderately social groups consisting of a female-calf pair escorted by a single male, and ultimately to highly social competitive groups comprised of multiple males aggressively competing for access to a female. Using the breeding ground data, competitive groups were further split into three ‘intensity’ levels, defined by a progressive increase in their levels of aggression and arousal.

A novel quantitative method was used to classify the acoustic repertoire of social calls in humpback whales and determine which call types were structurally stereotyped (‘discrete’) or structurally variable (‘graded’). The repertoire was found to consist of 15 call types, made up of six discrete and

ii nine graded calls. However, even the discrete calls showed significant variability in their acoustic structure, especially during social interactions. As acoustic variation is correlated with relatively high informative value, this indicates that all humpback whale calls have the potential to convey a large amount of information. Some calls also likely contain ‘multiple messages’ and may convey information on more than one attribute, such as temporary dominance status and individual identity.

In female-calf only groups, humpback whales produced calls at lower rates and used fewer call types. As the number and diversity of social interactions increased (defined as the addition of one or more male escorts), individual call rates increased, as did the use of long bouts of calls. Further, new discrete call types were introduced into the repertoire which were never heard in the more simple female-calf dyads. This meant that some calls demonstrated a degree of context-specificity, and were only heard in groups with one or more escorts. The use of certain calls in limited contexts is another characteristic linked with increased communicative capacity. This may be especially true if they also display acoustic features that are indicative of certain motivational states or arousal levels.

Graded call types were used more frequently than discrete call types across all levels of social interaction. However, during competitive behaviour, there was a significant increase in the overall proportion of graded calls used compared with discrete calls as the levels of aggression and group arousal increased. There was also a significant increase in the rate of individual call production with increased intensity. These results demonstrate a striking similarity between the displays and ritualised fighting in terrestrial animals, a concept not described in other cetaceans.

This thesis represents a dedicated and comprehensive assessment of the information content of humpback whale social calls, and one of the first to investigate the relationship between social interactions and communicative complexity in a marine mammal. The results of this research have provided insight into the repertoire of the humpback whale, and which call types are likely to contain information relevant during social interactions related to breeding. Overall, the results presented here support high communicative complexity in this species, both in the repertoire and within call types. Not only does this represent a critical step towards identifying the function of calls, it has broad implications for animal behaviour and communication in general. Using the classification methods for future analyses of other whale species will allow for comparisons which can inform researchers about the possible drivers of communicative complexity in this group of animals.

iii

Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, financial support and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my higher degree by research candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co-authors for any jointly authored works included in the thesis.

Publications included in this thesis

No publications included.

Submitted manuscripts included in this thesis

The manuscript appearing as Chapter 2 ‘Humpback whale social call production reflects both motivational state and arousal’ has been accepted to the peer-reviewed journal Bioacoustics. This manuscript was a collaborative effort, with specific author contributions listed on page 17.

Other publications during candidature

Peer-reviewed papers

1. Parks SE, Cusano DA, Van Parijs SM, Nowacek DP. 2019. Evidence for acoustic crypsis in communication by right whale mother-calf pairs on the calving grounds. Biology Letters 15:1- 6. 2. Parks SE, Cusano DA, Van Parijs SM, Nowacek DP. 2019. North Atlantic right whale (Eubalaena glacialis) acoustic behavior on the calving grounds. Journal of the Acoustical Society of America 146: EL15-21. 3. Cusano DA, Conger LA, Van Parijs SM, Parks SE. 2018. Implementing conservation measures for the North Atlantic right whale: considering the behavioral ontogeny of mother- calf pairs. Animal Conservation 22: 228-237. 4. Root-Gutteridge H, Cusano DA, Shiu Y, Nowacek DP, Van Parijs SM, Parks SE. 2018. A lifetime of changing calls: North Atlantic right whales, Eubalaena glacialis, refine call production as they age. Animal Behaviour 137: 21-34. 5. Petraccione J, Root-Gutteridge H, Cusano DA, Parks SE. 2017. Exploring the early social affiliations and behaviour of a captive Asian elephant (Elephas maximus) calf. Journal of Zoo and Aquarium Research 5: 131-136. 6. Hunsinger E, Root-Gutteridge H, Cusano DA, Parks SE. 2017. A description of defensive hiss types in the flat horned hissing cockroach (Aeluropoda insignis). Bioacoustics 27: 261-271. 7. Parks SE, Cusano DA, Bocconcelli A, Friedlaender AS, Wiley DN. 2016. Noise impacts on social sound production by foraging humpback whales. Proceedings of Meetings on Acoustics 27: 01009.

8. Cusano DA, Matthews LP, Grapsten E, Parks SE. 2016. Effects of Increasing Temperature on Acoustic Advertisement in the Tettigoniidae. Journal of Orthoptera Research 25: 39-47.

Conference abstracts

1. Cusano DA, Noad MJ, Dunlop RA. 2019. The communication and behaviour of humpback whales in competitive groups on their breeding ground. The Ecological Society of Australia Annual Conference, Launceston, Tasmania, Australia. 2. Cusano DA, Noad MJ, Dunlop RA. 2018. Motivational information within social sounds of humpback whales (Megaptera novaeangliae). The 4th Australia/New Zealand Student Chapter Meeting of the Society for Marine Mammalogy, Brisbane, Queensland, Australia. 3. Parks SE, Cusano DA, Van Parijs SM, Nowacek D. 2017. Acoustic communication of North Atlantic right whale (Eubalaena glacialis) mother-calf pairs on the calving grounds. 22nd Biennial Conference on the Biology of Marine Mammals, Halifax, Nova Scotia, Canada. 4. Cusano DA, Noad MJ, Dunlop RA. 2017. The repertoire, call rate, and information content of non-song vocalizations in humpback whale social groups. 22nd Biennial Conference on the Biology of Marine Mammals, Society of Marine Mammalogy Biennial Meeting. Halifax, Nova Scotia, Canada. 5. Cusano DA, Noad MJ, Dunlop RA. 2016. Motivational information within vocalisations of humpback whales (Megaptera novaeangliae). 3rd Conference of the Australia New Zealand Student Chapter of the Society for Marine Mammalogy, Adelaide, South Australia. 6. Parks SE, Cusano DA, Bocconcelli A, Friedlaender AS, Wiley DN. 2016. Noise impacts on social sound production by foraging humpback whales. Effects of Noise on Aquatic Life IV, Dublin, Ireland.

Contributions by others to the thesis

Chapter 2: Cusano DA, Indeck KL, Noad MJ, Dunlop RA. Humpback whale social call production reflects both motivational state and arousal. KI contributed a portion of the data, assisted with data analysis, and critically reviewed the final manuscript; MN and RD obtained funding for and led the BRAHSS project, oversaw fieldwork, assisted with the interpretation of results, and edited the final manuscript; RD additionally contributed to the development of the research concept; DC was responsible for the majority of data processing and analysis, contributed to the development of the research concept, interpreted the results, and prepared the final manuscript.

Chapter 3: Cusano DA, Noad MJ, Dunlop RA. Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering. MN and RD obtained funding for and led the BRAHSS project, oversaw fieldwork, assisted with the interpretation of results, and edited the final manuscript; RD additionally contributed to the development of the research concept; DC was responsible for all data processing and analysis, contributed to the development of the research concept, interpreted the results, and prepared the final manuscript.

Chapter 4: Cusano DA, Indeck KL, Noad MJ, Dunlop RA. Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale. KI contributed a portion of the data, assisted with data analysis, and critically reviewed the final manuscript; MN and RD obtained funding for and led the BRAHSS project, oversaw fieldwork, assisted with the interpretation of results, and edited the final manuscript; RD additionally contributed to the development of the research concept; DC was responsible for the majority of data processing and analysis, contributed to the development of the research concept, interpreted the results, and prepared the final manuscript.

Chapter 5: Cusano DA, Noad MJ, Dunlop RA. The differential use of discrete and graded calls during intraspecific conflict in the humpback whale. MN and RD assisted with the interpretation of results, and edited the final manuscript; RD additionally contributed to the development of the research concept; DC was responsible for obtaining funding for field data collection, oversaw data collection, processed and analysed all data, contributed to the development of the research concept, interpreted the results, and prepared the final manuscript.

Statement of parts of the thesis submitted to qualify for the award of another degree

No works submitted towards another degree have been included in this thesis.

Research involving human or animal subjects

The data collected for this thesis were approved by The University of Queensland’s Animal Ethics Committee for Native/Exotic Wildlife and Marine Animals.

vii

The majority of data on tagged female humpback whales were collected during previous associated projects (2010, 2011, 2014) under permit numbers SVS/230/10(NF) and CURTIN/SVS/283/13. Data collected from tagged female humpback whales in 2017 were collected under permit number SVS/174/17. Data from competitive groups were collected on the Great Barrier Reef, Australia from 2016-2019 under permit numbers SVS/245/16 and SVS/129/19.

Copies of all animal ethics permits can be found in Appendix 5.

viii

Acknowledgements

This thesis would not have been possible with the substantial support from friends, family, and colleagues. First and foremost, I would like to thank my supervisors Bec Dunlop and Mike Noad. Your support and guidance has been invaluable, both in the field and in the office. Bec, your jumpers are now safe. To my lab mates: Kate Indeck, Jenny Allen, Katya Ovsyanikova, Barry McGovern, Marjoleine ‘Skip’ Roos, Léonie Huijser, Elisa Girola, and Jake Linsky; and the former (and honorary) members who have maintained their group chat status for a reason: Lara Pogson- Manning, Gretta Chaplin, Angelique Burden, and Sean Johnston. You guys are the best support network, and a fine group of human beings. A special thank you to Elisa and Maëlle Torterotot for the development of the custom MATLAB code, to the Moreton Bay Research Station for hosting the CEAL Group, and to the Quandamooka people, who are the traditional owners of the land on which the research station was built. Additional thanks to my examiners, Julia Fischer and Laela Sayigh, and my thesis committee, Justine Gibson, Deanne Whitworth, and Malcolm Jones, all of whom have provided invaluable feedback to make this thesis stronger.

Much of the data were collected as part of the Behavioural Response of Australian Humpback whales to Seismic Surveys (BRAHSS). I would like to thank the many people involved in BRAHSS, especially the numerous volunteers. Field work for this project was made possible by funding from the Joint Industry Programme on E&P Sound and Marine Life, managed by the International Association of Oil and Gas Producers.

Data collection in the Great Barrier Reef was in collaboration with the Great Barrier Reef Whale and Dolphin Research Programme. Financial support for my participation in this project was provided by the Ecological Society of Australia and the Holsworth Wildlife Research Endowment. To the countless people involved in this research, thank you for making data collection in this remarkable place even possible. Special thanks to Dave Paton for running the show, Damien Morales for managing everything else and driving countless hours searching for and chasing whales, Corey Lardner for jumping in head first, driving like a champ, and not breaking my mug, Lesley Douglas for being the organisational guru, Sally Hughes-Allan for cooking for far too many people and catering to those who hate onions, and the research assistants that have returned again and again to support this project: legacy member Mark Cornish, Daryn McKenny, Andrew Penfold, Peter Lake, Cassie Smith, Virginie Rousseaux, Annie Swain, and my ‘children’, Sarah ‘Blard’ Amblard, Alicia Forbes, Kendall Fitzgerald, and Miranda Wyeth.

Next, to all my friends, family, and friends who are basically family, but especially my mom and dad, my sister Lauren, my brother-in-law Jason, my nephews Camden and Chase, my niece Tenley, my Uncle Dave and Aunt Pat, my Uncle ‘Chubby’, my aunts Lucy, Mary Beth, and Carol, my partner Damien, my liver Shiva, my pack Kim, my dumpling Kate, the Lady Elizabeth, my oldest friend PJ, and lastly, the love of my life, my dog Luke Tiberius Skywalker. Thank you all for being there, but most importantly, thank you for not being there sometimes, and acknowledging when I need space or find myself in ‘Dana land’. Which, let’s be honest, is more often than not. Thank you for accepting me as I am and for your support, which for many of you comes from half a world away.

Finally, I want to thank those who have mentored me on this journey to being a scientist and who may not realise how grateful and indebted I am to have ever had the pleasure of working with them: Susan Parks, Dave Wiley, and Michael Noonan. You’re my science heroes.

Financial support

This research was supported by a University of Queensland International Living Scholarship, a University of Queensland Research Training Tuition Fee Offset, a University of Queensland Research Training Program Career Development Scholarship, and a scholarship from the American Australian Association.

BRAHSS fieldwork (2010, 2011, 2014) was funded by the E&P Sound and Marine Life Joint Industry Programme and the U.S. Bureau of Ocean Energy Management. Great Barrier Reef fieldwork (2017, 2018, 2019) was funded by the Holsworth Wildlife Research Endowment and the Ecological Society of Australia (#2017001004). East Australia tagging fieldwork (2017) was funded by the Holsworth Wildlife Research Endowment and the Ecological Society of Australia (#2017002648).

Keywords acoustic communication, clustering, communicative complexity, competitive group, flexible information, graded calls, humpback whale, intraspecific conflict, social complexity hypothesis

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 060201, Behavioural Ecology, 40% ANZSRC code: 060801, Animal Behaviour, 40% ANZSRC code: 020301, Acoustics and Acoustical Devices, 20%

Fields of Research (FoR) Classification

FoR code: 0602, Ecology, 40% FoR code: 0608, Zoology, 40% FoR code: 0203, Physics, 20%

Dedications

For my parents.

“Do. Or do not. There is no try.” – Yoda

xii

Table of Contents

Abstract ...... ii

Declaration by author ...... iv

Publications included in this thesis ...... v

Submitted manuscripts included in this thesis ...... v

Other publications during candidature ...... v

Contributions by others to the thesis ...... vi

Statement of parts of the thesis submitted to qualify for the award of another degree ...... vii

Research involving human or animal subjects ...... vii

Acknowledgements ...... ix

Financial support ...... xi

Keywords ...... xi

Australian and New Zealand Standard Research Classifications (ANZSRC) ...... xi

Fields of Research (FoR) Classification ...... xi

Dedications ...... xii

Table of Contents ...... xiii

List of Figures ...... xvii

List of Tables ...... xix

List of abbreviations ...... xxii

Chapter 1: Introduction ...... 1

1.1 Thesis Overview and Aims ...... 1

1.2 Acoustic Communication ...... 1

1.2.1 Discrete acoustic signals ...... 2

1.2.2 Graded acoustic signals ...... 3

1.2.3 Call classification methods ...... 4

xiii

1.3 Animal Social Complexity ...... 5

1.4 Communicative Complexity ...... 7

1.5 The Humpback Whale ...... 9

1.5.1 Humpback whale social behaviour ...... 9

1.5.2 Humpback whale communication ...... 11

1.5.3 Communicative complexity in humpback whales ...... 13

1.6 Thesis Outline ...... 16

1.7 Literature Cited ...... 17

Chapter 2: Humpback whale social call production reflects both motivational state and

arousal ...... 34

2.1 Abstract ...... 35

2.2 Introduction ...... 35

2.3 Methods ...... 38

2.3.1 Call repertoire ...... 41

2.3.2 Call structure...... 43

2.3.3 Call bouts ...... 43

2.3.4 Call rate...... 44

2.4 Results ...... 44

2.4.1 Call type ...... 44

2.4.2 Call structure...... 49

2.4.3 Call bouts ...... 50

2.4.4 Call rate...... 51

2.5 Discussion ...... 51

2.6 References ...... 54

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy

clustering ...... 63

xiv

3.1 Abstract ...... 63

3.2 Introduction ...... 63

3.3 Methods ...... 66

3.3.1 Aural-visual (AV) qualitative classification ...... 68

3.3.2 Classification and Regression Rrees (CART) and Random Forests (RF) ...... 68

3.3.3 ‘Soft’ clustering with fuzzy k-medoids (FKM) ...... 69

3.4 Results ...... 70

3.4.1 Aural-visual (AV) qualitative classification ...... 70

3.4.2 Classification and regression trees (CART) and random forests (RF) ...... 72

3.4.3 ‘Soft’ clustering with fuzzy k-means (FKM) ...... 74

3.5 Discussion ...... 78

3.6 References ...... 81

Chapter 4: Support for a link between the social complexity hypothesis and communicative

complexity in the humpback whale ...... 88

4.1 Abstract ...... 88

4.2 Introduction ...... 89

4.3 Methods ...... 92

4.3.1 Social states ...... 93

4.3.2 Acoustic behaviour ...... 94

4.4 Results ...... 96

4.4.1 Social states ...... 96

4.4.2 Acoustic behaviour ...... 98

4.5 Discussion ...... 103

4.6 References ...... 107

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in

the humpback whale ...... 119

5.1 Abstract ...... 119

5.2 Introduction ...... 120

5.3 Methods ...... 123

5.3.1 Behavioural data collection ...... 124

5.3.2 Acoustic data collection ...... 126

5.3.3 Statistical analysis...... 128

5.4 Results ...... 129

5.4.1 Intensity level ...... 129

5.4.2 Acoustic behaviour ...... 130

5.5 Discussion ...... 135

5.6 References ...... 138

Chapter 6: Discussion ...... 146

6.1 Introduction ...... 146

6.2 Summary of work performed ...... 147

6.3 The presumed function of humpback calls ...... 149

6.4 Humpback whales have high communicative complexity ...... 152

6.5 Humpbacks fit the SCHCC in a novel way ...... 154

6.6 Limitations and future directions ...... 155

6.7 Concluding remarks ...... 157

6.8 References ...... 158

Appendices ...... 166

Appendix 1: Humpback whale social call production ...... 166

Appendix 2: Quantifying the number of call types ...... 168

Appendix 3: A link between social and communicative complexity ...... 172

Appendix 4: Discrete and graded calls during intraspecific conflict ...... 177

Appendix 5: Animal ethics approval ...... 179

xvi

List of Figures

Figure 2.1 Map of the primary study area in southern Queensland, Australia. The majority of data collection occurred just south of Peregian Beach ...... 39

Figure 2.2 Spectrograms of the identified call types, (a) and (b) ’calf’ calls, (c) ‘low frequency’ call, (d) ‘squeak’, (e) ‘modulated’ call, (f) ‘paired croaks’, (g) ‘knock’, (h) ‘snort’, and (i) ‘spiccato’ ...... 45

Figure 2.3 Results of the k-means cluster analysis, indicating (a) two clusters of calls for female- calf pairs, (b) two clusters of calls for female-calf-escort groups, and (c) the addition of a third cluster for female-calf-multiple escort groups ...... 48

Figure 3.1 Spectrograms of the 12 low frequency call types identified by aural-visual examination: (a) ‘bop’, (b) ‘crow’, (c) ‘grunt’, (d) ‘wop’, (e) ‘moan’, (f) ‘wup’, (g) ‘paired croaks’, (h) ‘snort’, (i) ‘thwop’, (j) ‘knock’, (k) ‘spiccato’, and (l) ‘grumble’...... 71

Figure 3.2 Spectrograms of the four high frequency call types identified by aural-visual examination: (a) ‘eeaw’, (b) ‘meow’, (c) ‘squawk’, and (d) ‘squeak’ ...... 72

Figure 3.3 Spectrograms of calls from each cluster: (a) cluster one, ‘low entropy snort/knock’, (b) cluster two, ‘wup/low frequency eeaw’, (c) cluster three, ‘grumble/long snort’, (d) cluster four, ‘discrete snort’, (e) cluster five, ‘broadband bop’, (f) cluster six, ‘high frequency squeak’, (g) cluster seven, ‘squeak/high frequency eeaw’, (h) cluster eight, ‘low entropy bop’, (i) cluster nine, ‘high frequency bop’, and (j) cluster ten, ‘high entropy bop’. Purple stars indicate calls classified as ‘snorts’ during the aural-visual analysis, and blue stars indicate calls classified as ‘bops’. The placement of these calls into separate clusters highlights the ability of the fuzzy cluster analysis to differentiate calls that sound similar to a human observer ...... 75

Figure 3.4 Histogram of average typicality coefficients (TC) for each cluster from the fuzzy k- means analysis. Dashed lines indicate the typicality threshold, whereby calls and clusters with an average TC below the left dashed line are considered ‘atypical’ (graded), and calls and clusters with an average TC above the right dashed line are considered ‘typical’ (discrete) ...... 77

Figure 4.1 Composition of the discrete call type repertoire for the four social states, indicating the proportions of each call type ...... 100

xvii

Figure 4.2 Composition of the graded call type repertoire for the four social states, indicating the proportions of each call type ...... 101

Figure 5.1 Map of the study area in the Great Barrier Reef, indicating the primary survey area around Whitsunday Island. The majority of competitive groups were found to the northeast, between Whitsunday Island and Bait Reef ...... 124

Figure 5.2 Spectrograms of the call types detected in competitive groups during this study, (a) ‘discrete snort’, (b)‘thwop’, (c) ‘wop’, (d) ‘spiccato’, (e) a series of ‘paired croaks’, (f) ‘meow’, (g) a ‘song unit social sound’ followed by a graded call, and (h) a graded call ...... 128

Figure 5.3 Average call rate (calls per whale per hour) within each intensity level ...... 131

Figure 5.4 Proportions of discrete and graded call types detected in the three intensity levels ..... 132

Figure 5.5 Proportions of discrete call types detected in the three intensity levels ...... 134

Figure A3.1 Spectrograms of calls from each cluster: (a) cluster one, ‘low entropy snort/knock’, (b) cluster two, ‘wup/low frequency eeaw’, (c) cluster three, ‘grumble/long snort’, (d) cluster four, ‘discrete snort’, (e) cluster five, ‘broadband bop’, (f) cluster six, ‘high frequency squeak’, (g) cluster seven, ‘squeak/high frequency eeaw’, (h) cluster eight, ‘low entropy bop’, (i) cluster nine, ‘high frequency bop’, and (j) cluster ten, ‘high entropy bop’ ...... 172

xviii

List of Tables

Table 2.1 A list of the 12 acoustic measurements that were used in the call classification and cluster analyses along with their associated abbreviation and description ...... 41

Table 2.2 Mean ± SD of the acoustic measurements for the 8 identified call types, as well as the number of calls in parentheses. A list of the abbreviations can be found in Table 2.1...... 46

Table 3.1 Description of the measurements used for repertoire classification. Frequency measurements were logged prior to analysis ...... 67

Table 3.3 Results of the fuzzy cluster analysis indicating the number of calls in each cluster, the AV call type that was represented the most, the average typicality coefficients (mean ± SD), the percent of calls that fell above and below the typicality threshold, and the designation of graded or discrete. Any cluster with calls above the threshold was considered relatively discrete. Note that clusters contain multiple call types from the aural-visual analysis ...... 78

xix

Table 4.1 Mean ± SD of the acoustic measurements for each call type. D: discrete, G: graded ..... 95

Table 4.2 Results of the social state cluster analysis, indicating the mean ± SD of each variable and the proposed social state for each cluster. The proportion of time at the surface is measured from the tagged animal. FC: female-calf pair, FCE: female-calf-escort group, FCME: female-calf-multiple escort group ...... 97

Table 4.3 Proportion ± SE of each call type and category that comprise the four social states. Note that columns, not rows, add up to 1.0. D: discrete, G: graded ...... 98

Table 4.5 Comparison of the coefficient of variation (CV) in discrete and graded calls for the four acoustic parameters across each social state. A significant result (p < 0.05) from the modified signed-likelihood ratio test is indicated by a * next to the parameter, with the highest and lowest coefficients indicated in bold. D: discrete, G: graded, MSLRT: modified signed-likelihood ratio test ...... 102

Table 5.1 Behavioural ethogram for humpback whale competitive groups. * indicates aggressive behaviour, ** indicates highly aggressive behaviour, and * or ** indicates level of aggression is determined by whether body contact is made or presumed to be attempted ...... 125

Table 5.2 Intensity scale developed for competitive groups based on the behaviours outlined in the ethogram and observations of speed and breathing rates ...... 130

Table 5.3 Results of the generalised linear models with the probability of detecting discrete and graded calls in each intensity level in the first three columns, and the pairwise comparisons in the last three columns. An asterisk indicates statistical significance at the p < 0.05 level ...... 132

Table A1.2 Results of the k-means analysis, providing the mean and standard deviation of the acoustic parameters for each cluster for the three group compositions. Abbreviations for the variables can be found in Table 2.1. FC: female-calf pair, FCE: female-calf-escort group, FCME: female-calf-multiple escort group ...... 167

Table A2.2 Results of the fuzzy cluster analysis, indicating how many of each call type from the AV analysis were assigned to each cluster. A call was assigned to the cluster in which it had the largest typicality coefficient ...... 169

Table A2.3 Summary of the acoustic parameters (mean ± SD) by call type. The description was based on the AV sound types most prevalent in the cluster and/or the defining features of the calls ...... 170

Table A3.1 Summary of the 24 focal follows used in analysis. The number of calls excludes calls with an SNR < 5 dB and without accompanying behavioural data ...... 173

Table A3.2 Comparison of the coefficient of variation (CV) within each call type for the four acoustic parameters across each social state. A significant result (p < 0.05) is indicated by a * next to the parameter, with the highest and lowest CVs indicated in bold. ‘Meows’ were only detected once in the ‘low social complexity’ state, so no CV is available. CV: coefficient of variation, D: discrete, G: graded, MSLRT: modified signed-likelihood ratio test ...... 174

xxi

List of abbreviations

Aggregate entropy (AgEnt) Aural-visual analysis (AV) Behavioural Responses of Australian Humpback whales to Seismic Surveys (BRAHSS) Bout end criterion (BEC)

Centre frequency (FC) Classification and Regression Trees (CART) Coefficient of variation (CV) Digital acoustic recording tag (DTAG) Duration (Dur) Fast Fourier Transform (FFT) Female-calf pair (FC) Female-calf-escort group (FCE) Female-calf-multiple escort group (FCME)

Fifth centile frequency (FC05)

First quartile frequency (FQ1)

Frequency trend (FTREND) Fuzzy k-means clustering (FKM) Generalised linear model (GLM) Generalised linear mixed-effects model (GLMM)

Inter-centile bandwidth (FIC)

Inter-quartile bandwidth (FIQ) Knots (kts) Linear mixed-effects model (LMM)

Maximum frequency (FMAX)

Minimum frequency (FMIN) Modified signed-likelihood ratio test (MSLRT)

Ninety-fifth centile frequency (FC95) Out-of-bag error rate (OOB)

Peak frequency (FP) Restricted maximum likelihood (REML) Signal-to-noise ratio (SNR) Social complexity hypothesis for communicative complexity (SCHCC)

xxii

Standard deviation (SD) Standard error (SE)

Third quartile frequency (FQ3) Typicality coefficient (TC)

xxiii

Chapter 1

Introduction

1.1 Thesis Overview and Aims

Animals living in a social environment typically have frequent and diverse relationships. In order to mediate these interactions, social species often have complex vocal repertoires, including a large number of call types and high variability both between and within calls. While this concept has been validated in several taxa, not all species appear to conform to these trends. Historically, humpback whales (Megaptera novaeangliae) have been described as having a simple social system characterised by small and unstable groups. Despite this, they have evolved a complex social call repertoire, although a quantitative assessment of this complexity is lacking. Although these interactions are undoubtedly mediated using acoustic signals, how potential information is conveyed (e.g. through call types and/or through changes in the structure of the calls) is poorly understood. This is partly due to the challenges involved in analysing an acoustic repertoire characterised by high variability. This thesis investigates the potential information in the social calls of humpback whales during social interactions related to breeding, with a particular emphasis on within-call structural variability (e.g. changes in frequency, duration, or bandwidth). This variability has the potential for conveying information on the signaller’s internal motivational state or arousal level, which may be particularly important for social species or during complex social interactions. As humpback whales are thought to have a simple social system, this thesis also aims to determine the potential link between the complex communication system of this species and social interactions during behaviours associated with breeding. The methods and results presented can provide a basis for furthering our understanding of the function of animal acoustic signals and for making comparisons between the social and communicative complexity of other species. This chapter provides an introduction to the main topics that are covered throughout this thesis, including the potential information conveyed in acoustic signals and animal social and communicative complexity. It will then present information on the current knowledge of these concepts in humpback whales. Finally, it concludes with an outline of the remainder of the thesis.

1.2 Acoustic Communication

Communication between animals serves to mediate social behaviour (Ridley 1995).

1 Chapter 1: Introduction

Communicative signals provide potential information to receivers, although information is only generated once it is processed by the receiver (Seyfarth and Cheney 2003; Rendall et al. 2009; Scarantino 2010; Fischer 2011, 2013). This potential information includes attributes related to the signaller (e.g. sex or hormone levels) or the environment, and can therefore function to reduce uncertainty about subsequent events (Ey and Fischer 2009; Fischer 2011, 2013). Communication signals can take many different forms, including visual, auditory, chemical, and olfactory cues (Halliday and Slater 1983; McGregor 2005; Bradbury and Vehrencamp 2011). The production of signals has evolved to maximize efficiency, and the effectiveness of each signal modality is largely dependent on environmental factors (Ey and Fischer 2009). Vocal communication is especially useful in animals where other forms of communication are limited, such as in aquatic environments. Selection pressures related to environmental constraints have impacted both the production of vocal signals and their acoustic structure (Ey and Fischer 2009).

1.2.1 Discrete acoustic signals A signal which has less within-call variability than between-call variability is classified as ‘discrete’ (Marler and Vandenbergh 1979). Discrete calls are thus perceived as distinct categories (Marler and Vandenbergh 1979), and tend to be relatively stereotyped in acoustic structure, both between and within contexts (Green and Marler 1979; Morton 1982). It is thought that discrete calls evolved for communication in environments with restricted visual access, such as dense forest or the marine environment (Marler 1975; Ey and Fischer 2009). This is because they are clearly defined, precise, and unambiguous, even in the presence of background noise. Further, they are distinguishable even in the absence of other signal modalities (e.g. visual or olfactory) (Marler 1976; Green and Marler 1979). Discrete calls are therefore often used for specific functions, such as territorial and long-distance signals [e.g. guenons, genus Cercopithecus (Marler 1975); black and white colobus, Colobus quereza (Marler 1975); spotted hyenas, Crocuta crocuta (East and Hofer 1991a, b)], contact or group-specific calls [e.g. greater spear-nosed bats, Phyllostomus hastatus (Wilkinson and Wenrick Boughman 1998); killer whales, Orcinus orca (Ford 1989); southern right whales, Eubalaena australis (Clark 1990); North Atlantic right whales, Eubalaena glacialis (Parks and Tyack 2005; Parks and Clark 2007)], and alarm calls [e.g. Campbell’s monkeys, Cercopithecus campbelli (Ouattara et al. 2009; Keenan et al. 2013); dwarf mongooses, Helogale parvula (Beynon and Rasa 1989; Manser et al. 2014); Japanese great tits, Parus major minor (Suzuki 2014)]. Because of their stereotypy, discrete calls are often associated with conveying fixed information that is related to the physical attributes of the producer (Marler 1961, 1977; Green and Marler 1979). This type of information is considered fixed, or static, because it is relatively stable

Chapter 1: Introduction over time and independent of both environmental and social context (Green and Marler 1979). Fixed information includes cues related to sex [e.g. giant pandas, Ailuropoda melanoleuca (Charlton et al. 2009); manatees, Trichechus manatus manatus (Sousa-Lima et al. 2008); spotted hyenas (Mathevon et al. 2010); yellow-bellied marmots, Marmota flaviventris (Blumstein and Munos 2005)], age class [e.g. humans (Endres et al. 1971); North Atlantic right whales (McCordic et al. 2016; Root-Gutteridge et al. 2018); chacma baboons, Papio cynocephalus ursinus (Fischer et al. 2004); red deer, Cervus elaphus (Reby and McComb 2003); manatees (Sousa-Lima et al. 2008)], and individual identity [e.g. bottlenose dolphins, Tursiops truncatus (Caldwell and Caldwell 1965); black-capped chickadees, Poecile atricapillus (Christie et al. 2004); tree shrews, Tupaia belangeri (Schehka and Zimmermann 2009); fallow deer, Dama dama (Reby et al. 1998); green frogs, Rana clamitans (Bee et al. 2001); hamadryas baboons, Papio hamadryas ursinus (Rendall 2003); African elephants, Loxodonta africana (Soltis et al. 2005)]. Fixed signals can be useful during complex social interactions, particularly those that are agonistic. As conflict can be costly, it is beneficial for animals to convey information on fitness and fighting ability, both before direct confrontation and as conflict escalates (Maynard Smith and Price 1973; Maynard Smith 1974). Fixed information related to body size is particularly common [e.g. toads, Bufo bufo (Davies and Halliday 1978); hamadryas baboons (Pfefferle and Fischer 2006); Blanchard’s cricket frogs, Acris crepitans blanchardi (Wagner, Jr 1992); Montezuma oropendolas, Psarocolius montezuma (Price et al. 2006); scops owls, Otus scops (Hardouin et al. 2007); fallow deer (Vannoni and McElligott 2008; Charlton and Reby 2011)]. In terms of competition, a larger animal will generally win in a physical contest against a smaller animal (Archer 1988). Communicating body size vocally can allow individuals to assess conspecifics remotely and potentially eliminate the need for direct confrontation (Zahavi 1982; Maynard-Smith and Harper 2003). This can have advantages for both animals by reducing the risk of injury for the smaller animal and wasted energy for the larger animal. Red deer stags for example, which roar during the breeding season, roar most often in situations where they are likely to be challenged, and when an opponent is closer than average (Clutton-Brock and Albon 1979). Further, red deer are capable of adjusting the features of their calls to exaggerate their size (Fitch and Reby 2001; Reby et al. 2005). These results suggest that roaring serves to provide individuals with an opportunity to assess opponents without the need for physical confrontation.

1.2.2 Graded acoustic signals In contrast to discrete calls, a signal with considerable within- and between-call variability is classified as ‘graded’ (Marler and Vandenbergh 1979). Graded calls can be thought of as falling

Chapter 1: Introduction along a continuum rather than into distinct categories like discrete calls (Marler 1961, 1975, 1976; Marler et al. 1992; Hauser 1996). The result is a vocal continua of graded calls which are similar in acoustic structure (Marler and Vandenbergh 1979), although receivers may be able to assign graded calls to discrete categories (Fischer 2006). Graded calls are thought to have evolved primarily for close-range communication because they can easily become indistinguishable after transmission (Marler 1976; Green and Marler 1979; Marler and Vandenbergh 1979). Further, the high variability can provide a large amount of potential information about the internal motivational state or arousal of the signaller, which may be more important during close intraspecific interactions (Marler 1961, 1976; Morton 1977; Manser 2010; Briefer 2012; Fischer et al. 2017b). This includes intraspecific conflict, where graded calls can be correlated with the signaller’s intent or level of aggression (Morton 1982). For example, gradation in the ‘tremolo’ call of the common loon (Gavia immer) is associated with the probability of attack (Barklow 1979; Morton 1982). Greater false vampire bats, Megaderma lyra (Bastian and Schmidt 2008) produce ‘aggressive calls’ during agonistic interactions between conspecifics, and the intensity of the interaction is reflected in the acoustic parameters of this call. ‘Aggressive calls’ are also produced in a wide variety of other taxa, including birds [e.g. Carolina wrens, Thryothorus ludovicianus (Morton 1982); corncrakes, Crex crex (Ręk et al. 2011)], anurans [e.g. gray tree frogs, Hyla versicolor (Reichert and Gerhardt 2013); Blanchard’s cricket frogs (Wagner 1989); spring peepers, Pseudacris crucifer (Schwartz 1989)], and Orthopteran insects (Alexander 1961). Graded acoustic signals can also be important during breeding interactions as a way for a receiver to remotely assess the dominance status, arousal, and/or reproductive success of a potential competitor [e.g. red deer (Clutton-Brock and Albon 1979; Reby and McComb 2003); chipping sparrows, Spizella passerina (Liu 2004); rock hyrax, Procavia capensis (Koren and Geffen 2009); chacma baboons (Fischer et al. 2004)]. This may be particularly important as conflict escalates beyond threats and displays, especially considering smaller animals with high motivation to fight are sometimes able to successfully dominate larger animals (Wagner 1989; Kotiaho et al. 1999; Hofmann and Schildberger 2001).

1.2.3 Call classification methods While within-call variability can provide a multitude of information, it naturally complicates classification methods. As such, repertoire size may be difficult to establish in species with considerable gradation between signal types (Green and Marler 1979; Hammerschmidt and Fischer 1998; Freeberg et al. 2012; Manser et al. 2014; Garland et al. 2015). Many classification techniques have relied heavily on human observers. Although these types of subjective analyses are successful

Chapter 1: Introduction in some studies, they are prone to observer bias and are generally not considered to be suitable for classifying graded calls (Janik 1999; Burghardt et al. 2012; Kershenbaum et al. 2016). More quantitative methods provide a welcome level of standardisation and objectivity, however most methods of automatic classification depend on techniques which are not well suited to account for gradation in call types (e.g. discriminant function analysis, hierarchical or k-means cluster analysis) (Marler 1975; Rekdahl et al. 2013; Wadewitz et al. 2015). The underlying assumption of these analyses is that calls are discrete, and can therefore be discretely categorised. With graded call types this assumption is violated, which can lead to inconsistent results even within a single species. For example, the number of call types in the repertoire of African elephants, which contains several graded call types, depends heavily on the choice of analysis (Leong et al. 2003; Soltis et al. 2005; Wood et al. 2005). There are thus challenges associated with establishing a definitive number of call types in repertoires that include graded calls (Green and Marler 1979; Hammerschmidt and Fischer 1998; Freeberg et al. 2012; Manser et al. 2014; Garland et al. 2015). However, using classification techniques that measure the amount of gradation in a call can provide information on the variation in the caller’s internal state, which may be as biologically-relevant as the use of specific call types (Fischer et al. 2017b), particularly for highly social species.

1.3 Animal Social Complexity

The social behaviour of animals involves the interaction between members of the same species. However the degree of complexity in these relationships is variable (Freeberg et al. 2012). Social complexity is often not well-defined, and may be characterized using different metrics, both currently and historically (Freeberg et al. 2012; Bergman and Beehner 2015; Fischer et al. 2017a; Allen 2019; Kappeler et al. 2019). While group size and stability are often considered standard metrics, they may only be applicable for some species (Freeberg et al. 2012; Silk et al. 2013; Bergman and Beehner 2015; Fischer et al. 2017a; Kappeler 2019; Peckre et al. 2019). Recent research has generally agreed that additional metrics should include the social roles of individuals, such as the presence of hierarchies, and the type of mating system (Freeberg et al. 2012; Allen 2019; Kappeler 2019; Peckre et al. 2019). Overall, although often poorly defined (Freeberg et al. 2012; Allen 2019), it is clear that a complex social system is one in which there are more parts, whether it be a larger number of interactions, greater diversity in these interactions, and/or relationships with more individuals. However, a universal index of social complexity is likely not appropriate, and different indices should apply to different taxa (Kappeler et al. 2019). In general, a species with a simple social system tends to have limited conspecific interaction, perhaps only in a few contexts (i.e. breeding or feeding), and few repeated interactions

Chapter 1: Introduction with the same individuals over time. For example, the slender mongoose (Galerella sanguinea) is primarily solitary, with pairs coming together only for reproduction (Manser et al. 2014). Further, the female is tasked with raising any offspring alone. In contrast, species with a complex social system tend to live in large, stable, and permanent societies where repeated interactions with the same individuals are common (Kappeler 2019). These societies are usually characterised by cooperative breeding, where group members assist in raising young, a feature often linked to kinship (Kappeler 2019). The potential benefits of this are improved reproductive success, increased inclusive fitness, and reduced fighting (Hamilton 1964; Clapham 1996). Bottlenose dolphins along the coast of Florida exhibit site fidelity and form ‘communities’, which are stable over time and across generations (Wells 2003). Dolphins within a community interact frequently, sharing ranges and social associates (Wells and Scott 1999). Elephants (family Elephantidae) are another well- known socially complex species that has been documented to form multiple and intricate social relationships (Payne 2003). Females in particular have a social network that, like bottlenose dolphins, spans multiple families and generations. Both of these species exhibit a high degree of behavioural individual variability that also contributes to the development of a highly complex social system (de Waal and Tyack 2003). A species characterised by a fluid or unstable group structure, or those which are only social during certain periods, may be more difficult to assess (Freeberg et al. 2012). In these species, it could be that characteristics associated with social complexity could be linked to specific behavioural interactions rather than sociality as a whole. For example, some terrestrial species which are considered asocial do display some cooperative feeding strategies, which is considered a metric of a more complex social system (Freeberg and Krams 2015). In addition, some species of pinnipeds (families Otariidae, Phocidae, and Odobenidae) only live in high densities during the breeding season. However this high density coupled with their polygamous mating system and breeding site fidelity indicate a higher social complexity, at least during the breeding season (Stirling and Thomas 2003). Naturally, these social characteristics are also correlated with an increased opportunity and motivation to engage in intraspecific conflict (Campagna 2009). One of the most common sources of agonistic interaction involves competition by sexually mature males over access to females, typical in polygamous mating systems (Campagna 2009). Conflict can be costly though, often involving high energy expenditure and injuries, which may equate to long-term fitness consequences (Huntingford and Turner 1987). In order to minimise these costs, it is beneficial for individuals to communicate information that may mediate these risky interactions (Maynard Smith and Price 1973; Maynard Smith 1974). This may be particularly true for species

Chapter 1: Introduction with complex social systems where they come into repeated contact with the same individuals (e.g. the ‘dear enemy’ phenomenon) (Temeles 1994).

1.4 Communicative Complexity

Lamarck (1809) and Darwin (1872) proposed that social animals have a greater need for, and dependence on, within-species communication than animals that are relatively more solitary. Highly social species, therefore, should have a rich and diverse communication system in order to convey a wider range of information, including individual or group identity, behaviour, and motivation (Freeberg et al. 2012). This concept is known as the ‘social complexity hypothesis for communicative complexity’, or the SCHCC (Freeberg et al. 2012). The SCHCC proposes that animals living in complex social environments have a greater need for complex communication systems in order to relay an enormous amount of information and mediate intricate social interactions (Freeberg 2006; Freeberg et al. 2012; Peckre et al. 2019). Indeed, as per the SCHCC, many highly social species demonstrate a high amount of communicative complexity [e.g. killer whales (Bain 1986; Ford 1989; Rehn et al. 2007, 2011); Barbary macaques, Macaca sylvanus (Hammerschmidt and Fischer 1998; Fischer and Hammerschmidt 2002); beluga whales, Delphinapterus leucas (Garland et al. 2015); ants, Camponotus socius (Hӧlldobler 1999); red- capped mangabeys, Cercocebus torquatus (Bouchet et al. 2013); pilot whales, Globicephala melaena (Busnel and Dziedzic 1966; Taruski 1979); banded mongooses, Mungos mungo (Manser et al. 2014); honeybees, genus Apis (von Frisch 1974)]. Like social complexity, communicative complexity is also often not well-defined, but can be generally thought of as a communication system with more parts. This can refer to both the number and types of elements in the repertoire, as well as the higher-order relationships between these elements (Fischer et al. 2017b). Measures of the former can manifest in many ways, one being the number of signals or displays in a species’ repertoire (Griebel and Oller 2008; Oller and Griebel 2008; Freeberg et al. 2012; Pika 2017). In terms of acoustic communication, a socially complex species may therefore have a larger number of acoustic call types compared to a species with a simple social system [e.g. North American wrens, family Troglodytidae (Kroodsma 1977); non- human primates (McComb and Semple 2005; Bouchet et al. 2013); seals, family Phocidae (Stirling and Thomas 2003)]. Additional metrics of communicative complexity have been relatively understudied historically, but can include the presence of nonlinear phenomena (Freeberg et al. 2012; Pika 2017). ‘Vocal nonlinearities’ are related to the physiology of the vocal production mechanism but independent of the fundamental frequency (Fitch and Hauser 2002; Fitch et al. 2002; Riede et al. 2004). Typical vocal nonlinearities include biphonation (i.e. the generation of a

Chapter 1: Introduction second frequency unrelated to the first), frequency jumps, deterministic chaos (i.e. broadband noise), entropy (i.e. disorder), and subharmonics. Nonlinear phenomena are proposed to generate complex vocalisations (Fitch et al. 2002) which may increase individual distinctiveness (Riede et al. 2004) or provide cues to arousal (Blumstein and Chi 2012; Briefer 2012; Blesdoe and Blumstein 2014). They may occur both in the overall repertoire or within individual sound units. For example, within a repertoire, high entropy in the sequence of calls is related to increased communicative complexity (Freeberg and Lucas 2012; Freeberg et al. 2012; Kershenbaum 2014; Fischer et al. 2017b; Kershenbaum et al. 2018). Within an individual call, entropy is related to the frequency range and energy distribution of a sound, with high entropy sounds having a mixture of frequencies (Kershenbaum et al. 2018). In contrast, low entropy sounds are frequently more tonal (Fitch et al. 2002; Briefer 2012). Additionally, there is evidence that vocal nonlinearities may provide information on age, with young animals presenting more nonlinearities in their calls [e.g. North Atlantic right whales (Root-Gutteridge et al. 2018); African elephants (Stoeger et al. 2011); sperm whales, Physeter catodon (Watkins et al. 1988)]. Lastly, the presence of a high level of both between- and within-call variation has been proposed as a metric of communicative complexity (Freeberg et al. 2012; Manser et al. 2014; Fischer et al. 2017b). As some graded calls can be perceived as discrete categories (Fischer 2006), potential information can lie with the call type itself as well as the flexible features that potentially convey information on the internal state of the signaller. A larger number of graded call types in a repertoire therefore offers a greater number of options for potential information to be perceived (Peckre et al. 2019; Rebout et al. 2020). However, quantitative assessments of communicative complexity and the variability in acoustic repertoires is lacking for most taxa, including marine mammals. Many marine mammals have a complex communication system which research suggests involves the use of both discrete and graded calls [e.g. bottlenose dolphins (Herzing 1996; Luís et al. 2016); killer whales (Bain 1986; Ford 1989; Rehn et al. 2007, 2011); harbour seals, Phoca vitulina (Hanggi and Schusterman 1994; Nicholson 2000; Hayes et al. 2004); Australian fur seals, Arctocephalus pusillus doriferus (Tripovich et al. 2008); beluga whales (Garland et al. 2015); pilot whales (Busnel and Dziedzic 1966; Taruski 1979); Weddell seals, Leptonychotes weddellii (Stirling and Thomas 2003; Collins et al. 2005); humpback whales, Megaptera novaeangliae (Silber 1986; Dunlop et al. 2008; Stimpert et al. 2011; Fournet et al. 2015)]. To date, little data exist on the degree to which social and communicative complexity are related in marine species (but see May-Collado et al. 2007), and what roles discrete and graded signals play in the communication systems of these animals.

Chapter 1: Introduction

1.5 The Humpback Whale

The order Cetacea includes all species of whales, dolphins, and porpoises, and is subdivided into the suborder Odontoceti (toothed whales), and Mysticeti (baleen whales). Humpback whales belong to the latter suborder, and are a medium sized whale growing to over 15 metres in length and weighing up to 45 tons (Chittleborough 1965; Lockyer 1976). There is a slight sexual dimorphism, with females being on average one to two metres longer than males (Chittleborough 1955). They are a cosmopolitan species, well known for their annual migrations between high latitude feeding grounds in the summer and low latitude breeding grounds during winter months (Chittleborough 1965; Dawbin 1966; Clapham 1996). On the breeding grounds and during most of the migration, humpbacks fast, relying on fat reserves (Chittleborough 1965; Clapham 1996). Although birth intervals are typically 2-3 years in most populations, annual births are reported in some areas (Clapham and Mayo 1990; Glockner-Ferrari and Ferrari 1990). Whaling data also indicated that some lactating females were also ovulating, providing further evidence of at least some animals undergoing postpartum oestrus (Chittleborough 1957). In the Southern Hemisphere, humpback whales feed in the waters around Antarctica, which is divided into six feeding areas (Areas I – VI) (IWC 2014). They then undergo one of the longest documented mammalian migrations to tropical breeding grounds (Stevick et al. 2010), where they are divided into seven breeding stocks based on natal site fidelity (Groups A – G) (IWC 2014). Two populations are centred around Australia – breeding stock D, which summers in Area IV (60oE – 120oE) and migrates up the coast of western Australia, and breeding stock E, which summers in Area V (120oE – 170oE) and migrates along the coasts of eastern Australia and New Zealand as well as across open ocean to their breeding grounds (Paterson and Paterson 1984). Group E can be further divided into populations based around separate breeding grounds – E1 on the Great Barrier Reef in eastern Australia; E2 in the waters of New Caledonia; and E3 around the islands of Tonga (IWC 2014). The Group E1 breeding grounds are poorly defined (Smith et al. 2012), and relatively unexplored compared to breeding areas for other populations of humpback whales, like those in the Atlantic Ocean.

1.5.1 Humpback whale social behaviour Similar to other baleen whales, humpback whales are typically classified as having a relatively simple social system (Berta and Sumich 1999; May-Collado et al. 2007). This is in part because they do not live in large, permanent, or stable groups (Clapham 1993b, 1996). Rather, humpback whale social groups tend to be small and unstable on both the feeding and breeding grounds, and while on migration (Whitehead 1983; Baker and Herman 1984; Mobley, Jr. and

Chapter 1: Introduction

Herman 1985; Clapham et al. 1992; Clapham 1993b, 1996, 2000; Mattila et al. 1994; Valsecchi et al. 2002; Ramp et al. 2010). The formation of social groups does not appear to be based on kinship, with related animals no more likely to associate than non-related individuals (Clapham 1993b; Pomilla and Rosenbaum 2006). This has been attributed to the fact that kin recognition is unlikely as females only give birth to a single offspring, raise calves alone, and calves are weaned before the birth of any siblings (Clapham 1996). Social groups instead are formed typically as the result of cooperative feeding to increase foraging capacity (Jurasz and Jurasz 1979; Clapham 1993b, 1996) and as associations related to breeding behaviour (Tyack and Whitehead 1983; Baker and Herman 1984; Clapham et al. 1992; Clapham 1996). Social behaviour that is associated with mating occurs primarily on the breeding grounds, but is also noted on the migration corridor to and from these areas (Brown and Corkeron 1995; Smith et al. 2008; Dunlop 2016; Dunlop and Noad 2016). Presumed mating strategies in the humpback include lone males singing, males 'escorting' a female and her calf if she is lactating, and large competitive groups of multiple males engaging in physical competition for access to a female (Glockner and Venus 1983; Clapham 1996). A solitary female-calf pair may have few social interactions with conspecifics, however the number and diversity of social interactions increases with the addition of escorts to a group. The nature of the relationships between group members also changes. For example, a single male escort can have a variety of effects on a female, with or without a calf, that indicate a range from tolerant and passive to intolerant and aggressive (Jones 2010). The addition of multiple escorts and the formation of competitive groups increases not only the number of animals, and therefore the number of potential social interactions, but the arousal level of the group (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Clapham et al. 1992; Clapham 1996; Felix and Novillo 2015). Humpback whale competitive groups generally contain three or more adults and have a definite structure (Tyack and Whitehead 1983). The nuclear animal can be identified by its central location in the group and a consistent close proximity to another animal, labelled the principal or primary escort. Other adults in the group are considered secondary escorts. Nuclear animals are almost always female and may or may not be accompanied by a calf, while other group participants are generally male (Baker and Herman 1984; Clapham et al. 1992; Brown and Corkeron 1995). The nuclear female is often unresponsive to approaches of other whales, however the primary escorts defend their close proximity to the female from the challenges of other escorts. In large and highly active groups, the composition and dynamic changes often, with principal escorts and secondary escorts changing positions and roles frequently (Tyack and Whitehead 1983; Clapham et al. 1992).

Chapter 1: Introduction

Competitive groups of humpback whales can exhibit a wide range of behaviours and intensity levels (Baker and Herman 1984). ‘Low intensity’ groups are characterised by lower aggression and arousal levels. There is typically no direct physical contact between group members, and instead they rely more on threats and displays (i.e. blowing bubble streams, jaw clapping, and chasing behaviour) (Darling 2001). More ‘moderate intensity’ groups are characterised by ‘intermediate’ aggressive behaviours, most commonly ‘head lunges’ (Baker and Herman 1984). ‘High intensity’ competitive groups display behaviours associated with higher levels of aggression, including ‘body thrashes’, ‘tail lashes’, and direct body contact (Tyack 1981; Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Darling 2001). Animals in high intensity competitive groups also display injuries, with one male recorded dead following observations of competitive behaviour (Pack et al. 1998). They also tend to move more erratically and have elevated respiration rates (Tyack and Whitehead 1983; Silber 1986; Clapham et al. 2008). While competitive groups are undoubtedly related to breeding, no mating has ever been observed (Clapham 2000; Pack et al. 2002). As such, the exact function of these groups is unclear, and particularly the function of the complex relationships between the animals involved (Felix and Novillo 2015). Most of the behaviour during competitive groups is agonistic, however there are reports of male humpback whales possibly working cooperatively during competitive behaviour (Clapham et al. 1992; Felix and Novillo 2015). There are also reports of multiple females in some competitive groups, which could support the idea of males cooperating (Pomilla and Rosenbaum 2006; Felix and Novillo 2015). Alternatively, this could suggest female cooperation to decrease male harassment or female intraspecific competition (Pomilla and Rosenbaum 2006). Lastly, all- male competitive groups have been reported (Clapham et al. 1992; Brown and Corkeron 1995). These groups are proposed to potentially function in dominance sorting (Clapham et al. 1992), although this may only be useful if individuals encounter each other frequently (Clapham 1993a). Overall, competitive groups are a complex social interaction that, although superficially well described, remain poorly understood.

1.5.2 Humpback whale communication Male humpback whales sing long, stereotyped songs (Payne and McVay 1971; Winn and Winn 1978). While songs are produced primarily on the breeding grounds, they may be heard to a lesser extent along the migration route and on the feeding grounds (Mattila et al. 1987; Charif et al. 2001; Garland et al. 2013). Despite extensive research, the exact function of humpback song is unknown (Payne and McVay 1971; Tyack 1981). As songs are a male display, and predominantly produced during the breeding season, it is likely that they serve a reproductive function (Winn and

Chapter 1: Introduction

Winn 1978). Possible functions of humpback whale songs are that they serve as a sexual display of fitness to females (Tyack 1981; Chu and Harcourt 1986; Smith et al. 2008); a male-male signal to mediate aggressive or cooperative social interactions (Winn and Winn 1978; Tyack 1981; Darling et al. 2006); a spacing mechanism (Frankel et al. 1995); prospecting for females (Smith et al. 2008); and/or a method of dominance sorting (Darling and Bérubé 2001). It is likely that song serves multiple functions however (Herman 2017), and may be a ‘multi-message’ signal similar to the songs of birds (Catchpole and Slater 2008). Humpback whales also produce non-song vocalisations, which include both surface generated sounds (e.g. breaches, tail slaps) and social vocalizations, or ‘calls’ (Silber 1986; Dunlop et al. 2007, 2008, 2010; Fournet et al. 2015). In contrast to song, calls are produced by all age and sex classes (Winn et al. 1979; Zoidis et al. 2008). In addition, they are diverse, produced in short bouts, and do not follow any particular pattern (Rekdahl et al. 2015). Additionally, calls are often produced at lower source levels than song, measured at 158 dB re 1 μPa @ 1 m during migration (Dunlop et al. 2013) and 137 dB re 1 μPa @ 1 m while foraging (Fournet et al. 2018b). Source levels are also reduced in lone female-calf pairs, which were recorded to have a median source level of 146 dB re 1 μPa on the migration corridor in east Australia (Indeck 2020). In contrast, song in the same study area was shown to have a range between 154 dB re 1 μPa @ 1 m and 171 dB re 1 μPa @ 1 m depending on the song unit and frequency (Girola et al. 2019). Overall, this means that, while songs can be detected over 10 km, calls can only be heard for approximately 4 km in the same environment (Dunlop et al. 2013; Dunlop 2018), indicating that they are likely used for close-range communication. Calls are produced in a wide variety of habitats and social contexts. On the feeding grounds in both the North Atlantic and Pacific, some calls have been linked to specific behaviours, including various types of cooperative foraging (Jurasz and Jurasz 1979; D’Vincent et al. 1985; Thompson et al. 1986; Cerchio and Dahlheim 2001; Sharpe 2001; Parks et al. 2014). While on migration, the use of calls continues to be pervasive across all social group compositions (Dunlop et al. 2008). While no calls on migration have been found to be unique to any context, certain calls are associated with some contexts more frequently than others. For example, ‘grunts’, ‘groans’, and ‘barks’ were used almost exclusively in joining groups, suggesting they may function in social integration (Dunlop et al. 2008). Another call commonly heard in the context of joining whales were ‘song unit social sounds’. These calls were identical to units present in the song, but they were used outside of their normal production in song (i.e. not patterned and structured). Aside from an increased use when joining, they were also produced in higher rates in lone males and in multiple-animal groups. They

Chapter 1: Introduction are therefore proposed to function similarly to song but in the context of within-group communication (Dunlop et al. 2008). Lastly, on the breeding grounds, calls seem to serve similar functions as on migration, including the coordination of group behaviours. Vocal behaviour in competitive groups on the breeding grounds is not well documented, however calls used by these groups in Hawaii were proposed to convey information on aggression (Silber 1986). In two separate experiments on the breeding grounds in Hawaii, playbacks of calls to humpback whales demonstrated that lone males and pairs of adults responded aggressively (measured by whether or not they ‘charged’ or rapidly approached the boat) when exposed to calls recorded from competitive groups (Tyack 1983; Mobley, Jr. et al. 1988). Conversely, groups containing a female and calf, and groups of more than three individuals, did not approach or moved away from the sound source during playbacks. Overall, this evidence suggests that the calls of competitive groups may convey potential information on aggression and arousal.

1.5.3 Communicative complexity in humpback whales Humpback whales do not fit the SCHCC in the way that social complexity has historically been defined because of their small group size and unstable social structure. However, they have been documented in a variety of complex social interactions where a more complicated communication system may be beneficial. For example, while social groups are generally short- term and unstable, some stable social associations have been recorded on the feeding grounds (Weinrich 1991; Clapham 1993b), with some associations occurring for up to six consecutive years (Ramp et al. 2010). Brown and Corkeron (1995) also demonstrated non-agonistic and occasionally cooperative social relationships among male humpback whales on migration, suggesting at least short-term stable bonds. Additionally, humpback whales are well known for forming large feeding and breeding aggregations, albeit temporarily. These aggregations function to increase foraging efficiency through cooperative foraging strategies (D’Vincent et al. 1985; Sharpe 2001; Wiley et al. 2011; Parks et al. 2014), and presumably as a mating strategy during the breeding season in the form of competitive groups (Tyack and Whitehead 1983; Silber 1986; Clapham 1996). Although not conforming to the traditional definitions of social complexity, it is clear that the frequency, number, diversity, and instability of relationships indicates a higher level of sociality in humpback whales. In turn, during these social interactions, humpback whales may benefit from conveying a multitude of potential information, including levels of aggression, submission, intent, physical features like body size and maturity, and/or possibly individuality.

Chapter 1: Introduction

There is substantial evidence to suggest that humpback whales have a complex communication system, likely in order to communicate this diverse range of potential information. For example, male song during the breeding season is arguably the most complex song in any animal, both because of its structure and its mode of transmission. Humpback whale song is hierarchically structured, composed of single sound units which are repeated in a predictable pattern to form a phrase (Payne & McVay 1971, Payne & Payne 1985). The repetition of phrases then forms a theme, and anywhere from 5 to 9 themes are repeated stereotypically to produce a song. The degree of complexity within a song, measured by the size and variety of sounds, can be highly variable (Allen et al. 2018, 2019). This is because, although males in a breeding population all sing the same song, this song differs between populations and years (Winn and Winn 1978; Payne et al. 1984; Payne and Payne 1985; Noad et al. 2000; Cerchio and Dahlheim 2001; Darling et al. 2014). Most changes are gradual, such as slight variations in the duration of the song, changes in the composition and presentation of themes, and changes in song structure such as theme modifications. These changes are classified as ‘cultural evolutions’ (Noad et al. 2000), and involve an increase in song complexity as the song evolves (Allen et al. 2018). In some years however, significant changes do occur. During these ‘cultural revolutions’, an entire population will sing an entirely new song from the previous year, a process which is undocumented in the vocal behaviour of other animals (Noad et al. 2000). In the South Pacific, changes in song spread across humpback populations unilaterally from west to east (Noad et al. 2000; Garland et al. 2011, 2013; Allen et al. 2018). Song change is thought to occur through social learning and horizontal cultural transmission, spreading amongst individuals within and between populations (Payne et al. 1984; Payne and Payne 1985; Garland et al. 2011). Vocal learning has been associated with a high level of communicative complexity, and in particular ‘lexical learning’, which involves the acquisition of new vocalisations (Ruch et al. 2018). This type of vocal learning can potentially lead to an increasing number of vocal signals in an individual’s repertoire, and thus increasing complexity, over its lifetime (Peckre et al. 2019). Evidence suggests that the acoustic repertoire of non-song calls is also highly complex. For example, they possess a large number of call types that are ‘stable’ (i.e. present across years), but also call types that are ‘inconsistent’ (i.e. only found in one or two years) (Dunlop et al. 2007; Rekdahl et al. 2013). East Australian humpbacks in particular also use song units in social contexts (Dunlop et al. 2007; Rekdahl et al. 2013). Overall, the repertoire for this population has been reported to be between 34 and 46 different call types. This is in stark contrast with other baleen whales, which typically have relatively few well described call types (2-8) [e.g. blue whale, Balaenoptera musculus (McDonald et al. 2001); fin whale, Balaenoptera physalus (Watkins et al.

Chapter 1: Introduction

1987; Širović et al. 2015); southern right whale (Clark 1982); North Atlantic right whale (Parks and Tyack 2005; Parks et al. 2011); gray whale, Eschrichtius glaucus (Cummings et al. 1968; Burnham et al. 2018)]. Aside from a large number of call types, the repertoire appears to include both discrete and graded calls (Jurasz and Jurasz 1979; D’Vincent et al. 1985; Cerchio and Dahlheim 2001; Dunlop et al. 2008; Parks et al. 2014; Fournet et al. 2015, 2018a; Dunlop 2017). Further, calls may convey flexible information related to motivational state or arousal. In migrating humpback whales, Dunlop et al. (2008) described the social sounds recorded in competitive groups of humpback whales off of east Australia. They identified ‘presumed underwater blows’ as common in situations where increased levels of aggression would be expected, such as in female-calf-multiple escort groups and in groups of more than two adults. As presumed underwater blows were found to be low-frequency, broadband sounds, these findings follow the trends of vocal signals in aggressive contexts in many terrestrial species (Morton 1977; Briefer 2012). Additionally, high-frequency sounds labelled ‘cries’ were also heard in these contexts. It was suggested that these social vocalisations were ‘appeasement’ or ‘fear’ signals. This could be due to shifts in group hierarchies, including the displacement of escorts, where animals may need to communicate whether or not they are aggressive and intend to fight (Dunlop et al. 2008). More recently, Dunlop (2017) tested social sounds and behavioural data from migrating humpback whales for further evidence that they followed these trends. The study found that social sounds formed clusters based on the measured acoustic parameters that were known to vary with motivation and arousal in terrestrial species (e.g. peak frequency, bandwidth, and duration). Two of the clusters represented acoustic ‘endpoints’. One endpoint cluster contained more high frequency tonal sounds and the other contained sounds with comparatively lower frequencies and large bandwidths. These clusters corresponded to ‘fear/appeasement’ and ‘aggressive’ endpoints (Morton 1977). Additionally, the signals matched the context in which they were produced. Groups which did not exhibit behaviours and group structures indicative of ‘competitive groups’ used almost exclusively signals that could be classified as ‘low arousal’ based on the prevailing terrestrial model (low in frequency and unmodulated). Larger groups exhibiting high-arousal behaviours such as erratic changes in speed and course used comparatively more signals that could be classified as ‘aggressive’ and ‘fearful’ (Morton 1977; Briefer 2012). Overall, the large number of call types, their instability, and the mix of discrete and graded signals is evidence for a high level of communicative complexity. However, previous research has concentrated on developing simple call repertoires, with few taking into consideration the behavioural context of signal production and none accounting for the degree of complexity present

Chapter 1: Introduction

(i.e. the amount of gradation in the repertoire). This thesis presents the first research dedicated to associating the calls of humpback whales with group behaviour on a fine scale in an attempt to determine whether a graded continuum exists in humpback calls that corresponds to levels of motivation and arousal, and how this might relate to signal function and communicative complexity.

1.6 Thesis Outline

Acoustic signals are undoubtedly important during the social interactions of humpback whales, particularly during breeding interactions. However, it is unknown how calls are used during this time, and if their use follows the trends observed in terrestrial species. Chapter 2 begins by testing which aspects of humpback whale acoustic behaviour were most important in communicating changes in motivation and arousal. To do this, statistical comparisons were made between the vocal behaviour of migrating female-calf pairs with and without male escorts to test for a correlation between social interactions (i.e. an increase in the number of escorts) and communicative behaviour. Specifically, whether they primarily use changes to their call repertoire (i.e. context specific calls), to call features (i.e. frequency, duration) and/or changes to the number of calls produced (i.e. call rate, the use of bouts). Humpbacks have a wide and variable repertoire of call types. However, previous classification methods have only aimed at discovering the number of call types (between-call variation), not the level of gradation (within-call gradation). Based on the importance of call types in Chapter 2, it was determined that a more dedicated analysis of the repertoire was needed. The same data from Chapter 2 were run through a fuzzy k-means cluster analysis, an approach not previously applied to the humpback whale acoustic repertoire. This analysis forms clusters of calls based on the similarity of acoustic features. Further, it also allows for the assessment of calls as relatively discrete or graded, and therefore more likely to contain fixed or flexible information, respectively. The resulting clusters of calls were considered call types, and provided the basis for the call types used in the remaining chapters. As prior research has concentrated solely on classifying call types, subsequent analyses have only been able to loosely describe the possible function of calls based on context without also considering the biological significance of discrete and graded calls. Chapter 4 uses the classification results of Chapter 3 to explore the function of calls based on the possible complexity of the social state in which they are used. First, the social states were established quantitatively using a cluster analysis, which incorporated data such as group composition (i.e. presence of escorts), speed of travel, course variation, and the number of behaviours. Next, calls were assigned to the social state

Chapter 1: Introduction in which they were detected, allowing for a comparison of the context of calls. Lastly, how acoustically variable each call type was in each social state was assessed to determine what effect the complexity of the social state had on the variation of acoustic features. The social state that was considered one of the most socially complex in Chapter 4 was the competitive group. On migration, social interactions involving breeding behaviour are still observed, but the intensity of these interactions and the overall level of aggression of these groups is lower than on the breeding grounds. The data for Chapter 5 were therefore collected on the breeding grounds in the Great Barrier Reef in order to investigate the use of discrete and graded call types during intraspecific conflict, a topic relatively well studied in terrestrial taxa but poorly understood in marine animals. Groups of humpback whales engaged in competitive behaviour were followed over the course of four field seasons, and data were recorded on their level of surface aggression and the calls produced by the group. Calls were assigned to a call type based on the results of Chapter 3. Competitive groups were split into different levels of group intensity (i.e. aggression and arousal), which allowed for correlations to be made regarding communicative and surface behaviour with progressive levels of intensity. As a final discussion, Chapter 6 provides an overall summary of the results found in the thesis, detailing their significance in the context of the study of communicative behaviour in animals. Additionally, knowledge gaps are addressed which have become apparent during the course of this thesis, and future research goals are suggested.

1.7 Literature Cited

Alexander RD (1961) Aggressiveness, territoriality, and sexual behavior in field crickets (Orthoptera: Gryllidae). Behaviour 17:130–223. Allen JA (2019) Community through culture: From insects to whales. How social learning and culture manifest across diverse animal communities. BioEssays 41:1–8. Allen JA, Garland EC, Dunlop RA, Noad MJ (2018) Cultural revolutions reduce complexity in the songs of humpback whales. Proceedings of the Royal Society B: Biological Sciences 285:2–7. Allen JA, Garland EC, Dunlop RA, Noad MJ (2019) Network analysis reveals underlying syntactic features in a vocally learnt mammalian display, humpback whale song. Proceedings of the Royal Society B: Biological Sciences 286:1–8. Archer J (1988) The Behavioural Biology of Aggression. Cambridge University Press, Cambridge, MA.

Chapter 1: Introduction

Bain DE (1986) Acoustic behavior of Orcinus: sequences, periodicity, behavioral correlates and an automated technique for call classification. In: Kirkevold BC, Lockard JS (eds) Behavioral biology of killer whales. Alan R. Liss, Inc., New York, pp 335–371. Baker CS, Herman LM (1984) Aggressive behavior between humpback whales (Megaptera novaeangliae) wintering in Hawaiian waters. Canadian Journal of Zoology 62:1922–1937. Barklow WE (1979) Graded Frequency Variations of the Tremolo Call of the Common Loon (Gavia immer). The Condor 81:53–64. Bastian A, Schmidt S (2008) Affect cues in vocalizations of the bat, Megaderma lyra, during agonistic interactions. The Journal of the Acoustical Society of America 124:598–608. Bee MA, Kozich CE, Blackwell KJ, Carl Gerhardt H (2001) Individual variation in advertisement calls of territorial male green frogs, Rana clamitans: Implications for individual discrimination. Ethology 107:65–84. Bergman TJ, Beehner JC (2015) Measuring social complexity. Animal Behaviour 103:203–209. Berta A, Sumich JL (1999) Marine mammals: Evolutionary biology. Academic Press, San Diego. Beynon P, Rasa OAE (1989) Do dwarf mongooses have a language-Warning vocalizations transmit complex information. South African Journal of Science 85:447–450. Blesdoe EK, Blumstein DT (2014) What is the sound of fear? Behavioral responses of white- crowned sparrows Zonotrichia leucophrys to synthesized nonlinear acoustic phenomena. Current Zoology 60:534–541. Blumstein DT, Chi YY (2012) Scared and less noisy: Glucocorticoids are associated with alarm call entropy. Biology Letters 8:189–192. Blumstein DT, Munos O (2005) Individual, age and sex-specific information is contained in yellow-bellied marmot alarm calls. Animal Behaviour 69:353–361. Bouchet H, Blois-Heulin C, Lemasson A (2013) Social complexity parallels vocal complexity: a comparison of three non-human primate species. Frontiers in Psychology 4:1–15. Bradbury JW, Vehrencamp SL (2011) Principles of Animal Communication, 2nd edn. Sinauer Associates, Inc., Sunderland, MA. Briefer EF (2012) Vocal expression of emotions in mammals: Mechanisms of production and evidence. Journal of Zoology 288:1–20. Brown M, Corkeron P (1995) Pod characteristics of migrating humpback whales (Megaptera novaeangliae) off the East Australian coast. Behaviour 132:163–179. Burghardt GM, Bartmess-Levasseur JN, Browning SA, Morrison KE, Stec CL, Zachau CE, Freeberg TM (2012) Perspectives - Minimizing observer bias in behavioral studies: A review and recommendations. Ethology 118:511–517.

Chapter 1: Introduction

Burnham R, Duffus D, Mouy X (2018) Gray whale (Eschrictius robustus) call types recorded during migration off the west coast of Vancouver Island. Frontiers in Marine Science 5:1–11. Busnel RG, Dziedzic A (1966) Acoustic signals of the Pilot whale Globicephala melaena and of the porpoises Delphinus delphis and Phocoena phocoena. In: Norris KS (ed) Whales, Dolphins and Porpoise. University of California Press, Berkeley, CA, pp 607–648. Caldwell MC, Caldwell DK (1965) Individualized whistle contours in bottle-nosed dolphins (Tursiops truncatus). Nature 207:434–435. Campagna C (2009) Aggressive Behavior, Intraspecific. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 18–24. Catchpole CK, Slater PJ (2008) Bird song: biological themes and variations. Cambridge University Press, Cambridge, UK. Cerchio S, Dahlheim M (2001) Variation in feeding vocalizations of humpback whales Megaptera novaeangliae from southeast Alaska. Bioacoustics 11:277–295. Charif RA, Clapham PJ, Clark CW (2001) Acoustic detections of singing humpback whales in deep waters off the British Isles. Marine Mammal Science 17:751–768. Charlton BD, Reby D (2011) Context-related acoustic variation in male fallow deer (Dama dama) groans. PLoS ONE 6:e21066. Charlton BD, Zhihe Z, Snyder RJ (2009) The information content of giant panda, Ailuropoda melanoleuca, bleats: Acoustic cues to sex, age and size. Animal Behaviour 78:893–898. Chittleborough RG (1965) Dynamics of two populations of the humpback whale, Megaptera novaeangliae (Borowski). Australian Journal of Marine and Freshwater Research 16:33–128. Chittleborough RG (1955) Puberty, physical maturity, and relative growth of the female humpback whale, Megaptera nodosa (Bonnaterre), on the Western Australian coast. Marine and Freshwater Research 6:315–327. Chittleborough RG (1957) The breeding cycle of the female humpback whale, Megaptera nodosa (Bonnaterre). Australian Journal of Marine and Freshwater Research 9:1–18. Christie PJ, Mennill DJ, Ratcliffe LM (2004) Chickadee song structure is individually distinctive over long broadcast distances. Behaviour 141:101–124. Chu K, Harcourt P (1986) Behavioral correlations with aberrant patterns in humpback whale songs. Behavioral Ecology and Sociobiology 19:309–312. Clapham PJ (1993a) Seasonal occurrence and annual return of humpback whales, Megaptera novaeangliae, in the southern Gulf of Maine. Canadian Journal of Zoology 71:440–443. Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mammal Review 26:27–49.

Chapter 1: Introduction

Clapham PJ (1993b) Social organization of humpback whales on a North Atlantic feeding ground. Zoological Symposium 66:131–145. Clapham PJ (2000) The humpback whale: seasonal feeding and breeding in a baleen whale. In: Mann J, Connor R, Tyack PL, Whitehead H (eds) Cetacean Societies: Field Studies of Dolphins and Whales. University of Chicago Press, Chicago, pp 173–196. Clapham PJ, Mattila DK, Palsbøll PJ (2008) High-latitude-area composition of humpback whale competitive groups in Samana Bay: Further evidence for panmixis in the North Atlantic population. Canadian Journal of Zoology 71:1065–1066. Clapham PJ, Mayo CA (1990) Reproduction of humpback whales (Megaptera novaeangliae) observed in the Gulf of Maine. Report of the International Whaling Commission 171–175. Clapham PJ, Palsboll PJ, Mattila DK, Vasquez O (1992) Composition and dynamics of humpback whale competitive groups in the West Indies. Behaviour 122:182–194. Clark CW (1982) The acoustic repertoire of the Southern right whale, a quantitative analysis. Animal Behaviour 30:1060–1071. Clark CW (1990) Acoustic behavior of mysticete whales. In: Thomas J, Kastelein R (eds) Sensory Abilities of Cetaceans. Plenum Press, New York, pp 571–583. Clutton-Brock TH, Albon SD (1979) The roaring of red deer and the evolution of honest advertisement. Behaviour 69:145–170. Collins KT, Rogers TL, Terhune JM, McGreevy PD, Wheatley KE, Harcourt RG (2005) Individual variation of in-air female “pup contact” calls in Weddell seals, Leptonychotes weddellii. Behaviour 142:167–189. Cummings WC, Thompson PO, Cook R (1968) Underwater sounds of migrating gray whales, Eschrichtius glaucus (Cope). The Journal of the Acoustical Society of America 44:1278–1281. D’Vincent CG, Nilson RM, Hanna RE (1985) Vocalizations and coordinated feeding behavior of the humpback whale in southeastern Alaska. Scientific Report of the Whales Research Institute 36:41–47. Darling JD (2001) Characterization of Behavior of Humpback Whales in Hawaiian Waters.Honolulu, HI. Darling JD, Acebes JM V., Yamaguchi M (2014) Similarity yet a range of differences between humpback whale songs recorded in the Philippines, Japan and Hawaii in 2006. Aquatic Biology 21:93–107. Darling JD, Bérubé M (2001) Interactions of singing humpback whales with other males. Marine Mammal Science 17:570–584.

Chapter 1: Introduction

Darling JD, Jones ME, Nicklin CP (2006) Humpback whale songs: Do they organize males during the breeding season? Behaviour 143:1051–1101. Darwin C (1872) The Expression of the Emotions in Man and Animals. John Murray, London. Davies NB, Halliday TR (1978) Deep croaks and fighting assessment in toads Bufo bufo. Nature 274:683–685. Dawbin WH (1966) The seasonal migratory cycle of humpback whales. In: Whales, Dolphins, and Porpoises. pp 145–170. de Waal FBM, Tyack PL (2003) Animal Social Complexity: Intelligence, Culture, and Individualized Societies. Harvard University Press, Cambridge. Dunlop RA (2016) Changes in vocal parameters with social context in humpback whales: considering the effect of bystanders. Behavioral Ecology and Sociobiology 70:857–870. Dunlop RA (2017) Potential motivational information encoded within humpback whale non-song vocal sounds. The Journal of the Acoustical Society of America 141:2204–2213. Dunlop RA (2018) The communication space of humpback whale social sounds in wind-dominated noise. The Journal of the Acoustical Society of America 144:540–551. Dunlop RA, Cato DH, Noad MJ (2008) Non-song acoustic communication in migrating humpback whales (Megaptera novaeangliae). Marine Mammal Science 24:613–629. Dunlop RA, Cato DH, Noad MJ (2010) Your attention please: increasing ambient noise levels elicits a change in communication behaviour in humpback whales (Megaptera novaeangliae). Proceedings of the Royal Society B: Biological Sciences 277:2521–2529. Dunlop RA, Cato DH, Noad MJ, Stokes DM (2013) Source levels of social sounds in migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 134:706–714. Dunlop RA, Noad MJ (2016) The “risky” business of singing: Tactical use of song during joining by male humpback whales. Behavioral Ecology and Sociobiology 70:2149–2160. Dunlop RA, Noad MJ, Cato DH, Stokes DM (2007) The social vocalization repertoire of east Australian migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 122:2893–2905. East ML, Hofer H (1991a) Loud calling in a female-dominated mammalian society II. Behavioural contexts and functions of whooping of spotted hyaenas, Crocuta crocuta. Animal Behaviour 42:651–669. East ML, Hofer H (1991b) Loud calling in a female-dominated mammalian society I. Structure and composition of whooping bouts of spotted hyaenas, Crocuta crocuta. Animal Behaviour 42:637–649.

Chapter 1: Introduction

Endres W, Bambach W, Flösser G (1971) Voice spectrograms as a function of age, voice disguise, and voice imitation. The Journal of the Acoustical Society of America 49:1842–1848. Ey E, Fischer J (2009) The “acoustic adaptation hypothesis”—A review of the evidence from birds, anurans and mammals. Bioacoustics 19:21–48. Felix F, Novillo J (2015) Structure and dynamics of humpback whales competitive groups in Ecuador. Animal Behavior and Cognition 2:56–70. Fischer J (2011) Where is the information in animal communication? Animal Thinking: Contemporary Issues in Comparative Cognition 151–161. Fischer J (2013) Information, inference and meaning in primate vocal behaviour. In: Stegmann U (ed) Animal Communication Theory: Information and Influence. Cambridge University Press, Cambridge, MA, pp 297–317. Fischer J (2006) Categorical perception in animals. In: Brown K (ed) Encyclopedia of language and linguistics, 2nd edn. Elsevier Ltd, London, pp 248–251. Fischer J, Farnworth MS, Sennhenn-Reulen H, Hammerschmidt K (2017a) Quantifying social complexity. Animal Behaviour 130:57–66. Fischer J, Hammerschmidt K (2002) An overview of the Barbary macaque, Macaca sylvanus, vocal repertoire. Folia Primatologica 73:32–45. Fischer J, Kitchen DM, Seyfarth RM, Cheney DL (2004) Baboon loud calls advertise male quality: Acoustic features and their relation to rank, age, and exhaustion. Behavioral Ecology and Sociobiology 56:140–148. Fischer J, Wadewitz P, Hammerschmidt K (2017b) Structural variability and communicative complexity in acoustic communication. Animal Behaviour 134:229–237. Fitch WT, Hauser MD (2002) Unpacking " honesty ": Vertebrate vocal production and the evolution of acoustic signals. In: Simmons A, Fay RR, Popper AN (eds) Acoustic Communication. Springer, New York (NY), pp 1–44. Fitch WT, Neubauer J, Herzel H (2002) Calls out of chaos: The adaptive significance of nonlinear phenomena in mammalian vocal production. Animal Behaviour 63:407–418. Fitch WT, Reby D (2001) The descended larynx is not uniquely human. Proceedings of the Royal Society B: Biological Sciences 268:1669–1675. Ford JKB (1989) Acoustic behaviour of resident killer whales (Orcinus orca) off Vancouver Island, British Columbia. Canadian Journal of Zoology 67:727–745. Fournet ME, Szabo A, Mellinger DK (2015) Repertoire and classification of non-song calls in Southeast Alaskan humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 137:1–10.

Chapter 1: Introduction

Fournet MEH, Gabriele CM, Sharpe F, Straley JM, Szabo A (2018a) Feeding calls produced by solitary humpback whales. Marine Mammal Science 34:851–865. Fournet MEH, Matthews LP, Gabriele CM, Mellinger DK, Klinck H (2018b) Source levels of foraging humpback whale calls. The Journal of the Acoustical Society of America 143:EL105– EL111. Frankel AS, Clark CW, Herman LM, Gabriele CM (1995) Spatial distribution, habitat utilization, and social interactions of humpback whales, Megaptera novaeangliae, off Hawai’i, determined using acoustic and visual techniques. Canadian Journal of Zoology 73:1134–1146. Freeberg T (2006) Social complexity can drive vocal complexity: Group size and information in chickadee calls in Carolina chickadees. The Journal of the Acoustical Society of America 17:557–561. Freeberg TM, Dunbar RIM, Ord TJ (2012) Social complexity as a proximate and ultimate factor in communicative complexity. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1785–1801. Freeberg TM, Krams I (2015) Does social complexity link vocal complexity and cooperation? Journal of Ornithology 156:125–132. Freeberg TM, Lucas JR (2012) Information theoretical approaches to chick-a-dee calls of carolina chickadees (Poecile carolinensis). Journal of Comparative Psychology 126:68–81. Garland EC, Castellote M, Berchok CL (2015) Beluga whale (Delphinapterus leucas) vocalizations and call classification from the eastern Beaufort Sea population. The Journal of the Acoustical Society of America 137:3054–3067. Garland EC, Gedamke J, Rekdahl ML, Noad MJ, Garrigue C, Gales N (2013) Humpback whale song on the Southern Ocean feeding grounds: Implications for cultural transmission. PLoS ONE 8:1–9. Garland EC, Goldizen AW, Rekdahl ML, Constantine R, Garrigue C, Hauser ND, Poole MM, Robbins J, Noad MJ (2011) Dynamic horizontal cultural transmission of humpback whale song at the ocean basin scale. Current Biology 21:687–691. Girola E, Noad MJ, Dunlop RA, Cato DH (2019) Source levels of humpback whales decrease with frequency suggesting an air-filled resonator is used in sound production. The Journal of the Acoustical Society of America 145:869–880. Glockner-Ferrari DA, Ferrari MJ (1990) Reproduction in the humpback whale (Megaptera novaeangliae) in Hawaiin waters, 1975-1088: the life history, reproductive rates, and behaviour of known individuals identified through surface and underwater photography. Reports of the International Whaling Commission 12:161–169.

Chapter 1: Introduction

Glockner DA, Venus SC (1983) Identification, growth rate, and behavior of humpback whale (Megaptera novaeangliae) cows and calves in the waters of Maui, Hawaii, 1977-79. In: Payne R (ed) Communication and Behavior of Whales. Westview Press, Boulder, pp 223–258. Green S, Marler P (1979) The Analysis of Animal Communication. In: Social Behavior and Communication. Plenum Press, New York (NY), pp 73–158. Griebel U, Oller DK (2008) Evolutionary forces favoring communicative flexibility. In: Griebel U, Oller DK (eds) Evolution of communicative flexibility: complexity, creativity, and adaptability in human and animal communication. MIT Press, Cambridge, MA, pp 9–41. Halliday TR, Slater PJB (eds) (1983) Animal Behaviour, vol. 2: Communication. Freeman, New York. Hamilton WD (1964) The genetical evolution of social behavior. II. Journal of Theoretical Biology 7:17–52. Hammerschmidt K, Fischer J (1998) The vocal repertoire of Barbary macaques: A quantitative analysis of a graded signal system. Ethology 104:203–216. Hanggi EB, Schusterman RJ (1994) Underwater acoustic displays and individual variation in male harbor seals, Phoca vitulina. Animal Behaviour 48:1275–1283. Hardouin LA, Reby D, Bavoux C, Burneleau G, Bretagnolle V (2007) Communication of male quality in owl hoots. American Naturalist 169:552–562. Hauser MD (1996) The evolution of communication. MIT Press, Cambridge, MA. Hayes SA, Kumar A, Costa DP, Mellinger DK, Harvey JT, Southall BL, Le Boeuf BJ (2004) Evaluating the function of the male harbour seal, Phoca vitulina, roar through playback experiments. Animal Behaviour 67:1133–1139. Herman LM (2017) The multiple functions of male song within the humpback whale (Megaptera novaeangliae) mating system: review, evaluation, and synthesis. Biological Reviews 92:1795– 1818. Herzing DL (1996) Vocalizations and associated underwater behavior of free-ranging Atlantic spotted dolphins, Stenella frontalis and bottlenose dolphins, Tursiops truncatus. Aquatic Mammals 22:61–79. Hofmann HA, Schildberger K (2001) Assessment of strength and willingness to fight during aggressive encounters in crickets. Animal Behaviour 62:337–348. Huntingford FA, Turner AA (1987) Animal Conflict. Chapman and Hall, London. Hӧlldobler B (1999) Multimodal signals in ant communication. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 184:129–141.

Chapter 1: Introduction

Indeck KL (2020) Acoustic communication of female-calf humpback whales during migration. PhD Thesis. The University of Queensland, Queensland. IWC (2014) Annex H: Report of the sub-committee on other Southern Hemisphere whale stocks.Bled, Slovenia. Janik VM (1999) Pitfalls in the categorization of behaviour: A comparison of dolphin whistle classification methods. Animal Behaviour 57:133–143. Jones ME (2010) Female humpback whale (Megaptera novaeangliae) reproductive class and male- female interactions during the breeding season. PhD Thesis. Antioch University New England, Keene (NH). Jurasz CM, Jurasz VP (1979) Feeding modes of the humpback whale, Megaptera noavaengliae, in southeast Alaska. Scientific Report of the Whales Research Institute 31:69–83. Kappeler PM (2019) A framework for studying social complexity. Behavioral Ecology and Sociobiology 73:1–14. Kappeler PM, Clutton-Brock T, Shultz S, Lukas D (2019) Social complexity: patterns, processes, and evolution. Behavioral Ecology and Sociobiology 73:1–6. Keenan S, Lemasson A, Zuberbühler K (2013) Graded or discrete? A quantitative analysis of Campbell’s monkey alarm calls. Animal Behaviour 85:109–118. Kershenbaum A (2014) Entropy rate as a measure of animal vocal complexity. Bioacoustics 23:195–208. Kershenbaum A, Blumstein DT, Roch MA, et al (2016) Acoustic sequences in non-human animals: A tutorial review and prospectus. Biological Reviews 91:13–52. Kershenbaum A, Déaux ÉC, Habib B, Mitchell B, Palacios V, Root-Gutteridge H, Waller S (2018) Measuring acoustic complexity in continuously varying signals: how complex is a wolf howl? Bioacoustics 27:215–229. Koren L, Geffen E (2009) Complex call in male rock hyrax (Procavia capensis): A multi- information distributing channel. Behavioral Ecology and Sociobiology 63:581–590. Kotiaho JS, Alatalo R V., Mappes J, Parri S (1999) Honesty of agonistic signalling and effects of size and motivation asymmetry in contests. Acta Ethologica 2:13–21. Kroodsma DE (1977) Correlates of song organization among North American wrens. The American Naturalist 111:995–1008. Lamarck JB (1809) Philosophie Zoologique. Hafner Publishing, New York. Leong KM, Ortolani A, Burks KD, Mellen JD, Savage A (2003) Quantifying acoustic and temporal characteristics of vocalizations for a group of captive african elephants Loxodonta africana. Bioacoustics 13:213–231.

Chapter 1: Introduction

Liu WC (2004) The effect of neighbours and females on dawn and daytime singing behaviours by male chipping sparrows. Animal Behaviour 68:39–44. Lockyer C (1976) Body weights of some species of large whales. ICES Journal of Marine Science 36:259–273. Luís AR, Couchinho MN, Dos Santos ME (2016) A quantitative analysis of pulsed signals emitted by wild bottlenose dolphins. PLoS ONE 11:e0157781. Manser MB (2010) The generation of functionally referential and motivational vocal signals in mammals. In: Brudzynski SM (ed) Handbook of Mammalian Vocalization - an integrative neuroscience approach. Academic Press, London (UK), pp 477–486. Manser MB, Jansen DAWAM, Graw B, Hollén LI, Bousquet CAH, Furrer RD, le Roux A (2014) Vocal complexity in meerkats and other mongoose species. Advances in the Study of Behavior 46:281–310. Marler P (1975) On the origin of speech from animal sounds. In: Kavanagh JF, Cutting J (eds) The Role of Speech in Language. MIT Press, Cambridge (MA), pp 11–37. Marler P (1976) Social organization, communication, and graded signals: The chimpanzee and the gorilla. In: Bateson PPG, Hinde RA (eds) Growing Points in Ethology. Cambridge University Press, Oxford (UK), pp 239–277. Marler P (1961) The logical analysis of animal communication. Journal of Theoretical Biology 1:295–317. Marler P (1977) The structure of animal communication sounds. In: Bullock T, Evans E (eds) Recognition of complex acoustic signals. Dahlem Konferenzen, Berlin, pp 17–35. Marler P, Evans CS, Hauser MD (1992) Animal signals: Motivational, referential, or both? In: Papousek H, Jurgens U, Papousek M (eds) Nonverbal vocal communication: comparative and developmental approaches. Cambridge University Press, Cambridge, UK, pp 66–86. Marler P, Vandenbergh JG (1979) Social Behavior and Communication. Plenum Press, New York. Mathevon N, Koralek A, Weldele M, Glickman SE, Theunissen FE (2010) What the hyena’s laugh tells: Sex, age, dominance and individual signature in the giggling call of Crocuta crocuta. BMC Ecology 10:1–16. Mattila DK, Clapham PJ, Vasquez O, Bowman RS (1994) Occurrence, population composition, and habitat use of humpback whales in Samana Bay, Dominican Republic. Canadian Journal of Zoology 72:1898–1907. Mattila DK, Guinee LN, Mayo CA (1987) Humpback whale songs on a North Atlantic feeding ground. Journal of Mammalogy 68:880–883.

Chapter 1: Introduction

May-Collado LJ, Agnarsson I, Wartzok D (2007) Phylogenetic review of tonal sound production in whales in relation to sociality. BMC Evolutionary Biology 7:1–20. Maynard-Smith J, Harper D (2003) Animal Signals. Oxford University Press, Oxford, UK. Maynard Smith J (1974) The theory of games and the evolution of animal conflicts. Journal of Theoretical Biology 47:209–221. Maynard Smith J, Price GR (1973) The logic of animal conflict. Nature 246:15–18. McComb K, Semple S (2005) Coevolution of vocal communication and sociality in primates. Biology Letters 1:381–385. McCordic JA, Root-Gutteridge H, Cusano DA, Denes SL, Parks SE (2016) Calls of North Atlantic right whales Eubalaena glacialis contain information on individual identity and age class. Endangered Species Research 30:157–169. McDonald MA, Calambokidis J, Teranishi AM, Hildebrand JA (2001) The acoustic calls of blue whales off California with gender data. The Journal of the Acoustical Society of America 109:1728–1735. McGregor PK (ed) (2005) Animal Communication Networks. Cambridge University Press, Cambridge. Mobley, Jr. JR, Herman LM (1985) Transience of social affiliations among humpback whales on the Hawaiian wintering grounds. Canadian Journal of Zoology 63:762–772. Mobley, Jr. JR, Herman LM, Frankel AS (1988) Responses of wintering humpback whales (Megaptera novaeangliae) to playback of recordings of winter and summer vocalizations and of synthetic sound. Behavioral Ecology and Sociobiology 23:211–223. Morton ES (1977) On the occurrence and significance of motivation-structural rules in some bird and mammal sounds. The American Naturalist 111:855–869. Morton ES (1982) Grading, discreteness, redundancy, and motivation-structural rules. In: Kroodsma DE, Miller MH (eds) Acoustic communication in birds. Academic Press, New York (NY), pp 183–212. Nicholson TE (2000) Social structure and underwater behavior of harbor seals in southern Monterey Bay, California. M.S. Thesis. San Francisco State University, San Francisco, CA. Noad MJ, Cato DH, Bryden MM, Jenner M-N, Jenner KCS (2000) Cultural revolution in whale songs. Nature 408:537. Oller DK, Griebel U (eds) (2008) Evolution of communicative flexibility: Complexity, creativity, and adaptability in human and animal communication. MIT Press, Cambridge, MA. Ouattara K, Lemasson A, Zuberbuhler K (2009) Campbell’s monkeys use affixation to alter call meaning. PloS one 4:1–7.

Chapter 1: Introduction

Pack AA, Herman LM, Craig AS, Spitz SS, Deakos MH (2002) Penis extrusions by humpback whales (Megaptera novaeangliae). Aquatic Mammals 28:131–146. Pack AA, Salden DR, Ferrari MJ, Glockner-Ferrari DA, Herman LM, Stubbs HA, Straley JM (1998) Male humpback whale dies in competitive group. Marine Mammal Science 14:861– 873. Parks SE, Clark CW (2007) Acoustic communication: Social sounds and the potential impacts of noise. In: Kraus SD, Rolland RM (eds) The Urban Whale: North Atlantic Right Whales at the Crossroads. Harvard University Press, Cambridge, MA, pp 310–332. Parks SE, Cusano DA, Stimpert AK, Weinrich MT, Friedlaender AS, Wiley DN (2014) Evidence for acoustic communication among bottom foraging humpback whales. Scientific Reports 4:7508. Parks SE, Searby A, Célérier A, Johnson MP, Nowacek DP, Tyack PL (2011) Sound production behavior of individual North Atlantic right whales: implications for passive acoustic monitoring. Endangered Species Research 15:63–76. Parks SE, Tyack PL (2005) Sound production by North Atlantic right whales (Eubalaena glacialis) in surface active groups. The Journal of the Acoustical Society of America 117:3297–3306. Paterson R, Paterson P (1984) A study of the past and present status of humpback whales in east Australian waters. Biological Conservation 29:321–343. Payne K (2003) Sources of social complexity in the three elephant species. In: de Waal FBM, Tyack PL (eds) Animal Social Complexity: Intelligence, Culture, and Individualized Societies. Harvard University Press, Cambridge, pp 57–86. Payne K, Payne R (1985) Large Scale Changes over 19 Years in Songs of Humpback Whales in Bermuda. Zeitschrift für Tierpsychologie 68:89–114. Payne K, Tyack PL, Payne R (1984) Progressive changes in the songs of humpback whales (Megaptera novaeangliae): A detailed analysis of two seasons in Hawaii. In: Payne R (ed) Communication and Behavior of Whales. Westview Press, Boulder, pp 9–57. Payne RS, McVay S (1971) Songs of humpback whales. Science 173:585–597. Peckre L, Kappeler PM, Fichtel C (2019) Clarifying and expanding the social complexity hypothesis for communicative complexity. Behavioral Ecology and Sociobiology 73:1–19. Pfefferle D, Fischer J (2006) Sounds and size: identification of acoustic variables that reflect body size in hamadryas baboons, Papio hamadryas. Animal Behaviour 72:43–51. Pika S (2017) Unpeeling the layers of communicative complexity. Animal Behaviour 134:223–227.

Chapter 1: Introduction

Pomilla C, Rosenbaum HC (2006) Estimates of relatedness in groups of humpback whales (Megaptera novaeangliae) on two wintering grounds of the Southern Hemisphere. Molecular Ecology 15:2541–2555. Price JJ, Earnshaw SM, Webster MS (2006) Montezuma oropendolas modify a component of song constrained by body size during vocal contests. Animal Behaviour 71:799–807. Ramp C, Hagen W, Palsbøll P, Bérubé M, Sears R (2010) Age-related multi-year associations in female humpback whales (Megaptera novaeangliae). Behavioral Ecology and Sociobiology 64:1563–1576. Rebout N, De Marco A, Lone JC, et al (2020) Tolerant and intolerant macaques show different levels of structural complexity in their vocal communication: Vocal complexity in macaques. Proceedings of the Royal Society B: Biological Sciences 287: Reby D, Joachim J, Lauga J, Lek S, Aulagnier S (1998) Individuality in the groans of fallow deer (Dama dama) bucks. The Journal of Zoology, London 245:79–84. Reby D, McComb K (2003) Anatomical constraints generate honesty: Acoustic cues to age and weight in the roars of red deer stags. Animal Behaviour 65:519–530. Reby D, McComb K, Cargnelutti B, Darwin C, Fitch WT, Clutton-Brock T (2005) Red deer stags use formants as assessment cues during intrasexual agonistic interactions. Proceedings of the Royal Society B: Biological Sciences 272:941–947. Rehn N, Filatova OA, Durban JW, Foote AD (2011) Cross-cultural and cross-ecotype production of a killer whale “excitement” call suggests universality. Naturwissenschaften 98:1–6. Rehn N, Teichert S, Thomsen F (2007) Structural and temporal emission patterns of variable pulsed calls in free-ranging killer whales (Orcinus orca). Behaviour 144:307–329. Reichert MS, Gerhardt HC (2013) Gray tree frogs, Hyla versicolor, give lower-frequency aggressive calls in more escalated contests. Behavioral Ecology and Sociobiology 67:795–804. Ręk P, Osiejuk TS, Budka M (2011) Functionally similar acoustic signals in the corncrake (Crex crex) transmit information about different states of the sender during aggressive interactions. Hormones and Behavior 60:706–712. Rekdahl ML, Dunlop RA, Goldizen AW, Garland EC, Biassoni N, Miller P, Noad MJ (2015) Non- song social call bouts of migrating humpback whales. The Journal of the Acoustical Society of America 137:3042–3053. Rekdahl ML, Dunlop RA, Noad MJ, Goldizen AW (2013) Temporal stability and change in the social call repertoire of migrating humpback whales. The Journal of the Acoustical Society of America 133:1785–1795.

Chapter 1: Introduction

Rendall D (2003) Acoustic correlates of caller identity and affect intensity in the vowel-like grunt vocalizations of baboons. The Journal of the Acoustical Society of America 113:3390–3402. Rendall D, Owren MJ, Ryan MJ (2009) What do animal signals mean? Animal Behaviour 78:233– 240. Riede T, Owren MJ, Arcadi AC (2004) Nonlinear acoustics in pant hoots of common chimpanzees (Pan troglodytes): Frequency jumps, subharmonics, biphonation, and deterministic chaos. American Journal of Primatology 64:277–291. Root-Gutteridge H, Cusano DA, Shiu Y, Nowacek DP, Van Parijs SM, Parks SE (2018) A lifetime of changing calls: North Atlantic right whales, Eubalaena glacialis, refine call production as they age. Animal Behaviour 137:21–34. Ruch H, Zürcher Y, Burkart JM (2018) The function and mechanism of vocal accommodation in humans and other primates. Biological Reviews 93:996–1013. Scarantino A (2010) Animal communication between information and influence. Animal Behaviour 79:1–5. Schehka S, Zimmermann E (2009) Acoustic features to arousal and identity in disturbance calls of tree shrews (Tupaia belangeri). Behavioural Brain Research 203:223–231. Schwartz JJ (1989) Graded aggressive calls of the spring peeper, Pseudacris crucifer. Herpetologica 45:172–181. Seyfarth RM, Cheney DL (2003) Signalers and receivers in animal communication. Annual Reviews in Psychology 54:145–173. Sharpe FA (2001) Social foraging of the southeast Alaskan humpback whale, Megaptera novaeangliae. PhD Thesis. Simon Fraser University, Ann Arbor, MI. Silber GK (1986) The relationship of social vocalizations to surface behavior and aggression in the Hawaiian humpback whale (Megaptera novaeangliae). Can. J. Zool. 64:2075–2080 Silk J, Cheney D, Seyfarth R (2013) A practical guide to the study of social relationships. Evolutionary Anthropology 22:213–225. Širović A, Rice A, Chou E, Hildebrand JA, Wiggins SM, Roch MA (2015) Seven years of blue and fin whale call abundance in the Southern California Bight. Endangered Species Research 28:61–76. Smith JN, Goldizen AW, Dunlop RA, Noad MJ (2008) Songs of male humpback whales, Megaptera novaeangliae, are involved in intersexual interactions. Animal Behaviour 76:467– 477.

Chapter 1: Introduction

Smith JN, Grantham HS, Gales N, Double MC, Noad MJ, Paton D (2012) Identification of humpback whale breeding and calving habitat in the Great Barrier Reef. Marine Ecology Progress Series 447:259–272. Soltis J, Leong K, Savage A (2005) African elephant vocal communication II: Rumble variation reflects the individual identity and emotional state of callers. Animal Behaviour 70:589–599. Sousa-Lima RS, Paglia AP, da Fonseca GAB (2008) Gender, age, and identity in the isolation calls of Antillean manatees (Trichechus manatus manatus). Aquatic Mammals 34:109–122. Stevick PR, Neves MC, Johansen F, Engel MH, Allen J, Marcondes MCC, Carlson C (2010) A quarter of a world away: female humpback whale moves 10000 km between breeding areas. Biology Letters 131–167. Stimpert AK, Au WWL, Parks SE, Hurst T, Wiley DN (2011) Common humpback whale (Megaptera novaeangliae) sound types for passive acoustic monitoring. The Journal of the Acoustical Society of America 129:476–482. Stirling I, Thomas JA (2003) Relationships between underwater vocalizations and mating systems in phocid seals. Aquatic Mammals 29:227–246. Stoeger AS, Charlton BD, Kratochvil H, Fitch WT (2011) Vocal cues indicate level of arousal in infant African elephant roars. The Journal of the Acoustical Society of America 130:1700– 1710. Suzuki TN (2014) Communication about predator type by a bird using discrete, graded and combinatorial variation in alarm calls. Animal Behaviour 87:59–65. Taruski AG (1979) The whistle repertoire of the North Atlantic pilot whale (Globicephala melaena) and its relationship to behavior and environment. In: Winn HE, Olla BL (eds) Behavior of Marine Mammals. Plenum Press, New York, pp 345–368. Temeles EJ (1994) The role of neighbours in territorial systems: when are they dear enemies? Animal Behaviour 47:339–350. Thompson PO, Cummings WC, Ha SJ (1986) Sounds, source levels, and associated behavior of humpback whales, Southeast Alaska. The Journal of the Acoustical Society of America 80:735–740. Tripovich JS, Canfield R, Rogers TL, Arnould JPY (2008) Characterization of Australian fur seal vocalizations during the breeding season. Marine Mammal Science 24:913–928. Tyack PL (1981) Interactions between singing Hawaiian humpback whales and conspecifics nearby. Behavioral Ecology 8:105–116. Tyack PL (1983) Differential response of humpback whales, Megaptera novaeangliae, to playback of song or social sounds. Behavioral Ecology and Sociobiology 13:49–55.

Chapter 1: Introduction

Tyack PL, Whitehead H (1983) Male competition in large groups of wintering humpback whales. Behaviour 83:132–154. Valsecchi E, Hale P, Corkeron P, Amos W (2002) Social structure in migrating humpback whales (Megaptera novaeangliae). Molecular Ecology 11:507–518. Vannoni E, McElligott AG (2008) Low frequency groans indicate larger and more dominant fallow deer (Dama dama) males. PLoS ONE 3:e3113. von Frisch K (1974) Decoding the language of the bee. Science 185:663–668. Wadewitz P, Hammerschmidt K, Battaglia D, Witt A, Wolf F, Fischer J (2015) Characterizing Vocal Repertoires — Hard vs . Soft Classification Approaches. PLoS ONE 10:1–16. Wagner, Jr WE (1992) Deceptive or honest signalling of fighting ability? A test of alternative hypotheses for the function of changes in call dominant frequency by male cricket frogs. Animal Behaviour 44:449–462. Wagner WE (1989) Graded aggressive signals in Blanchard’s cricket frog: Vocal responses to opponent proximity and size. Animal Behaviour 38:1025–1038. Watkins WA, Moore KE, Clark CW, Dahlheim M (1988) The sounds of sperm whale calves. In: Nachtigall PE, Moore PWB (eds) NATO ASI Science Series A: Life Sciences: Vol. 156. Animal Sonar. Springer, Boston (MA), pp 99–107. Watkins WA, Tyack P, Moore KE, Bird JE (1987) The 20-Hz signals of finback whales (Balaenoptera physalus). Journal of the Acoustical Society of America 82:1901–1912. Weinrich MT (1991) Stable social associations among humpback whales (Megaptera novaeangliae) in the southern Gulf of Maine. Canadian Journal of Zoology 69:3012–3019. Wells RS (2003) Dolphin social complexity: Lessons from long-term study and life history. In: de Waal FBM, Tyack PL (eds) Animal Social Complexity: Intelligence, Culture, and Individualized Societies. Harvard University Press, Cambridge, pp 32–56. Wells RS, Scott MD (1999) Bottlenose Dolphin Tursiops truncatus (Montagu, 1821). In: Ridgway SH, Harrison SR (eds) Handbook of Marine Mammals: The Second Book of Dolphins and the Porpoises. Academic Press, pp 137–182. Whitehead H (1983) Structure and stability of humpback whale groups off Newfoundland. Can J Zool 61:1391–1397. Wiley D, Ware C, Bocconcelli A, Cholewiak D, Friedlaender A, Thompson M, Weinrich M (2011) Underwater components of humpback whale bubble-net feeding behaviour. Behaviour 148:575–602. Wilkinson GS, Wenrick Boughman J (1998) Social calls coordinate foraging in greater spear-nosed bats. Animal Behaviour 55:337–350.

Chapter 1: Introduction

Winn HE, Beamish P, Perkins PJ (1979) Sounds of two entrapped humpback whales (Megaptera novaeangliae) in Newfoundland. Marine Biology 55:151–155. Winn HE, Winn LK (1978) The song of the humpback whale in the West Indies. Marine Biology 47:97–114. Wood JD, McCowan B, Langbauer Jr WR, Viljoen JJ, Hart LA (2005) Classification of african elephant loxodonta Africana rumbles using acoustic parameters and cluster analysis. Bioacoustics 15:143–161. Zahavi A (1982) The pattern of vocal signals and the information they convey. Behaviour 80:1–8. Zoidis AM, Smultea MA, Frankel AS, Hopkins JL, Day A, McFarland AS, Whitt AD, Fertl D (2008) Vocalizations produced by humpback whale (Megaptera novaeangliae) calves recorded in Hawaii. The Journal of the Acoustical Society of America 123:1737–1746.

Chapter 2

Humpback whale social call production reflects both motivational state and arousal

Cusano DA, Indeck KI, Noad MJ, Dunlop RA. Humpback whale social call production reflects both motivational state and arousal. Accepted by Bioacoustics.

Contributor Statement of Contribution % Initial concept 20 Analysis 70 Dana Cusano (Candidate) Preparation of figures 100 Writing of text 90 Proof-reading 40 Data collection 10 Kate Indeck Analysis 20 Proof-reading 20 Obtained funding 50 Michael Noad Data collection 45 Supervision, guidance 20 Proof-reading 10 Initial concept 80 Obtained funding 50 Data collection 45 Rebecca Dunlop Supervision, guidance 80 Analysis 10 Writing of text 10 Proof-reading 30

34 Chapter 2: Humpback whale social call production reflects both motivational state and arousal

2.1 Abstract

In many terrestrial species, there is a direct relationship between the acoustic features of vocal signals and the underlying arousal level of the signaller. High arousal contexts are often correlated with complex social interactions, like those associated with breeding behaviour. Acoustic indicators of increased arousal typically include changes to call production, like elevated call rates, long bouts of calls, and calls of longer duration. Research into how this information is used in the acoustic communication of marine mammals during complex social behaviours however is limited. Here, we examined changes in the calling behaviour of humpback whale female-calf groups with the addition of escorts to investigate if and how arousal information is encoded within their non- song vocal signals. Female-calf pairs had low overall call rates, however, the addition of one or more escorts to the group resulted in a significant increase in individual call rates, the use of long bouts of calls, and changes to some call features. Multiple escort groups, presumably the highest in arousal and complexity, demonstrated further changes by introducing, or significantly increasing the use of, specific call types to the repertoire. These results indicate that humpback whales convey arousal by modifying call production, as well as through the use of specific call types, which contributes to our growing understanding of their complex communication system.

Keywords: arousal; Megaptera novaeangliae; humpback whale; vocalisations; acoustic communication; vocal repertoire

2.2 Introduction

Animal vocal cues can contain a multitude of information about the sender, including fixed information independent of both environmental and social context, and flexible information related to internal state (Green and Marler 1979; Gerhardt 1992). Fixed information is generally related to the physiology of the signaller and is relatively stereotyped within calls of the same type In contrast, flexible acoustic information is linked to the internal physiological state of the signaller (Marler 1961; Morton 1977; Scherer 1986; Marler et al. 1992; Scherer 2003). Calls that convey flexible information vary in acoustic structure both between and within a context (Manser 2010). This variation provides listeners with information on subtle changes to the signaller’s internal attributes at the time of the call, including motivational state and arousal (Marler 1961, 1976; Morton 1977, 1982; Owings and Morton 1998; Briefer 2012). In non-human animal research, the term ‘motivation’ generally refers to the drive to carry out a particular behaviour and the biological mechanisms that generate these behaviours (Colgan 1989). It is a process that moves an animal toward action (Bradley 2000; Roughley 2016).

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Motivational states are determined by internal physiological changes (Kleinginna and Kleinginna 1981; Colgan 1989; Mendl et al. 2010), and are primarily considered a result of basic biological needs (Fernald and Fernald 1978), such as fight (‘aggression’), flight (‘avoidance’), sex (‘reproductive’), and hunger (‘feeding’) (Adler 1979). This information can be informative to conspecifics in order to identify the caller’s intent, or the tendency to exhibit a subsequent activity or behaviour (Zahavi 1982; Todt 1986; Fischer 2011). The term ‘arousal’ refers to the intensity of this internal state (Briefer 2012). Communicating arousal can convey potentially important information to conspecifics, including the level of urgency for a response (Green and Marler 1979; Todt 1986; Manser 2001; Fischer 2011). Acoustic cues to motivation have typically been thought to be reflected in changes to frequency measurements (Morton 1977). For example, call types produced in negative motivational states (e.g. aggression) are typically lower in frequency, cover a wider frequency range (bandwidth), and have fewer frequency modulations (Morton 1977; Briefer 2012). This has been observed in the ‘bugle calls’ of North American elk, Cervus elaphus, which were more broadband and had more low frequency harmonics in aggressive contexts (Feighny et al. 2006). Additionally, white-faced capuchins, Cebus capucinus, in aggressive contexts produced low frequency, broadband call types (Gros-Louis et al. 2008). However, there are many exceptions to these generalities (Hauser and Marler 1993; Schrader and Todt 1993; Fischer et al. 1995; Gouzoules and Gouzoules 2000), which indicates additional attributes (e.g. arousal) may be equally important in explaining variation in acoustic structure. Extensive research indicates that potential information on signaller arousal is most consistently related to changes in temporal parameters, including call rate, inter-call interval, the production of calls in sequences (‘bouts’), and call duration (Scherer 1986, 2003; Briefer 2012). However, frequency also appears to be consistently correlated with arousal. Overall, vocal signals produced in high arousal contexts are typically emitted at higher rates and frequencies, have decreased inter-call intervals, and are longer in duration (Briefer 2012; Fischer and Price 2017). These trends have been demonstrated in the calls of greater false vampire bats, Megaderma lyra (Bastian and Schmidt 2008) and ravens, Corvus corax (Szipl et al. 2017) which increased in duration and call rate during high arousal situations. In addition, increased arousal has also been correlated with higher and more variable frequencies and increased amplitude, as seen in the alarm calls of red-fronted lemurs, Eulemur fulvus rufus (Fichtel and Hammerschmidt 2002) and squirrel monkeys, Saimiri sciureus (Fichtel and Hammerschmidt 2003). Acoustic information on motivation and arousal level may be particularly important during complex social interactions. This may be more relevant for species that exhibit male competition for

Chapter 2: Humpback whale social call production reflects both motivational state and arousal access to females during the breeding season. In these species, males may utilise several mating strategies for ensuring reproductive success, including direct physical conflict between competitors, the coercion of receptive females, or mate guarding (Clutton-Brock 1989). During all of these strategies, acoustic signals could be used as a way for a receiver to remotely assess the arousal, dominance status, and/or reproductive status of a conspecific [e.g. red deer, Cervus elaphus (Clutton-Brock and Albon 1979; Reby and McComb 2003); greater sac-winged bats, Saccopteryx bilineata (Eckenweber and Knörnschild 2013); chipping sparrows, Spizella passerina (Liu 2004); Galapagos sea lions, Zalophus wollebaeki (Kunc and Wolf 2008); rock hyrax, Procavia capensis (Koren and Geffen 2009); baboons (Fischer et al. 2004); and tree shrews, Tupaia belangeri (Schehka et al. 2007)]. The use of these cues becomes increasingly necessary for species who exhibit these complex breeding strategies in the marine environment where visual communication is limited. Knowledge of the mechanism of vocal signal production in baleen whales is limited (Fitch 2006; Adam et al. 2013; Girola et al. 2019). However, there are indications that flexible information is conveyed in their vocal output in a similar way to terrestrial species. For example, calls produced in high arousal breeding interactions in southern right whales, Eubalaena australis, and North Atlantic right whales, Eubalaena glacialis, have more frequency modulations and are produced at higher rates (Clark 1982; Parks and Tyack 2005; Parks et al. 2011). The vocalizations of minke whales, Balaenoptera acutorostrata, on the breeding grounds are longer in duration and have shorter inter-pulse intervals than those on the feeding grounds, which is speculated to be related to increased arousal during the breeding season (Risch et al. 2013). Lastly, humpback whales, Megaptera novaeangliae, engaged in male competition produce calls at higher rates (Silber 1986) and in bouts (Rekdahl et al. 2015). Humpback whales are well known for singing complex songs (Payne and McVay 1971; Winn and Winn 1978). However, they also exhibit a wide repertoire of non-song social sounds (hereafter ‘calls’) that are used to coordinate social behaviour, including breeding interactions (Silber 1986; Thompson et al. 1986; Dunlop et al. 2007, 2008; Dunlop 2016, 2017; Stimpert et al. 2011; Parks et al. 2014; Fournet et al. 2015). These interactions are highly variable in arousal, ranging from a single male 'escorting' a female, to larger agonistic competitive groups, where multiple males aggressively compete with each other for access to a female (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Mattila et al. 1989; Clapham et al. 1992; Clapham 1996; Herman et al. 2007; Felix and Novillo 2015). As the sex ratio along migration and on the breeding grounds is biased towards males (Brown and Corkeron 1995; Palsbøll et al. 1997), females are a relatively limited resource and face the risk of increased harassment (Jones 2010). Females

Chapter 2: Humpback whale social call production reflects both motivational state and arousal accompanied by new calves may experience even higher risks, as the presence of escorts can be costly by causing increased energy expenditure, decreased time spent nursing, temporary calf separation, and the risk of injury (Pack et al. 2002; Cartwright and Sullivan 2009; Jones 2010). This is supported by evidence of female-calf pairs actively avoiding potential escorts (Tyack and Whitehead 1983; Mobley, Jr. et al. 1988; Jones 2010; Indeck 2020). Therefore, the addition of one or multiple escorts is expected to alter the underlying motivation of the female and calf and increase arousal levels, and these changes are likely to be reflected in changes to the individual vocal production of group members. Recent evidence suggests competitive groups used a larger proportion of call types with acoustic structures indicative of high arousal and aggression (Dunlop 2017). This change in vocal behaviour provided preliminary evidence that humpback whales may use specific call types in addition to modifying call features to convey motivation and arousal. Here, we tested whether humpback whale female-calf pairs exhibit changes in their vocal production with the presence of one or more escorts. Specifically, we tested whether escorted groups exhibited a more variable call repertoire (i.e. used more call types), changed the acoustic structure of calls, increased call rates, and/or increased the use of long bouts of calls - all flexible acoustic features indicative of a change in motivation and/or arousal. These results may provide information on the potential function of some humpback whale calls.

2.3 Methods

Data were collected on eastern Australian humpback whales in 2010, 2011, 2014, and 2017 during their southbound migration using acoustic recording tags (Figure 2.1). In 2010, 2011, and 2014, DTAGs (Johnson and Tyack 2003) were attached to adult females with calves via suction cups and programmed to detach after 2 to 4 hours. DTAGs are equipped with a sensor suite that records acoustic signals, pressure (depth), temperature, acceleration, and magnetic field along 3 axes to provide 3-dimensional movement data on pitch, roll, and heading (Johnson and Tyack 2003). Tags were programmed to sample 16-bit audio at either a 48 kHz or 96 kHz sampling rate, with a high-pass filter asymptotic to a decrease of 6 dB per octave from 1 kHz down to 50 Hz, and 18 dB per octave below 50 Hz. In 2017, one Acousonde (Greeneridge Sciences, Inc.) was deployed on an adult female accompanied by a calf. As with DTAGs, the Acousonde is a suction cup acoustic digital recording tag with a similar sensor suite. This tag was programmed to sample 16-bit audio at a 25 kHz sampling rate.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Figure 2.1 Map of the primary study area in southern Queensland, Australia. The majority of data collection occurred just south of Peregian Beach.

For all years, boat-based focal follows were conducted on the tagged animal for the duration of the tag deployment to collect data on group composition (e.g. splitting and joining of escorts to the group) and surface behaviour. Group compositions included female-calf (FC) only groups, female-calf-single escort (FCE) groups, and female-calf-multiple escort (FCME) groups. The motivational states and arousal levels of the three group compositions were assumed to be different, with levels of arousal increasing with the addition of an escort, and increasing further with the presence of multiple escorts. Further, it was assumed motivational state would change between the three group compositions, with no indication of aggression in FC groups and high aggression in FCME groups. This was based on previous research demonstrating that FC groups spend less time resting and more time travelling with the addition of one or more escorts (Cartwright and Sullivan

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

2009; Jones 2010), that females and calves can become separated during larger FCME groups (Pack et al. 2002), and that animals in competitive groups display aggressive and combative behaviour (Tyack 1981; Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Mattila et al. 1989; Clapham et al. 1992; Clapham 1996; Herman et al. 2007; Felix and Novillo 2015). To eliminate confounding variables associated with changing social structure, only stable group associations (more than 10 minutes before another animal joined and/or after an animal split from the group) were used. The first ten minutes of the tag deployment were also excluded to allow for the tagged animal to return to ‘pre-tagging’ behaviour (Williamson et al. 2016). Spectrograms of recordings were browsed visually and aurally by an experienced analyst in Raven Pro 1.5 (Bioacoustics Research Program 2017) using a Hann window, Fast Fourier Transform size of either 4096 or 2048 samples (depending on the sampling rate), and 50% overlap. All humpback whale calls were marked and extracted for further analysis, as well as clips of surrounding noise (at least 0.5 seconds) either just prior to or after the call. Using a custom-written MATLAB script (MathWorks Inc., 2017), energy from the ambient environment was removed by subtracting the spectrum of the noise clips from that of the associated sounds (Girola et al. 2019). In order to correct for the high-pass filter, tags were calibrated prior to deployment at the Woronora Dam in New South Wales, Australia, using a standard reference hydrophone. The results showed a high pass filter in the DTAGs that was asymptotic to a decrease of 6 dB per octave from 1 kHz down to 50 Hz, and 18 dB per octave below this. From 1 – 2 kHz, DTAG sensitivity was approximately -170 dB V/µPa but decreased and became variable at frequencies above this. For every frequency value, the received sound level (RL) at the tag was calculated using the output voltage of the tag and the sensitivity of the tag as measured during calibration. Then the resulting values were used to correct for the roll-off of each tag recording in Matlab before measuring the calls. In total, 12 acoustic parameters were measured that were partially chosen based on call classification analyses from previous studies on humpback whale calls (Dunlop et al. 2007; Stimpert et al. 2011; Rekdahl et al. 2013) (Table 2.1). Frequency measurements were initially made on a linear scale, but were converted to a logarithmic scale for subsequent analyses to better account for the mammalian perception of pitch (Richardson et al. 1995; Cardoso 2013). To increase the confidence with which calls could be ascribed to the focal group, a signal-to- noise ratio (SNR) cut-off was used. This cut-off was based on a data simulation of a whale sound using the MATLAB script (Girola et al. 2019). A tonal sound was created in Matlab that was similar to a humpback whale ‘moan’ (i.e. a sine wave with a fundamental of 200 Hz, harmonics at multiple integers, and decreasing amplitude). White noise was created to simulate water turbulence, the most common background noise detected on the tags. Both the simulated moan and the white

Chapter 2: Humpback whale social call production reflects both motivational state and arousal noise were of known amplitude and intensity. The sound and the noise were combined and denoised following the techniques developed by Boll (1979) and outlined in Girola et al. (2019). The intensity of the denoised sound was then measured and compared to the intensity of the original tonal sound. This was repeated with decreasing SNR values until the SNR value was reached where the intensity of the denoised sound no longer matched the intensity of the original tonal sound. This signified that the denoising algorithm is no longer efficient and occurred at an SNR ≤ -3 dB. Although recent research has indicated an SNR higher than 20 dB may be necessary to confidently assign calls to the focal animal (Stimpert et al. 2020), identifying the focal animal was not necessary for this analysis and the cut-off of > -3 dB was used (Girola et al. 2019). Calls that were overlapping or interrupted by acute noise (e.g. whale song or tag rubbing) were also removed to preserve the accuracy of the measurements.

2.3.1 Call repertoire To test if groups used a different call repertoire with changes in group membership, calls were initially assigned a subjective call type classification based on qualitative features and previous catalogues for eastern Australian humpback whales (Dunlop et al. 2007, 2008; Rekdahl et al. 2013). To statistically validate the qualitative classifications, we used a recursive partitioning analysis with a classification and regression tree (CART) and a random forest (RF) model. Analyses were run in R (R Core Team 2018) with the packages rpart (Therneau and Atkinson 2018) and randomForest (Liaw and Wiener 2002). To build the classification trees, all 12 acoustic variables were included (Table 2.1). The acoustic features of some humpback whale calls are known to fall along a gradient rather than into discrete call types (Dunlop et al. 2007, 2008; Stimpert et al. 2011; Fournet et al. 2015). Therefore, the call types that performed poorly in the CART and RF (i.e. they did not form terminal nodes in the CART analysis and had the highest misclassification rates), were subsequently grouped with the call type to which they were mostly misclassified.

Table 2.1 A list of the 12 acoustic measurements and their associated abbreviations that were used in the call classification and cluster analyses. Measurement Abbreviation Description

Duration (s) Dur Total duration of the sound

Minimum frequency (Hz) FMIN Minimum frequency of the sound

Maximum frequency (Hz) FMAX Maximum frequency of the sound

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Peak frequency (Hz) FP Frequency of the spectral peak

Frequency midpoint dividing the sound Centre frequency (Hz) F C into two equal energy intervals Frequency at which 25% of the sound’s 1st quartile frequency (Hz) F Q1 energy occurs Frequency at which 75% of the sound’s 3rd quartile frequency (Hz) F Q3 energy occurs Difference between the 3rd and 1st quartile Inter-quartile bandwidth (Hz) F IQ frequencies Frequency at which 5% of the sound’s 5th centile frequency (Hz) F C05 energy occurs Frequency at which 95% of the sound’s 95th centile frequency (Hz) F C95 energy occurs Difference between the 95th and 5th centile Inter-centile bandwidth (Hz) F IC frequencies

Frequency trend (ratio) FTREND Start frequency/end frequency

As a test of whether there were any acoustic differences in calls between groups of different arousals, the call repertoire was considered as a whole, with all calls run through a k-means cluster analysis (MacQueen 1967). This method partitions the data (each call) into a specified number of clusters in such a way as to minimise the sum of squares from the data points to their assigned cluster centres. Each cluster should therefore contain calls that are similar in structure to each other. Data were scaled prior to analysis to account for the variable measurement units. The optimal number of clusters was determined using the package NbClust (Charrad et al. 2014), which proposes the ideal clustering scheme based on multiple indices. The same 12 acoustic measurements used in the previous analysis were included (Table 2.1). A non-hierarchical k-means cluster analysis was then run using the stats package. Separate analyses were run for each of the three group compositions, with the resulting clusters representing the acoustic endpoints of the repertoire for each group. Differences in the way that the clusters were formed (based on loadings, or relative importance of acoustic variables), or in the number of clusters between groups, represents changes in the acoustic features of calls and/or changes in the use of certain call types between the groups.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

2.3.2 Call structure Linear mixed-effects models (LMMs) were used (Cnaan et al. 1997) to investigate whether the observed change in the repertoire between group compositions was due to the addition of new call types or changes in acoustic features within call types. LMMs provide least squares means rather than observed means, which are adjusted to predict the effect of the factor variables on the response assuming equal sample sizes, and are therefore more accurate for unbalanced data (Harvey 1960). Separate LMMs for each call type were run using the packages lme4 (Bates et al. 2015) and emmeans (Lenth 2018). Three acoustic parameters that are known to encode motivational arousal in other species (Marler et al. 1992; Dunlop 2017; McGrath et al. 2017; Mandl et al. 2019) were selected as response variables: duration (Dur), peak frequency (FP), and inter-centile (90%) bandwidth (FIC). Group composition was considered the fixed-effect and Tag ID was added as a random effect to account for potential individual differences.

2.3.3 Call bouts To define call bouts, we first established the inter-call interval between all calls. We then used the bout end criterion (BEC) calculated by Rekdahl et al. (2015) of 3.9 seconds to designate whether a call was part of a bout or represented a single call. Due to the difficulties of assigning calls to individual animals, and the decreased likelihood of capturing all calls from a group with an increase in group membership, it was not possible to deduce whether any changes in call production were due to an increase in the number of animals calling or an increase in the number of calls each animal makes. Therefore, the data were standardised by dividing the number of calls detected by group size. Exploratory data analysis revealed that the mean number of calls per whale per bout was 1.29, however, there were several cases of extreme outliers that resulted in highly skewed data. As a result, the inter-quartile range was calculated and used to find the upper limit beyond which data points could be considered outliers. This led to the designation of single calls, short bouts (≤ 1.29 calls/whale), medium bouts (> 1.29 but < 2.75 calls/whale), and long bouts (> 2.75 calls/whale). To compare the proportion of bouts indicative of high arousal (long bouts), and low arousal (short bouts and single calls) in each group, we used a generalised linear mixed-effects model (GLMM) for proportions. Due to the nature of proportion data (non-constant variance, non-normal errors, and strictly bounded data) a binomial error distribution was used (Crawley 2013). Separate models were run for each bout type, with the internally calculated proportion of bouts within each tag record as the response variable, and group composition as the predictor variable. Tag ID was added as a random effect to account for repeated observations of individuals. GLMMs were run in R with the packages lme4 and emmeans.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

2.3.4 Call rate Standardised call rates (calls per hour per whale) were compared to investigate differences in the rate of call production. FCME groups were eliminated from statistical comparisons because of low sample size (three groups, Appendix 1, Table A1.1). Again, call rates per group were standardised by dividing the number of calls detected by group size. A GLMM with a quasi-poisson distribution was then used to compare call rates between the FC and FCE groups (Crawley 2013), with post-hoc analyses using emmeans. An offset was used to account for the variable amount of time of each tag deployment and to provide a comparison of rates rather than counts. Tag ID was included as a random effect. Observed differences in call type usage may be a result of changes in the proportion of the repertoire that certain calls make up and/or in the production rate of specific call types. To explore whether or not there was a change in the call rate of adult and/or calf calls specifically, two call types were selected for further analyses: ‘snorts’ and ‘calf calls’. These call types were detected in relatively high numbers in both group compositions. Additionally, ‘snorts’ are most likely produced by adult females (Indeck et al. 2020) potentially providing a representation of both female and calf calling behaviour. Separate GLMMs were run for each call type.

2.4 Results

A total of 3,309 calls were detected in stable groups of humpback whales from 26 tags, comprising nearly 62 hours of recordings. Of these 26 tags, 15 were deployed in female-calf (FC) only groups, eight were deployed in female-calf-escort groups (FCE), and the remaining three were deployed in female-calf-multiple escort groups (FCME) (Appendix 1, Table A1.1). As whales are migrating steadily through the area, it is assumed that each tag corresponds to a different individual (Cato et al. 2013). When calls were removed that overlapped, were interrupted by acute noise, or had an SNR of less than -3 dB, 3,029 were retained for further analyses: 330 calls from FC groups; 1,362 calls from FCE groups; and 1,337 calls from FCME groups.

2.4.1 Call type Subjective manual classification resulted in 16 discrete call types, all of which were included in the CART and RF analyses. The CART had a high misclassification rate (OOB estimate of error = 45.3%), as did the RF analysis (OOB estimate of error = 41.2%). Six call types performed poorly in the CART (did not form terminal nodes) and RF analyses (> 75% misclassification rate) and so were grouped with the calls to which they were most often misclassified, resulting in eight

Chapter 2: Humpback whale social call production reflects both motivational state and arousal updated sound types: ‘calf’ calls (Indeck et al. 2020), ‘low frequency’ calls, ‘modulated’ calls, ‘snorts’ , ‘paired croaks’, ‘knocks’, ‘spiccatos’, and ‘squeaks’ (Figure 2.2).

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

When the CART and RF analyses were rerun using these new call types, the misclassification rate decreased to 33.7% and 32.8%, respectively. Duration (Dur) was the most important variable for classification in both analyses, suggesting call types are initially split st th according to their duration. Additional important variables were 1 quartile frequency (FQ1) and 5 centile frequency (FC05). The acoustic measurements for the eight call types, as well as the number of each call type detected, are provided in Table 2.2.

Table 2.2 Mean ± SD of the acoustic measurements for the 8 identified call types, as well as the number of calls in parentheses. A list of the abbreviations can be found in Table 2.1.

Call Type Dur FMIN FMAX FP Calf (1,075) 0.15±0.07 332.0±275.8 2931.8±2407.9 534.9±435.9 Knock (244) 0.10±0.04 121.3±153.1 2676.6±2694.4 248.7±234.7 Low Frequency (373) 0.68±0.38 123.2±171.7 3082.1±2181.1 200.3±242.0 Modulated (228) 0.54±0.34 256.3±211.4 6157.5±5672.7 507.9±533.3 Paired Croaks (127) 0.41±0.06 77.4±51.0 1960.0±967.2 120.1±57.0 Snort (712) 0.20±0.09 157.2±234.2 2885.5±2312.4 270.9±334.7 Spiccato (70) 1.65±1.36 140.9±111.2 3391.7±2581.0 202.8±154.2 Squeak (200) 0.19±0.12 1157.0±853.9 5241.0±4418.6 1559.4±912.7

FC FQ1 FQ3 FIQ Calf 592.6±425.8 484.4±352.8 738.0±508.8 253.5±268.9 Knock 287.0±237.7 224.3±212.0 370.2±262.6 145.9±107.7 Low Frequency 256.9±269.5 194.9±228.8 345.5±334.9 150.6±157.2 Modulated 584.0±516.8 456.5±419.9 820.5±688.4 364.0±442.8 Paired Croaks 154.3±48.3 118.2±44.8 201.2±55.3 83.0±34.7 Snort 321.8±356.8 249.0±298.5 434.1±460.7 185.1±223.8 Spiccato 258.8±162.6 202.8±136.7 343.9±202.8 141.1±101.8 Squeak 1652.7±998.3 1456.0±922.1 1907.4±1193.9 451.4±597.3

FC05 FC95 FIC FTREND Calf 386.2±307.3 1096.7±705.5 710.5±565.2 1.32±1.04 Knock 160.8±189.3 598.9±386.2 438.1±303.0 1.44±1.02 Low Frequency 144.9±196.3 600.8±537.7 455.9±441.1 1.91±5.00 Modulated 318.3±260.3 1322.6±892.3 1004.2±785.4 1.42±2.33 Paired Croaks 88.4±47.6 308.5±135.6 220.1±140.8 1.75±4.80 Snort 184.8±257.5 727.1±671.1 542.3±507.8 1.86±5.02 Spiccato 160.1±120.3 623.2±392.3 463.1±305.9 1.76±4.12 Squeak 1277.6±875.6 2373.4±1836.5 1095.9±1482.0 1.02±0.58

The proposed number of clusters for FC and FCE groups was two, increasing to three in FCME groups (Figure 2.3). The number of calls that fell into each cluster, as well as the proportion

Chapter 2: Humpback whale social call production reflects both motivational state and arousal of the cluster repertoire they represent, can be found in Table 2.3. For all three group compositions, Components 1 and 2 explained over 70% of the variability. Frequency measurements and bandwidth loaded into both components, and Dur loaded only into component 2. For FC and FCE groups, calls in cluster 1 were generally shorter in duration, higher in frequency, and more broadband than those in cluster 2. The cluster output changed, however, for FCME groups. Here, an additional third cluster was formed that was characterised by long average duration (Figure 2.3). Two call types in particular were responsible for the formation of this third cluster: ‘paired croaks’ and ‘spiccatos’. ‘Paired croaks’ were rarely detected with only a single escort present (3 occurrences), however their use increased significantly with the addition of more than one escort (Figure 2.2f). ‘Spiccatos’, while rare (n = 70), were only detected in the presence of multiple escorts (Figure 2.2i). Both of these call types are relatively low in frequency and bandwidth measurements compared with the other call types. In addition, ‘spiccatos’ and sequences of paired croaks were the longest in duration of any of the call types. ‘Spiccatos’ were on average 1.65 ± 1.36 seconds in duration (range 0.25 to 7.07 seconds) and paired croaks were only produced in sequences (range 2 to 18 pairs), with some series up to 11.7 seconds long. The means for frequency and bandwidth measurements of calls in cluster 3 fell between clusters 1 and 2 (Appendix 1, Table A1.2).

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Table 2.3 Number of calls, broken down by call type, that comprised each cluster. The proportion of the cluster that each call type represents is in parentheses. FC: female-calf pair, FCE: female- calf-escort group, FCME: female-calf-multiple escort group. Call Type Group Cluster Calf Knock Low Mod. Paired Snort Spiccato Squeak freq. croaks 1 119 1 13 11 8 19 FC 0 0 (N=171) (0.70) (0.0) (0.08) (0.06) (0.05) (0.11) 2 40 4 41 7 65 FC 0 0 2 (0.01) (N=159) (0.25) (0.03) (0.26) (0.04) (0.41)

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

1 272 16 20 109 130 123 FCE 0 0 (N=670) (0.41) (0.02) (0.03) (0.16) (0.19) (0.19) 2 126 64 171 44 3 281 3 FCE 0 (N=692) (0.18) (0.10) (0.25) (0.06) (0.0) (0.41) (0.0) 1 201 7 17 10 17 5 48 FCME 0 (N=305) (0.66) (0.02) (0.05) (0.03) (0.06) (0.02) (0.16) 2 277 86 36 36 13 75 36 4 FCME (N=563) (0.16) (0.18) (0.12) (0.04) (0.20) (0.24) (0.06) (0.0) 3 40 66 75 11 111 136 29 1 FCME (N=469) (0.09) (0.14) (0.16) (0.02) (0.24) (0.29) (0.06) (0.0)

2.4.2 Call structure

There were few significant differences in the acoustic parameters (Dur, FP, and FIC) within call types between the three group compositions (Table 2.4). This indicates that changes in the way calls were clustered in FCME groups were due to new call types being used rather than changes to the structure of existing calls in the repertoire. However, there were some exceptions. Duration was significantly different between only two call types, with the duration of ‘squeaks’ longer in FC (0.36 ± 0.04) compared to FCE (0.28 ± 0.04, p = 0.0316) groups, and shorter in ‘low frequency’ calls of FC groups (0.55 ± 0.05) compared to FCE groups (0.72 ± 0.04, p = 0.0285) (Table 2.4).

There were significant differences in the FIC of ‘calf’ calls between FC (2.64 ± 0.03) and FCME groups (2.80 ± 0.06, 9=0.0448); ‘knocks’ between FC (2.23 ± 0.12) and both FCE (2.58 ± 0.04, p = 0.0156) and FCME (2.67 ± 0.07, p = 0.0009) groups; and ‘snorts’ between FC (2.47 ± 0.05) and FCE (2.60 ± 0.04, p = 0.0279) groups. In all cases, calls in FC groups were more narrow-band compared to groups with one or more escort (Table 2.4). There were no significant differences in FP in any call type between the three group compositions.

Table 2.4 Results of the linear mixed models for the call types and parameters that were significantly different between the group compositions. There were no significant differences between FCE and FCME groups. An asterisk indicates statistical significance at the p < 0.05 level. A list of the abbreviations can be found in Table 2.1. FC: female-calf pair, FCE: female-calf-escort group, FCME: female-calf-multiple escort group. FC FCE FCME FC-FCE FC-FCME Call Type Parameter (mean ± SE) (mean ± SE) (mean ± SE) (est. ± SE) (est. ± SE) -0.10 ± 0.04 -0.15 ± 0.06 Calf FIC 2.64 ± 0.03 2.74 ± 0.04 2.80 ± 0.06 t ratio = -2.15 t ratio = -2.49 p = 0.0860 p = 0.0448* -0.34 ± 0.12 -0.44 ± 0.12 Knock FIC 2.23 ± 0.12 2.58 ± 0.04 2.67 ± 0.07 t ratio = -2.85 t ratio = -3.68 p = 0.0156* p= 0.0009*

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

-0.17 ± 0.07 -0.10 ± 0.08 Low Dur 0.55 ± 0.05 0.72 ± 0.04 0.65 ± 0.06 t ratio = -2.67 t ratio = -1.25 Frequency p = 0.0285* p = 0.4543 -0.14 ± 0.05 -0.12 ± 0.09 Snort FIC 2.47 ± 0.05 2.60 ± 0.04 2.58 ± 0.08 t ratio = -2.59 t ratio = -1.30 p = 0.0279* p = 0.4013 0.07 ± 0.03 0.19 ± 0.08 Squeak Dur 0.36 ± 0.04 0.28 ± 0.04 0.16 ± 0.07 t ratio = 2.49 t ratio = 2.29 p = 0.0316* p = 0.0868

2.4.3 Call bouts A total of 526 bouts were detected, 366 of which were short (≤ 1.29 calls/whale), 117 of which were of medium length (> 1.29 and < 2.75 calls/whale) and 43 of which were long (> 2.75 calls/whale). In general, calls were more often produced in bouts (2,264 calls, 75%) than as single calls (765 calls, 25%), agreeing with Rekdahl et al. (2015). However, this result was not consistent across group composition. FC only groups produced more single calls (62%) than calls in bouts. With the addition of one or more escorts, groups shifted from using single calls to producing calls in bouts, with only 28% of calls produced singly in FCE groups and 13% in FCME groups. FCME groups used a significantly higher number of long bouts (32%) compared to FC (7%) and FCE (15%) groups (Table 2.5).

Table 2.5 Results of the generalised linear models with the proportion of each call bout type in each group composition given in the first three columns and the pairwise contrasts (comparisons) in the last three columns. A negative estimate of the pairwise contrasts and z-ratio indicates there is a lower probability of that call type occurring in the first of the two group compositions listed. An asterisk indicates statistical significance at the p < 0.05 level. FC: female-calf pair, FCE: female- calf-escort group, FCME: female-calf-multiple escort group. Bout FC FCE FCME FC-FCE FC-FCME FCE-FCME Type (prop. ±SE) (prop .±SE) (prop.± SE) (est. ± SE) (est. ± SE) (est. ± SE) 1.43 ± 0.13 2.39 ± 0.14 1.0 ± 0.10 Single 0.62 ± 0.03 0.28 ± 0.01 0.13 ± 0.01 z ratio = 11.1 z ratio = 17.1 z ratio = 9.53 p < 0.0001* p < 0.0001* p < 0.0001* -0.48 ± 0.15 -0.52 ± 0.15 -0.05 ± 0.08 Short 0.22 ± 0.02 0.31 ± 0.01 0.32 ± 0.01 z ratio = -3.25 z ratio = -3.58 z ratio = -0.57 p = 0.0030* p = 0.0009* p = 0.8297 -1.24 ± 0.20 -1.11 ± 0.20 0.13 ± 0.09 Medium 0.09 ± 0.02 0.26 ± 0.01 0.23 ± 0.01 z ratio = -6.18 z ratio = -5.51 z ratio = 1.47 p < 0.0001* p < 0.0001* p = 0.2907 -0.84 ± 0.22 -1.79 ± 0.22 -0.95 ± 0.10 Long 0.07 ± 0.01 0.15 ± 0.01 0.32 ± 0.01 z ratio = -3.73 z ratio = -8.13 z ratio = -9.97 p = 0.0005* p < 0.0001* p < 0.0001*

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

2.4.4 Call rate FC only groups had a relatively low overall call rate (estimate = 4.5 ± 2.6 calls/hr), which increased significantly with the addition of an escort (estimate = 25.0 ± 8.7 calls/hr, p = 0.0110). The standardised call rates of ‘snorts’ were significantly higher in FCE groups (estimate = 7.5 ± 2.5 calls/hr) than FC only groups (0.96 ± 0.6 calls/hr, p = 0.0050), as were the standardised rates of ‘calf’ calls (FCE estimate = 7.3 ± 2.3 calls/hr, FC estimate = 2.1 ± 0.9 calls/hr, p = 0.0174).

2.5 Discussion

Humpback whales exhibit a diverse range of social interactions during the breeding season, including females alone with their calves, accompanied by a male escort, or the focus of a competitive group. Previous research indicates that female-calf pairs attempt to avoid male escorts (Tyack and Whitehead 1983; Mobley, Jr. et al. 1988; Jones 2010; Indeck 2020). If unsuccessful, the addition of escorts changes the behaviour of these pairs, potentially increasing energy expenditure and the potential for calf separation or injury (Pack et al. 2002; Cartwright and Sullivan 2009; Jones 2010). Recently, it has been suggested that flexible information is likely contained within acoustic cues during encounters with escorts, which may reflect these changes in motivational state or arousal (Dunlop 2017). Here, we provide further evidence that flexible information is potentially conveyed in acoustic features through the use of specific call types, increases in call rates, the use of long bouts of calls, and changes to some call features. These results indicate that humpback whales use multiple ‘strategies’ to communicate flexible information by modifying their vocal production. Female-calf (FC) pairs of humpback whales produced significantly less calls/hour/whale than female-calf-escort (FCE) groups and used a significantly higher proportion of single calls as opposed to bouts of calls. The call rate of ‘snorts’ and ‘calf’ calls also significantly increased with group membership. These results indicate that the low rate of signal production in FC groups could, in part, be the result of low arousal, as low call rates indicate low arousal in many species. This is a reasonable assumption considering the predominance of resting and slow travelling in lone FC pairs observed in this and other habitats (Cartwright and Sullivan 2009; Indeck 2020, this study). However, FC groups of humpback whales (Dunlop et al. 2008; Videsen et al. 2017; Indeck et al. 2020), southern right whales (Nielsen et al. 2019), and North Atlantic right whales (Cusano et al. 2018; Parks et al. 2019) have all been documented to produce calls at a very low rate, potentially as a method of avoiding detection by predators or conspecifics. Alternatively, it could indicate a reduced need to communicate when the pair are together, as females and calves are known to increase call rates when separated (Indeck 2020). It is likely that a combination of factors is thus responsible for the low call production in FC groups.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

In the present study, increased group membership was also correlated with changes to the acoustic features of some call types, although not always in a predictable way. It was predicted that here, as in other species, increased duration would be correlated with increased arousal and social complexity. The duration of ‘squeaks’, however, decreased with the addition of escorts. In several species, the duration of alarm calls is shorter in situations of increased urgency or arousal (Blumstein and Arnold 1995; Manser 2001), or compared to calls produced in low arousal contexts (Fischer et al. 2001). This could function to reduce conspicuousness to predators (Briefer 2012). Additionally, the duration of screams produced by chimpanzees, Pan troglodytes, are generally shorter in high arousal agonistic situations (Siebert and Parr 2003). However, this is variable depending on the sex of the caller and whether they are pursuing or being chased (i.e. the aggressor or victim), with longer screams produced by females being chased. Therefore, it is likely that both motivational state and arousal affect the duration of calls, albeit in potentially different ways. Further, while there is a general trend of broader bandwidth observed in many other species for situations of increased arousal, here ‘calf’ calls, ‘knocks’, and ‘snorts’ were all shown to be more narrow-band in the presence of one or more escorts. Therefore, while arousal appears to play a large role in the results presented here, there are undoubtedly other contributing factors to the observed changes, such as behavioural and motivational state. For example, narrow-band calls are often correlated with ‘fearful’, ‘appeasing’, or ‘aversive’ motivational contexts (Morton 1977; August and Anderson 1987; Briefer 2012). Multiple escort groups of humpback whales are often engaged in direct competition (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Clapham 1996), and there are elevated levels of aggression as evidenced by an increased number of agonistic behaviours (i.e. tail slashes, lunges, and direct body contact) and visible fresh wounds. In addition, it has been shown that females with calves in competitive groups experience higher energetic costs (Jones 2010; Craig et al. 2014), including an increased risk of injury to, and separation from, the calf (Baker and Herman 1984). These groups are therefore likely to have individuals in not only aggressive motivational states, but fearful/aversive states as well. Indeed, Dunlop (2017) provided evidence of not only increased aggression in the acoustic features of calls detected in larger groups of humpback whales in this area, but features indicative of fear. The observed trend of decreasing bandwidth with the addition of escorts may therefore be more correlated with motivational state than arousal. Unfortunately, in the present study, it is unclear which animal is producing the sounds. The ability to reliably assign calls to individuals, and the inclusion of underwater video, will be one of the only clear ways to begin distinguishing between the effects of these two related attributes.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

While few acoustic features were modified across group composition, the results of the cluster analysis clearly showed a difference in the repertoire of FCME groups compared to FC and FCE groups with the addition of a third cluster of calls. This could indicate that the changes in the acoustic behaviour of FCME groups were primarily the result of the increased use or introduction of particular call types. In particular, three call types exhibited some specificity in call production. One such call type were ‘knocks’, which occurred almost exclusively in the presence of an escort, and increased in use with the number of escorts. These signals are relatively low frequency and very short in duration. ‘Knocks’ are acoustically similar to the short, low amplitude, pulsed calls produced by North Atlantic right whale FC pairs (Parks et al. 2019) and humpback whale calves in Hawai’i (Zoidis et al. 2008). It was proposed that these call types may serve as a cryptic way for females and their calves to remain in acoustic contact. While ‘knocks’ were relatively infrequent in FC only groups (five occurrences, < 2% of the FC repertoire), the use of this call type increased in FCE (6% of the repertoire) and FCME groups (12% of the repertoire). If the ‘knocks’ described here function in a similar way to pulsed calls of right whale FC pairs, this increase in use could reflect the need for maintained contact between a female and her calf during high arousal activity. Further, ‘knocks’ increased in bandwidth as escorts joined FC pairs. If an increase in bandwidth increases the detectability of a call, this change in the bandwidth of ‘knocks’ provides additional support that this is a contact call between females and calves. Two additional call types were detected only in the presence of escorts. ‘Spiccatos’ were only produced in FCME groups. ‘Paired croaks’ were almost always produced in FCME groups, with only 3 occurrences in FCE groups and none detected in FC only groups. Both call types were distinct in their temporal features. ‘Spiccatos’ were the longest in duration of any of the call types (up to 5.8 seconds), while ‘paired croaks’ sequences were up to 11.7 seconds in duration. As vocal signals emitted in high arousal contexts are generally longer and produced at higher rates, the use of these long call types almost exclusively in FCME groups strongly indicates a potential function in conveying high arousal. Further, they are both relatively low in frequency, indicating both ‘spiccatos’ and ‘paired croaks’ may function to specifically communicate an aggressive motivational state. Although historically the benefit of signalling a state of high arousal and/or aggression has been contested (Maynard Smith 1982), it is clear that many conflicts are resolved without direct fighting (Zahavi 1982). Fixed features like size may not always predict the winner in an escalated contest, and smaller animals with higher levels of motivation and/or arousal are sometimes able to dominate larger opponents (Wagner 1989; Kotiaho et al. 1999; Hofmann and Schildberger 2001). ‘Spiccatos’ and ‘paired croaks’ may function in a similar way, providing information on intent to engage or continue engaging as conflict escalates beyond threats.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Compared with other baleen whales, humpbacks are well known for having a wide and variable repertoire of non-song calls (Silber 1986; Dunlop et al. 2007, 2008; Stimpert et al. 2011; Fournet et al. 2015; Rekdahl et al. 2017). Further, it has been suggested that the repertoire of humpback whales likely contains calls that are both discrete (i.e. each call type acoustically distinct) and graded (i.e. one call type blending into the next) (Dunlop et al. 2007, 2008; Stimpert et al. 2011; Fournet et al. 2015, 2018; Dunlop 2017; Rekdahl et al. 2017). However, it is currently unclear why such a complex communication system has evolved in this species. Here, we have demonstrated that humpback whales may encode arousal and motivational information in call type, as well as through changes in call features and call production. Combined with past research, our results begin to help explain the large number of call types that have been catalogued in this species, as well as how flexible information is communicated. Further studies should consider behaviour on a finer scale (i.e. behavioural states, splitting and joining of escorts, speed, number of aggressive behaviours) concurrent with arousal information to parse apart context-related changes in vocal features. Additionally, effort should be made to characterise whether calls are discrete or graded, as this potentially reveals more biologically relevant information (Insley et al. 2003; Fischer et al. 2017). This will help to create a more complete picture of the functional significance of specific call types and begin to explain the complex communication system evident in this species.

2.6 References

Adam O, Cazau D, Gandilhon N, Fabre B, Laitman JT, Reidenberg JS (2013) New acoustic model for humpback whale sound production. Applied Acoustics 74:1182–1190. Adler NT (1979) On the Physiological Organization of Social Behavior: Sex and Aggression. In: Marler P, Vandenbergh JG (eds) Social Behavior and Communication. Plenum Press, New York (NY), pp 29–72. August P V., Anderson JGT (1987) Mammal sounds and motivation-structural rules: a test of the hypothesis. Journal of Mammalogy 68:1–9. Baker CS, Herman LM (1984) Aggressive behavior between humpback whales (Megaptera novaeangliae) wintering in Hawaiian waters. Canadian Journal of Zoology 62:1922–1937. Bastian A, Schmidt S (2008) Affect cues in vocalizations of the bat, Megaderma lyra, during agonistic interactions. The Journal of the Acoustical Society of America 124:598–608. Bates D, Mächler M, Bolker B, Walker S (2015) Fitting Linear Mixed-Effects Models using lme4. Journal of Statistical Software 67:1–48. Bioacoustics Research Program (2017) Raven Pro: Interactive Sound Analysis Software. Cornell Laboratory of Ornithology, Ithaca (NY).

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Blumstein DT, Arnold W (1995) Situational specificity in alpine-marmot alarm communication. Ethology 100:1–13. Boll SF (1979) Suppression of Acoustic Noise in Speech Using Spectral Subtraction. IEEE Transactions on Acoustics, Speech, and Signal Processing 27:113–120. Bradley MM (2000) Emotion and motivation. In: Cacioppo JT, Tassinary LG, Berntson GG (eds) Handbook of psychophysiology, 2nd edn. Cambridge University Press, pp 602–642. Briefer EF (2012) Vocal expression of emotions in mammals: Mechanisms of production and evidence. Journal of Zoology 288:1–20. Brown M, Corkeron P (1995) Pod characteristics of migrating humpback whales (Megaptera novaeangliae) off the East Australian coast. Behaviour 132:163–179. Cardoso GC (2013) Using frequency ratios to study vocal communication. Animal Behaviour 85:1529–1532. Cartwright R, Sullivan M (2009) Associations with multiple male groups increase the energy expenditure of humpback whale (Megaptera novaeangliae) female and calf pairs on the breeding grounds. Behaviour 146:1573–1600. Cato DH, Noad MJ, Dunlop RA, et al (2013) A study of the behavioural response of whales to the noise of seismic air guns, design, methods and progress. Acoustics Australia 41:88–97. Charrad M, Ghazzali N, Boiteau V, Niknafs A (2014) NbClust: an R package for determining the relevant number of clusters in a data set. Journal of Statistical Software 61:1–36. Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mammal Review 26:27–49. Clapham PJ, Palsboll PJ, Mattila DK, Vasquez O (1992) Composition and dynamics of humpback whale competitive groups in the West Indies. Behaviour 122:182–194. Clark CW (1982) The acoustic repertoire of the Southern right whale, a quantitative analysis. Animal Behaviour 30:1060–1071. Clutton-Brock TH (1989) Mammalian mating systems. Proceedings of the Royal Society B: Biological Sciences 236:339–372. Clutton-Brock TH, Albon SD (1979) The roaring of red deer and the evolution of honest advertisement. Behaviour 69:145–170. Cnaan A, Laird NM, Slasor P (1997) Tutorial in biostatistics: using the general linear mixed model to analyse unbalanced repeated measures and longitudinal data. Statistics in Medicine 16:2349–2380. Colgan P (1989) Animal Motivation. Chapman and Hall, London. Craig AS, Herman LM, Pack AA, Waterman JO (2014) Habitat segregation by female humpback

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

whales in Hawaiian waters: Avoidance of males? Behaviour 151:613–631. Crawley MJ (2013) The R Book, 2nd edn. John Wiley & Sons, Ltd., Chichester (UK). Cusano DA, Conger LA, van Parijs SM, Parks SE (2018) Implementing conservation measures for the North Atlantic right whale: considering the behavioral ontogeny of mother-calf pairs. Animal Conservation 22:1–10. Dunlop RA (2017) Potential motivational information encoded within humpback whale non-song vocal sounds. The Journal of the Acoustical Society of America 141:2204–2213. Dunlop RA, Cato DH, Noad MJ (2008) Non-song acoustic communication in migrating humpback whales (Megaptera novaeangliae). Marine Mammal Science 24:613–629. Dunlop RA, Noad MJ, Cato DH, Stokes DM (2007) The social vocalization repertoire of east Australian migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 122:2893–2905. Eckenweber M, Knörnschild M (2013) Social influences on territorial signaling in male greater sac- winged bats. Behavioral Ecology and Sociobiology 67:639–648. Feighny JA, Williamson KE, Clarke JA (2006) North American elk bugle vocalizations: male and female bugle call structure and context. Journal of Mammalogy 87:1072–1077. Felix F, Novillo J (2015) Structure and dynamics of humpback whales competitive groups in Ecuador. Animal Behavior and Cognition 2:56–70. Fernald L, Fernald P (1978) Introduction to psychology, 4th edn. Houghton Mifflin, Boston (MA). Fichtel C, Hammerschmidt K (2002) Responses of redfronted lemurs to experimentally modified alarm calls: Evidence for urgency-based changes in call structure. Ethology 108:763–777. Fichtel C, Hammerschmidt K (2003) Responses of squirrel monkeys to their experimentally modified mobbing calls. The Journal of the Acoustical Society of America 113:2927–2932. Fischer J (2011) Where is the information in animal communication? Animal Thinking: Contemporary Issues in Comparative Cognition 151–161. Fischer J, Hammerschmidt K, Cheney DL, Seyfarth RM (2001) Acoustic features of female chacma baboon barks. Ethology 107:33–54. Fischer J, Hammerschmidt K, Todt D (1995) Factors affecting acoustic variation in Barbary‐ macaque (Macaca sylvanus) disturbance calls. Ethology 101:51–66. Fischer J, Kitchen DM, Seyfarth RM, Cheney DL (2004) Baboon loud calls advertise male quality: Acoustic features and their relation to rank, age, and exhaustion. Behavioral Ecology and Sociobiology 56:140–148. Fischer J, Price T (2017) Meaning, intention, and inference in primate vocal communication. Neuroscience and Biobehavioral Reviews 82:22–31.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Fischer J, Wadewitz P, Hammerschmidt K (2017) Structural variability and communicative complexity in acoustic communication. Animal Behaviour 134:229–237. Fitch T (2006) Production of vocalizations in mammals. In: Brown K (ed) Encyclopedia of language and linguistics. Elsevier Ltd, Oxford (UK), pp 115–121. Fournet ME, Szabo A, Mellinger DK (2015) Repertoire and classification of non-song calls in Southeast Alaskan humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 137:1–10. Fournet MEH, Jacobsen L, Gabriele CM, Mellinger DK, Klinck H (2018) More of the same: allopatric humpback whale populations share acoustic repertoire. PeerJ 6:e5365. Gerhardt HC (1992) Multiple messages in acoustic signals. Seminars in Neuroscience 4:391–400. Girola E, Noad MJ, Dunlop RA, Cato DH (2019) Source levels of humpback whales decrease with frequency suggesting an air-filled resonator is used in sound production. The Journal of the Acoustical Society of America 145:869–880. Gouzoules H, Gouzoules S (2000) Agonistic screams differ among four species of macaques: The significance of motivation-structural rules. Animal Behaviour 59:501–512. Green S, Marler P (1979) The Analysis of Animal Communication. In: Social Behavior and Communication. Plenum Press, New York (NY), pp 73–158. Gros-Louis JJ, Perry SE, Fichtel C, Wikberg E, Gilkenson H, Wofsy S, Fuentes A (2008) Vocal repertoire of Cebus capucinus: Acoustic structure, context, and usage. International Journal of Primatology 29:641–670. Harvey W (1960) Least-squares analysis of data with unequal subclass numbers. Agricultural Research Service, United States Dept. of Agriculture, Washington (DC). Hauser MD, Marler P (1993) Food-associated calls in rhesus macaques (Macaca mulatta): I. Socioecological factors. Behavioral Ecology 4:194–205. Herman EYK, Herman LM, Pack AA, Marshall G, Shepard CM, Bakhtiari M (2007) When whales collide: CRITTERCAM offers insight into the competitive behavior of humpback whales on their Hawaiian wintering grounds. Marine Technology Society 41:35–43. Hofmann HA, Schildberger K (2001) Assessment of strength and willingness to fight during aggressive encounters in crickets. Animal Behaviour 62:337–348. Indeck KL (2020) Acoustic communication of female-calf humpback whales during migration. PhD Thesis. The University of Queensland, Queensland. Indeck KL, Girola E, Torterotot M, Noad MJ, Dunlop RA (2020) Adult female-calf acoustic communication signals in migrating east Australian humpback whales. Bioacoustics Insley SJ, Phillips A V., Charrier I (2003) A review of social recognition in pinnipeds. Aquatic

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Mammals 29:181–201. Johnson MP, Tyack PL (2003) A digital acoustic recording tag for measuring the response of wild marine mammals to sound. IEEE Journal of Oceanic Engineering 28:3–12. Jones ME (2010) Female humpback whale (Megaptera novaeangliae) reproductive class and male- female interactions during the breeding season. PhD Thesis. Antioch University New England, Keene (NH). Kleinginna PR, Kleinginna AM (1981) A categorized list of emotion definitions, with suggestions for a consensual definition. Motivation and Emotion 5:345–379. Koren L, Geffen E (2009) Complex call in male rock hyrax (Procavia capensis): A multi- information distributing channel. Behavioral Ecology and Sociobiology 63:581–590. Kotiaho JS, Alatalo R V., Mappes J, Parri S (1999) Honesty of agonistic signalling and effects of size and motivation asymmetry in contests. Acta Ethologica 2:13–21. Kunc HP, Wolf JBW (2008) Seasonal changes of vocal rates and their relation to territorial status in male Galápagos sea lions (Zalophus wollebaeki). Ethology 114:381–388. Lenth R V. (2018) emmeans: estimated marginal means, aka least-squares means. R Core Team. Liaw A, Wiener M (2002) Classification and Regression by randomForest. R News 2:18–22. Liu WC (2004) The effect of neighbours and females on dawn and daytime singing behaviours by male chipping sparrows. Animal Behaviour 68:39–44. MacQueen J (1967) Some methods of classification and analysis of multi-variate observations. In: Proceedings of the Fifth Berkeley symposium in mathematical statistics and probability. pp 281–297. Mandl I, Schwitzer C, Holderied M (2019) Sahamalaza sportive lemur, Lepilemur sahamalaza, vocal communication: Call use, context and gradation. Folia Primatologica 90:336–360. Manser MB (2001) The acoustic structure of suricates’ alarm calls varies with predator type and the level of response urgency. Proceedings of the Royal Society B: Biological Sciences 268:2315– 2324. Manser MB (2010) The generation of functionally referential and motivational vocal signals in mammals. In: Brudzynski SM (ed) Handbook of Mammalian Vocalization - an integrative neuroscience approach. Academic Press, London (UK), pp 477–486. Marler P (1961) The logical analysis of animal communication. Journal of Theoretical Biology 1:295–317. Marler P (1976) Social organization, communication, and graded signals: The chimpanzee and the gorilla. In: Bateson PPG, Hinde RA (eds) Growing Points in Ethology. Cambridge University Press, Oxford (UK), pp 239–277.

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

Marler P, Evans CS, Hauser MD (1992) Animal signals: Motivational, referential, or both? In: Papousek H, Jurgens U, Papousek M (eds) Nonverbal vocal communication: comparative and developmental approaches. Cambridge University Press, Cambridge, UK, pp 66–86. Mattila DK, Clapham PJ, Katona SK, Stone GS (1989) Population composition of humpback whales, Megaptera novaeangliae, on Silver Bank, 1984. Canadian Journal of Zoology 67:281– 285. Maynard Smith J (1982) Do animals convey information about their intentions? Journal of Theoretical Biology 97:1–5. McGrath N, Dunlop R, Dwyer C, Burman O, Phillips CJC (2017) Hens vary their vocal repertoire and structure when anticipating different types of reward. Animal Behaviour 130:79–96. Mendl M, Burman OHP, Paul ES (2010) An integrative and functional framework for the study of animal emotion and mood. Proceedings of the Royal Society B: Biological Sciences 277:2895–2904. Mobley, Jr. JR, Herman LM, Frankel AS (1988) Responses of wintering humpback whales (Megaptera novaeangliae) to playback of recordings of winter and summer vocalizations and of synthetic sound. Behavioral Ecology and Sociobiology 23:211–223. Morton ES (1977) On the occurrence and significance of motivation-structural rules in some bird and mammal sounds. The American Naturalist 111:855–869. Morton ES (1982) Grading, discreteness, redundancy, and motivation-structural rules. In: Kroodsma DE, Miller MH (eds) Acoustic communication in birds. Academic Press, New York (NY), pp 183–212. Nielsen MLK, Bejder L, Videsen SKA, Christiansen F, Madsen PT (2019) Acoustic crypsis in southern right whale mother – calf pairs: infrequent, low-output calls to avoid predation? Journal of Experimental Biology 222:1–6. Owings D, Morton ES (1998) Animal Vocal Communication: A New Approach. Cambridge University Press, Cambridge (UK). Pack AA, Herman LM, Craig AS, Spitz SS, Deakos MH (2002) Penis extrusions by humpback whales (Megaptera novaeangliae). Aquatic Mammals 28:131–146. Palsbøll PJ, Allen J, Bérubé M, et al (1997) Genetic tagging of humpback whales. Nature 388:767– 769. Parks SE, Cusano DA, Parijs SM Van, Nowacek DP (2019) North Atlantic right whale (Eubalaena glacialis) acoustic behavior on the calving grounds. JASA Express Letters 146:15–21. Parks SE, Searby A, Célérier A, Johnson MP, Nowacek DP, Tyack PL (2011) Sound production behavior of individual North Atlantic right whales: implications for passive acoustic

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

monitoring. Endangered Species Research 15:63–76. Parks SE, Tyack PL (2005) Sound production by North Atlantic right whales (Eubalaena glacialis) in surface active groups. The Journal of the Acoustical Society of America 117:3297–3306. Payne RS, McVay S (1971) Songs of humpback whales. Science 173:585–597. R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Reby D, McComb K (2003) Anatomical constraints generate honesty: Acoustic cues to age and weight in the roars of red deer stags. Animal Behaviour 65:519–530. Rekdahl ML, Dunlop RA, Goldizen AW, Garland EC, Biassoni N, Miller P, Noad MJ (2015) Non- song social call bouts of migrating humpback whales. The Journal of the Acoustical Society of America 137:3042–3053. Rekdahl ML, Dunlop RA, Noad MJ, Goldizen AW (2013) Temporal stability and change in the social call repertoire of migrating humpback whales. The Journal of the Acoustical Society of America 133:1785–1795. Rekdahl ML, Tisch C, Cerchio S, Rosenbaum H (2017) Common nonsong social calls of humpback whales (Megaptera novaeangliae) recorded off northern Angola, southern Africa. Marine Mammal Science 33:365–375. Richardson WJ, Greene, Jr. CR, Malme CI, Thomson DH (1995) Marine Mammals and Noise. Academic Press, San Diego (CA). Risch D, Clark CW, Dugan PJ, Popescu M, Siebert U, Van Parijs SM (2013) Minke whale acoustic behavior and multi-year seasonal and diel vocalization patterns in Massachusetts Bay, USA. Marine Ecology Progress Series 489:279–295. Roughley N (2016) Wanting and Intending: Elements of a Philosophy of Practical Mind. Springer, New York. Schehka S, Esser KH, Zimmermann E (2007) Acoustical expression of arousal in conflict situations in tree shrews (Tupaia belangeri). Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 193:845–852. Scherer KR (2003) Vocal communication of emotion: A review of research paradigms. Speech Communication 40:227–256. Scherer KR (1986) Vocal affect expression: A review and model for further research. Psychological Bulletin 99:143–165. Schrader L, Todt D (1993) Contact call parameters covary with social context in common marmosets, Callithrix J. Jacchus. Animal Behaviour 46:1026–1028. Siebert ER, Parr LA (2003) A structural and contextual analysis of chimpanzee screams. Annals of

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

the New York Academy of Sciences 1000:104–109. Silber GK (1986) The relationship of social vocalizations to surface behavior and aggression in the Hawaiian humpback whale (Megaptera novaeangliae). Can. J. Zool. 64:2075–2080 Stimpert AK, Au WWL, Parks SE, Hurst T, Wiley DN (2011) Common humpback whale (Megaptera novaeangliae) sound types for passive acoustic monitoring. The Journal of the Acoustical Society of America 129:476–482. Stimpert AK, Lammers MO, Pack AA, Au WWL (2020) Variations in received levels on a sound and movement tag on a singing humpback whale: Implications for caller identification. The Journal of the Acoustical Society of America 147:3684–3690. Szipl G, Ringler E, Spreafico M, Bugnyar T (2017) Calls during agonistic interactions vary with arousal and raise audience attention in ravens. Frontiers in Zoology 14:1–13. Therneau TM, Atkinson EJ (2018) rpart: Recursive partioning and regression trees. R Foundation for Statistical Computing, Vienna, Austria. Thompson PO, Cummings WC, Ha SJ (1986) Sounds, source levels, and associated behavior of humpback whales, Southeast Alaska. The Journal of the Acoustical Society of America 80:735–740. Todt D (1986) Hinweis-Charakter und Mittler-Funktion von Verhalten. Zeitschriftfür Semiot 8:183–232. Tyack PL (1981) Interactions between singing Hawaiian humpback whales and conspecifics nearby. Behavioral Ecology 8:105–116. Tyack PL, Whitehead H (1983) Male competition in large groups of wintering humpback whales. Behaviour 83:132–154. Videsen SKA, Bejder L, Johnson M, Madsen PT (2017) High suckling rates and acoustic crypsis of humpback whale neonates maximise potential for mother–calf energy transfer. Functional Ecology 31:1561–1573. Wagner WE (1989) Graded aggressive signals in Blanchard’s cricket frog: Vocal responses to opponent proximity and size. Animal Behaviour 38:1025–1038. Williamson MJ, Kavanagh AS, Noad MJ, Kniest E, Dunlop RA (2016) The effect of close approaches for tagging activities by small research vessels on the behavior of humpback whales (Megaptera novaeangliae). Marine Mammal Science 32:1234–1253. Winn HE, Winn LK (1978) The song of the humpback whale in the West Indies. Marine Biology 47:97–114. Zahavi A (1982) The pattern of vocal signals and the information they convey. Behaviour 80:1–8. Zoidis AM, Smultea MA, Frankel AS, Hopkins JL, Day A, McFarland AS, Whitt AD, Fertl D

Chapter 2: Humpback whale social call production reflects both motivational state and arousal

(2008) Vocalizations produced by humpback whale (Megaptera novaeangliae) calves recorded in Hawaii. The Journal of the Acoustical Society of America 123:1737–1746.

Chapter 3

Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

3.1 Abstract

Social animals typically need to communicate a large amount of information to conspecifics, such as identity, group behaviour, and arousal. This information is usually conveyed within a complex repertoire through the use of discrete, stereotyped calls, and more graded, variable calls. In order to make comparisons amongst species, typical repertoire analyses concentrate primarily on determining the number of call types. These analyses assume that call types are discrete, and therefore do not adequately account for call gradation. However, the amount of gradation in calls is likely equally biologically relevant for highly social species. Here, we applied a combination of methods to classify the repertoire of a species with a complex communication system, the humpback whale (Megaptera novaeangliae). First, qualitative classification techniques were used, followed by typical Classification and Regression Trees and Random Forest algorithms. Fuzzy k- means clustering was then implemented, a clustering technique not widely implanted in classifying marine mammal repertoires. From this combination of techniques, we identified 15 call types in the repertoire, six of which were discrete and nine that were graded. This method provides more biologically-relevant information than just the identification of call types, and also may be more appropriate for graded repertoires like the humpback whale which are not well suited for discrete categorisation.

Keywords: vocalisation; classification; fuzzy clustering; communication; graded repertoire; humpback whale

3.2 Introduction

The vocal repertoires of non-human animals are presumably constrained to a finite number of call types (Hammerschmidt and Fischer 2008; Jansen et al. 2012; Manser 2013). As a result, animals have developed additional ways to convey potential information in calls, such as only using certain call types in specific situations to create context specificity (Premack 1975; Marler et al. 1992; Macedonia and Evans 1993; Evans 1997; Seyfarth and Cheney 2003; Wheeler and Fischer 2012). Alternatively, variation in the acoustic features of calls can convey information related to the

63 Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering signaller. This includes fixed information, which remains relatively stable over time (e.g. sex, individual identity, or age class) (Green and Marler 1979; Titze 1994; Taylor and Reby 2010; Briefer and McElligott 2011), as well as flexible information, which varies with the internal state of the signaller (e.g. motivation or arousal) (Marler 1977; Briefer 2012). Acoustic signals that contain fixed information tend to be stereotyped (‘discrete’) both between and within contexts, and are often perceived as distinct categories (Marler 1961, 1977; Green and Marler 1979). Discrete calls are unambiguous and can function in the absence of other signal modalities (i.e. visual or olfactory) (Marler 1976; Green and Marler 1979). They are therefore thought to have evolved in species which live in habitats with restricted visual access or with a high level of background noise (Marler 1975). Recent research indicates that other selection pressures also likely play a large role in discrete signal evolution, including the need for species recognition (Marler 1961; Ey and Fischer 2009; Fischer et al. 2017). As a result, discrete calls are often used for long-distance, species-specific territorial signals [e.g. blue monkey, Cercopithecus mitis, and red-tailed monkey, Cercopithecus ascanius (Marler 1975); black and white colobus, Colobus quereza (Marler 1975); gibbons, Nomascus spp. (Thinh et al. 2011)], or as predator- specific alarm calls [e.g. Campbell’s monkeys, Cercopithecus campbelli (Ouattara et al. 2009; Keenan et al. 2013); vervet monkeys, Cercopithecus aethiops (Seyfarth et al. 1980; Price et al. 2015); dwarf mongooses, Helogale parvula (Beynon and Rasa 1989; Manser et al. 2014)]. Calls that contain primarily flexible information are more variable in acoustic structure (‘graded’) and are often characterised by continuous acoustic variation between signals (Marler 1961, 1975, 1976; Marler et al. 1992; Hauser 1996). The acoustic structure of these calls typically fall along a continuum, rather than separate into distinct categories. Graded calls can therefore be more ambiguous and are thought to have evolved primarily for close-range signalling (Marler 1976; Green and Marler 1979). They are particularly useful for group-based social situations, where the graded structure of calls provides information about the internal motivational state or arousal of the signaller (Marler 1961, 1976; Morton 1977; Owings and Morton 1998; Manser 2010; Briefer 2012). For example, the low-frequency ‘rumbles’ of African elephants (Loxodonta africana) are emitted at higher and more variable fundamental frequencies when the signaller is in a negative, high intensity situation (Wood et al. 2005; Soltis et al. 2009). Horse (Equus caballus) ‘whinnies’ are longer in duration and higher in fundamental frequency when the signaller is in a negative, high arousal state (Briefer et al. 2015). Lastly, the ‘barks’ of domestic dogs (Canis familiaris) are more graded, shorter in duration, and higher in frequency in positive ‘play’ contexts (Yin and McCowan 2004).

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Most research into animal vocal communication is centred around establishing call catalogues (between-call variability). However, current methods of classification depend heavily on techniques which are unsuitable for graded data (e.g. discriminant function analyses and ‘hard’ clustering; see Rekdahl et al. 2013; Marler 1975; and Wadewitz et al. 2015 for detailed arguments). These types of analyses assume that calls are discrete and can be neatly categorised. The impracticality of this assumption is highlighted by the variation in the number of call types depending on the choice of analysis. For example, the repertoire of chacma baboons (Papio ursinus) is considered relatively discrete. However the use of three types of analyses did not reveal an obvious solution as to the number of call types (Wadewitz et al. 2015). In one captive population of African elephants, ‘rumbles’ were classified as multiple sub-types in several analyses, however a separate analysis failed to distinguish sub-types of this call (Leong et al. 2003; Soltis et al. 2005; Wood et al. 2005). Lastly, in humpback whales (Megaptera novaeangliae) off the coast of east Australia, research has indicated between 13 and 34 call types (Dunlop et al. 2007) as well as between 12 and 46 call types (Rekdahl et al. 2013). Although these studies used different data sets from different years, and some call types are only present in the repertoire for a few years, it is reasonable to assume at least some of the variability could be due to differences in the classification techniques used. The expansive data available on the humpback whale off of east Australia presents an opportunity to reassess their calls using new techniques that account for the variability evident in the repertoire. Research suggests humpback whales use a mix of discrete and graded call types (Silber 1986; Dunlop et al. 2008; Stimpert et al. 2011; Fournet et al. 2015). However, as with many other species, previous repertoire analyses for the humpback whale have concentrated primarily on establishing the number of call types and/or the stability of these call types over time (Dunlop et al. 2007; Rekdahl et al. 2013, 2017; Fournet et al. 2018a). While Fournet et al. (2015) provided initial insight into the gradation present in the humpback whale repertoire, their methods were dependent on agreement between classification techniques and not a dedicated analysis of discreteness. To date, no classification analyses have attempted to account for the presence of both discrete and graded call types. This study presents a classification of the call repertoire of humpback whales that includes a measure of within-call variability that can be used to separate discrete stereotyped calls from graded calls. Similar to Fournet et al. (2015), calls were first analysed using a combination of qualitative and quantitative techniques to investigate the degree of variability within each call type. Next, a classification technique that may be more suitable for such complex data known as fuzzy k-means

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering clustering was used to quantify the amount of gradation within each call type. These results were then used to inform the size of the call repertoire in terms of the number of different call types, as well as which could be considered discrete or graded. Results here will provide the foundation necessary for more comprehensive research into the function of graded calls in humpback whales and the potential flexible information they may convey. Further, these results can be used to make comparisons of the humpback whale vocal repertoire with other species and populations.

3.3 Methods

Data were collected on humpback whales off the coast of eastern Australia during their southward migration from their breeding grounds in the Great Barrier Reef to their Antarctic feeding grounds (Chittleborough 1959, 1965). While humpback whales do not typically form long- term associations, individuals do frequently interact during this time (Corkeron et al. 1994; Brown and Corkeron 1995; Kavanagh et al. 2017). To capture their vocal behaviour during these social interactions, non-invasive, suction-cup acoustic recording tags were attached to adult females with calves in 2010, 2011, 2014, and 2017. These tags are equipped with a sensor suite that records sound, pressure (depth), temperature, and acceleration and magnetic field along 3 axes to provide pitch, roll, and heading of the tagged animal. In 2010, 2011, and 2014, DTAGs (Johnson and Tyack 2003) were deployed and programmed to sample 16-bit audio at either 48 kHz (2011) or 96 kHz (2010, 2014) with a 400 Hz Butterworth high-pass filter to reduce low frequency flow noise. Sensitivity was approximately -170 dB re V/µPa from 1 – 2 kHz, but decreased above 2 kHz. In 2017, Acousondes (Greeneridge Sciences, Inc., http://www.acousonde.com) were deployed and programmed to sample 16-bit audio at 25 kHz. Sounds extracted from the DTAGs were corrected for the high-pass filter, but no correction was needed for the Acousonde (see Chapter 2 for details). Spectrograms of recordings were browsed visually and aurally in Raven Pro 1.5 (Bioacoustics Research Program 2017) by an experienced analyst using a Hann window, Fast Fourier Transform size of 2048 or 4096 samples (depending on the sampling rate), and 50% overlap. All humpback whale calls were marked and extracted for further analysis. Immediately before or after the call, a clip of background noise of at least 0.5 seconds in duration was also extracted. The noise clip covered the same bandwidth as the associated call clip. The signal-to-noise ratio (SNR) of each sound was calculated using the clip of surrounding noise with custom script (Girola et al. 2019) in Matlab (MathWorks 2018). In order to exclude calls outside of the group containing the focal animal, a SNR cut-off was used. The point at which the frequency measurements of a simulated call (a 0.5 second signal with a shape between 50 Hz and 1 kHz) were

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering no longer reliable was -3 dB SNR (Girola et al. 2019, see Chapter 2 for details). However, only sounds with a SNR greater than 5 dB were used to ensure the highest quality calls were included, while still maintaining a relatively high sample size to compare with previous humpback whale repertoire analyses (Dunlop et al. 2007; Stimpert et al. 2011; Fournet et al. 2015). Although recent research has indicated an SNR higher than 20 dB may be necessary to confidently assign calls to the focal animal (Stimpert et al. 2020), identifying the focal animal was not necessary for this analysis. Calls that were overlapping with other vocalisations (i.e. song or other calls) or acute noise (i.e. tag rubbing) were also removed from analyses. Background noise was removed from the call clips by subtracting the spectrum of the noise clips from the corresponding calls in Matlab (Girola et al. 2019). The custom script then automatically measured 11 common acoustic features from the de-noised call clips. Additionally, aggregate entropy (i.e. the disorder in a sound) was measured in Raven from the original (not denoised) sound clips as it was not included in the Matlab script (Girola et al. 2019). This resulted in a total of 12 features for analyses (Table 3.1). Measurements were chosen based on previous studies of humpback whale social call classification (Dunlop et al. 2007; Stimpert et al. 2011; Fournet et al. 2015). Numeric variables were scaled to account for differing units of measure. Frequency measurements were converted to a logarithmic scale to account for the mammalian perception of pitch (Richardson et al. 1995; Cardoso 2013).

Table 3.1 Description of the measurements used for repertoire classification. Frequency measurements were logged prior to analysis. Measurement Abbreviation Description Disorder in the sound based on the energy Aggregate entropy (unit-less) 1 AgEnt distribution summed over all frequencies Duration (s) Dur Total duration of the sound Time at which the minimum power occurs in Minimum time (s) T MIN the signal Time at which the maximum power occurs in Maximum time (s) T MAX the signal Lower limit of the frequency band of the Minimum frequency (Hz) 1 F MIN signal Upper limit of the frequency band of the Maximum frequency (Hz) 1 F MAX signal Frequency at which the maximum amplitude Peak frequency (Hz) F P level in the sound occurs

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Energy midpoint that divides the sound into Center frequency (Hz) F C two equal frequency intervals Difference between the 3rd and 1st quartile Inter-quartile bandwidth (Hz) F IQ frequencies Peak frequency of 10% of the sound centered Peak frequency of T (Hz) F Q1 Q1T at the 1st quartile time Peak frequency of 10% of the sound centered Peak frequency of T (Hz) F Q3 Q3T at the 3rd quartile time Ratio of 1st quartile frequency to 3rd quartile Frequency trend (ratio) F TREND frequency 1 based on measurements of the Raven spectrogram

3.3.1 Aural-visual (AV) qualitative classification Initial classification of call types was carried out via aural and visual (AV) inspection of sounds in Raven by an experienced analyst (DC; primary author). Another experienced analyst then performed a separate AV classification while blind to the results from the first reviewer. Both analysts qualitatively assigned calls based on previous call catalogues for east Australian humpback whales (Dunlop et al. 2007, 2008; Rekdahl et al. 2013). Any call type classification assignment that did not match between the two reviewers was reassessed until there was agreement for all calls.

3.3.2 Classification and Regression Rrees (CART) and Random Forests (RF) Two non-parametric recursive partitioning techniques were implemented: Classification and Regression Trees (CART) and Random Forests (RF) (Breiman et al. 1984; Breiman 2001). These two techniques are commonly used tools for classifying animal vocal repertoires in both marine (Van Opzeeland and Van Parijs 2004; Rekdahl et al. 2013; Garland et al. 2015; Webster et al. 2016; Fournet et al. 2018b; Sharpe et al. 2019) and terrestrial species (Melendez et al. 2006; Armitage and Ober 2010; McGrath et al. 2017). Classification trees explain the variation in a response variable (e.g. call type) using multiple explanatory variables (e.g. acoustic features). They are equipped to correlated variables and pseudo-replication, which allows for the inclusion of calls that may be from the same individual (Breiman et al. 1984; De’ath and Fabricius 2000). Trees were built by first finding the variable that best split the call types into two groups or nodes. Further splits were made recursively until all calls were used, the node contained a minimum number of calls, or no improvements could be made with a further split (De’ath and Fabricius 2000). The minimum number of calls for a terminal node was set to 20 as this was the minimum sample size for all AV classified call types. All explanatory variables were initially considered for

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering each split. Variables were then ranked according to the Gini index, a measure of impurity, and the split that minimised this impurity was used. The tree was first overgrown and then V-fold cross validation with 50 subsets was performed. In V-fold cross validation, the data are divided randomly into v equal parts. The CART then runs on v-1 parts and tested against the remaining part, providing an estimate of the error rate (risk) and the standard error of the risk. The tree was then pruned upward until the tree with the smallest estimated cross-validated error within one standard error of the best tree was reached (1-SE rule) (Breiman et al. 1984; Hastie et al. 2008). The number of terminal nodes in which a call type was placed was considered a metric of variability; the call was preliminarily designated discrete if it was assigned to only one node, or graded if assigned to more than one node. The CART analysis was run in R Studio (R Core Team 2018) using the package rpart (Therneau and Atkinson 2018). Next, a RF analysis was used as an extension of the CART to grow an ensemble of trees (a forest). This method provides an estimate of the classification uncertainty of each tree and the relative importance of each predictor variable (i.e. acoustic measurement; Table 3.1) (Breiman 2001; Garland et al. 2015). The classification uncertainty is termed the out-of-bag (OOB) error. The splitting of nodes is based on a specified number of randomly selected predictors rather than the full set of variables (Breiman 2001). Simulations were run to calculate the stability of the classification uncertainty, with a high level of stability correlating to a low OOB error. Based on the lowest OOB error, six variables were sampled at each split and 1000 trees were grown (following Garland et al. 2015). The OOB errors for each call type were considered another preliminary measure of how discrete or graded call types were. The RF analysis was run in R using the package randomForest (Liaw and Wiener 2002).

3.3.3 ‘Soft’ clustering with fuzzy k-medoids (FKM) A common approach to the classification of animal acoustic repertoires is with ‘hard’ clustering techniques (e.g. k-means, hierarchical clustering) (Kaufman and Rousseeuw 1990). However, the most obvious limitation to hard clustering is the assumption that objects belong to distinct classes, i.e. each data-point can only occur in one cluster (Jain 2010). For graded communication systems, this is usually not the case as calls can have acoustic features that are typical of more than one cluster (Wadewitz et al. 2015). ‘Soft’ classification methods, based on ‘fuzzy set theory’ (Zadeh 1965), are an alternative approach that allow for the notion of imperfect membership. They are therefore better suited for classifying graded data (Ruspini 1969; Dunn 1973; Bezdek 1974, 1981). One such soft classification method is fuzzy k-means (FKM) (Risch et al.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

2013; Ferraro and Giordani 2015; Wadewitz et al. 2015; Fischer et al. 2017). As with all clustering techniques, data-points (here, individual calls) are partitioned into clusters (here, assumed to be call types) based on a set of features. In contrast to hard clustering, the FKM assigns each data point a membership value to each of the clusters ranging from 0 (does not match the cluster properties) to 1 (fully matches the cluster properties) with allowance for intermediate membership between clusters (Bezdek 1981). Analyses were run in R using the packages fclust (Ferraro and Giordani 2015) and DoTC (Wadewitz et al. 2016). The focus of this portion of the analysis was to classify call types not well separated using traditional techniques. Therefore, the five call types considered highly discrete based on the above analyses were first removed before running the FKM. To determine the optimal number of clusters, multiple FKM iterations with varying degrees of the fuzziness parameter (μ, 1.5-2.5) were run to establish the most stable solution (Wadewitz et al. 2015; Fischer et al. 2017). The maximum number of clusters was set above the expected number of clusters based on AV analyses even after the removal of the most discrete call types. A typicality coefficient (TC) was calculated for each call by subtracting the second largest cluster membership value from the largest membership value (Wadewitz et al. 2015; Fischer et al. 2017). The halved mean absolute deviation of all TCs was then used to establish a threshold whereby a call above this value was considered ‘typical’ (distinctive), and below this value ‘atypical’ (indistinctive). Clusters with any calls that fell above this threshold were classified as discrete, and otherwise as graded.

3.4 Results

From over 96 hours of acoustic recording data, 4,695 calls were detected. After removing calls of poor quality and low SNR (< 5 dB), a total of 2,376 calls were retained for further analyses.

3.4.1 Aural-visual (AV) qualitative classification AV classification resulted in 16 call types: ‘bop’ (n = 565), ‘crow’ (n = 20), ‘eeaw’ (n = 103), ‘grumble’ (n = 104), ‘grunt’ (n = 70), ‘meow’ (n = 165), ‘moan’ (n = 55), ‘paired croaks’ (n = 43), ‘knock’ (n = 149), ‘snort’ (n = 416), ‘spiccato’ (n = 40), ‘squawk’ (n = 101), ‘squeak’ (n = 184), ‘thwop’ (n = 23), ‘wop’ (n = 44), and ‘wup’ (n = 294). See Figures 3.1 and 3.2 for representative spectrograms.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Figure 3.2 Spectrograms of the four high frequency call types identified by aural-visual examination: (a) ‘eeaw’, (b) ‘meow’, (c) ‘squawk’, and (d) ‘squeak’.

3.4.2 Classification and regression trees (CART) and random forests (RF) The CART analysis had a correct classification rate of 69%. The variables ranked highest in importance (contributed to the most splits) were aggregate entropy (AgEnt) and duration (Dur), although the initial split was by duration. The number of splits was based on the lowest cross- validated error. This formed four main branches with 35 terminal nodes, indicating the CART detected 35 different call types in the data set. This was significantly higher than the initial qualitative AV call type classification, however it was similar to the results of Dunlop et al. (2007) and Rekdahl et al. (2013). Eight AV call types were present in more than one terminal node: ‘snorts’ in six terminal nodes; ‘bops’, ‘wups’, and ‘squeaks’ in four terminal nodes each; ‘eeaws’, ‘grumbles’, and ‘squawks’ in three terminal nodes each; and ‘paired croaks’ in two terminal nodes (Table 3.2). This distribution of call types amongst numerous terminal nodes could indicate these call types are either relatively graded compared with other call types, or they need to be split further.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

The RF analysis had a classification success rate of 71% (OOB estimate of error = 29%). The acoustic variables most important in decreasing the Gini index were again aggregate entropy and duration. ‘Thwops’ were misclassified the most (83% of the time, Table 3.2), but were classified mostly as ‘wops’ (Appendix 2, Table A2.1). Similarly, ‘wops’ were misclassified at a relatively high rate (43% of the time), 26% of which resulted from incorrect classification as ‘thwops’. These two call types are both low-frequency, harmonic upsweeps that are structurally similar with the only major difference being that the ‘thwop’ is broken into two parts (Dunlop et al. 2007, Figure 3.1). Therefore, the high misclassification is likely due to minute differences and a low sample size rather than gradation. ‘Spiccatos’ were also misclassified at a high rate (48%, Table 3.2), although this is likely due to the high variability in the duration of this call type (Appendix 2, Table A2.1). Some call types were not only misclassified at relatively high rates (≥ 25%), they were also not classified in a consistent way, e.g. they were not misclassified as one particular call type in the RF. This included ‘grumbles’, ‘snorts’, ‘squawks’, and ‘wups’. Again, this indicates certain call types are graded, particularly in the case of these four call types as they fell into multiple terminal nodes in the CART (Table 3.2; Appendix 2, Table A2.1).

Table 3.2 Results of the Classification and Regression Tree (CART) and Random Forest (RF) analyses, indicating the total number of each call type, the number of terminal nodes resulting from the CART, and the misclassification error rate from the RF. AV: aural-visual. CART: RF: AV Call Type Terminal Misclassification Nodes Error Rate Bop (n = 565) 4 0.13 Crow (n = 20) 1 0.70 Eeaw (n = 103) 3 0.20 Grumble (n = 104) 3 0.41 Grunt (n = 70) 1 0.30 Meow (n = 165) 1 0.10 Moan (n = 55) 1 0.55 Paired croaks (n = 43) 2 0.23 Knock (n = 149) 1 0.53 Snort (n = 416) 6 0.28 Spiccato (n = 40) 0 0.48

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Squawk (n = 101) 3 0.58 Squeak (n = 184) 4 0.27 Thwop (n = 23) 0 0.83 Wop (n = 44) 1 0.43 Wup (n = 294) 4 0.34

3.4.3 ‘Soft’ clustering with fuzzy k-means (FKM) Based on the results of the AV, CART, and RF, the five call types considered relatively discrete were removed prior to running the FKM: ‘meows’ and ‘paired croaks’ because of their distinctiveness in the AV classification, low misclassification rates, and low number of terminal nodes; ‘spiccatos’ because of their distinctiveness; and ‘thwops’ and ‘wops’ because of their distinctiveness, misclassification primarily with each other, and their use and discrete nature in this and other populations of humpback whales (Dunlop et al. 2007, 2008; Stimpert et al. 2011; Rekdahl et al. 2013, 2017; Seger 2016; Fournet et al. 2018b, a). The FKM produced three stable clustering solutions: 10 (μ = 1.7), 8 (μ = 1.85), and 6 (μ = 1.95). Based on validity indices, the optimal number of clusters was determined to be 10. As with the CART analysis, many call types fell into multiple clusters, which is typical of most clustering techniques for classifying humpback whale and other animal vocalisations (Stimpert et al. 2011; Fournet et al. 2015). For example, ‘bops’, ‘snorts’, and ‘squeaks’ were split between several clusters each, while ‘crows’ and ‘moans’ did not correspond to any particular cluster (Appendix 2, Table A2.2). This could be due to the high level of gradation present in the data, and highlights the limitations of attempting to classify a highly graded repertoire with purely quantitative methods. Spectrograms of two calls from each cluster are in Figure 3.3, highlighting calls that were classified as a single call type in the AV analysis, but fell into more than one cluster in the FKM.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

‘snorts’ during the aural-visual analysis, and blue stars indicate calls classified as ‘bops’. The placement of these calls into separate clusters highlights the ability of the fuzzy cluster analysis to differentiate calls that sound similar to a human observer.

Typicality coefficients (TC) were calculated for individual calls in order to determine a threshold for which each call could be considered typical (e.g., representative of its cluster). In general, data-points (calls) with large average TCs (close to 1) are well separated from others and can be considered more discrete, with the opposite for low (close to 0) average TCs (Wadewitz et al. 2015; Fischer et al. 2017). Results indicated a call or cluster could be considered typical if the typicality coefficient (TC) was above 0.93, and atypical if it was below 0.07. Here, the average TCs for all individual calls as well as each cluster were significantly skewed towards 0, indicating a highly graded data set overall (Figure 3.4). No cluster had all calls above the 0.93 threshold, however some individual calls fell above this threshold. These calls were all confined to cluster 4, which was comprised primarily of snorts and knocks, and thus only this cluster was considered to be discrete (Figure 3.4, Table 3.3). Combined with the five call types deemed discrete from the previous analyses, this resulted in a total of 15 call types, six of which were relatively discrete and nine which were more graded. The acoustic parameters for the final 15 call types can be found in Appendix 2, Table A2.3.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Table 3.3 Results of the fuzzy cluster analysis indicating the number of calls in each cluster, the aural-visual (AV) call type that was represented the most, the average typicality coefficients (mean ± SD), the percent of calls that fell above and below the typicality threshold, and the designation of graded or discrete. Any cluster with calls above the threshold was considered relatively discrete. Note that clusters contain multiple AV call types.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

% Atypical % Typical Call Cluster Cluster Description TC Calls Calls Category 1 (n = 248) Low entropy snort/knock 0.15 ± 0.12 100% 0% Graded 2 (n = 387) Wup/low frequency eeaw 0.09 ± 0.08 100% 0% Graded 3 (n = 220) Grumble/long snort 0.06 ± 0.07 100% 0% Graded 4 (n = 152) Discrete snort 0.46 ± 0.31 96% 4% Discrete 5 (n = 128) Broadband bop 0.09 ± 0.10 100% 0% Graded 6 (n = 113) High frequency squeak 0.24 ± 0.20 100% 0% Graded 7 (n = 140) Squeak/high frequency eeaw 0.08 ± 0.09 100% 0% Graded 8 (n = 273) Low entropy bop 0.29 ± 0.24 100% 0% Graded 9 (n = 182) High frequency bop 0.23 ± 0.18 100% 0% Graded 10 (n = 218) High entropy bop 0.08 ± 0.08 100% 0% Graded

3.5 Discussion

Classification analyses for graded vocal repertoires are typically hindered by the amount of variability within calls. Here we combined traditional techniques with fuzzy clustering, a classification method not previously applied to a marine animal. Our results extend traditional methods by both determining the number of call types in the data (15) and whether call types were discrete or graded. This method provides information that may be more biologically relevant than just the number of call types in a repertoire. It may therefore be more appropriate for species which display gradation in their repertoire and are not well suited for typical classification methods. Using fuzzy clustering can also create an opportunity to investigate the communication system of humpback whales in a new way, and to compare between populations and species. Quantitative measurements of the gradation in a call, as presented here, are currently considered to be more adequate metrics than repertoire size for assessing the communicative complexity of a species (Peckre et al. 2019). According to this study, the humpback whale acoustic repertoire contains a combination of discrete and graded calls. Humpback whales do not conform to the classical definition of a complex social system (May-Collado et al. 2007), but they do demonstrate social behaviour that involves both coordination and cooperation (Whitehead 1983; Clapham 1993; Parks et al. 2014). During the breeding season, they also engage in competitive interactions that involve frequently changing group dynamics, unstable hierarchies, and potentially even temporary coalitions (Tyack and Whitehead 1983; Silber 1986; Clapham et al. 1992; Clapham 1996). Viewed in this aspect, the presence of so many graded calls is not surprising as they may need to convey complex information to conspecifics, including a combination of temporary dominance status, testosterone levels, arousal, body size, and/or motivational state. However, there

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering were still some discrete call types detected. This could be related to the selection pressures that may have influenced the evolution of their acoustic repertoire. Humpback whales live in an environment which can be heavily impacted by natural noise independent of anthropogenic influence such as abiotic (e.g. wind and breaking waves) and biotic sounds (e.g. humpback song and snapping shrimp) (Cato 1976; Cato and McCauley 2002). In these situations, signals can be easily masked or degraded (Erbe et al. 2016), and discrete sounds would potentially be favoured. This has been demonstrated previously in migrating east Australian humpbacks, which switch to using discrete surface-generated sounds during periods of increased wind noise (Dunlop et al. 2010, 2013; Dunlop 2016, 2018). Therefore, a variety of selection pressures (e.g. social and habitat) have likely contributed to the shaping of the humpback acoustic repertoire, as is suspected for most taxa (Marler 1975; Ey and Fischer 2009; Freeberg et al. 2012; Fischer et al. 2017; Peckre et al. 2019). While providing new insight into the gradation in calls within the humpback whale repertoire, this study highlights major issues regarding the classification of acoustic repertoires in species that use graded calls. Here, the results of the traditional analyses strongly indicated several call types should be split. The fuzzy clustering, however, did not separate these call types into distinct clusters. Similar to the classification of African elephant rumbles (see introduction), these results provide evidence that gradation within call types may preclude establishing discrete categories even if they do exist, regardless of the type of analysis. However, variation within a signal does not automatically exclude a call type from being discrete (van Hooff and Preuschoft 2003). For example, killer whales have a type of call class deemed ‘aberrant’ calls, which are variations of discrete call types (Ford 1989; Rehn et al. 2007). These calls are used significantly more often in high arousal socialising states than lower arousal states (e.g. foraging and travelling). The presence of aberrant calls has not been specifically assessed in humpback whales. However, aberrant versions of discrete calls are anecdotally evident, at least in the current data set, which included periods of high arousal with socialising animals. Quantitative techniques might not be able to differentiate an aberrant discrete call from a graded call. This could be one reason that the classification techniques used here were not as successful as in other studies of humpback whales, where behavioural contexts likely did not include high arousal social interactions and calls that could not be definitively classified by aural-visual analyses were omitted (Fournet et al. 2018a, b). The aural-visual analysis differentiated 16 call types, whereas the Classification and Regression Tree suggested 35. The combined classification trees/random forest/fuzzy clustering method determined 15. The inconsistent results suggest that relying on one method may not be suitable for classifying graded repertoires, or different classifications may be useful for different

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering purposes. In particular, the use of qualitative measures may be necessary when repertoires contain graded calls, particularly when call types include sequences, as in the case of ‘paired croaks’ in the present study. Fournet et al. (2015) found that some call features that were identified as important indicators of humpback whale call type through a visual analysis were not uniquely identified as discriminating features using quantitative classification techniques (discriminant function analysis and hierarchical agglomerative cluster analysis). Similar results were found in beluga whales (Delphinapterus leucas), also known to have graded call types, where a cluster analysis grouped together various whistle call types which a concurrent aural-visual analysis clearly differentiated (Chmelnitsky and Ferguson 2012). Using a combination of quantitative techniques and human observers strengthened the results of Fournet et al. (2015) and Chmelnitsky and Ferguson (2012). We suggest a similar approach but using the methods outlined in the present study, with some modifications to improve classification success. For example, additional acoustic parameters that may help in detecting differences could be incorporated, as it is likely that important acoustic features that would help in classification were missed. This includes pulse repetition rate and inflection points, which are important in other classification studies (Rekdahl et al. 2013, 2017; Garland et al. 2015). Another potential technique could be to qualitatively determine initial broad call classes before running more quantitative analyses on smaller subsets of data. Improving classification success will be important for future research into the function of call types, especially for obviously discrete call types that are not clustered as such. In conclusion, classifying a highly graded repertoire is challenging, and no one optimal solution likely exists (Wadewitz et al. 2015; Fischer et al. 2017). Some combination of analyses, such as those presented here, could provide the subjectivity needed for initial classification as well as the objectivity needed as validation. This method is generally accepted in most classification analyses at present, at least for cetaceans. However, the use of fuzzy clustering to quantify gradation is a promising addition to current techniques that has yet to be widely applied. While fuzzy clustering does not necessarily provide an unequivocal answer to the question of the number of call types, especially in a heavily graded repertoire, it does allow for fine-scale details of call types (albeit as clusters) to be captured which are missed using traditional clustering techniques (Wadewitz et al. 2015; Fischer et al. 2017). Additionally, it provides information on whether calls should be considered discrete or graded, which may be more biologically meaningful than the number of calls in an acoustic repertoire. Developing this analysis further, and incorporating it to current classification schemes, could open the door to future comparisons of call types and

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering communicative complexity between populations and species, as well as the function of animal vocalisations.

3.6 References

Armitage DW, Ober HK (2010) A comparison of supervised learning techniques in the classification of bat echolocation calls. Ecological Informatics 5:465–473. Beynon P, Rasa OAE (1989) Do dwarf mongooses have a language-Warning vocalizations transmit complex information. South African Journal of Science 85:447–450. Bezdek JC (1974) Cluster validity with fuzzy sets. Journal of Cybernetics 3:58–73. Bezdek JC (1981) Pattern Recognition with Fuzzy Objective Function Algorithms. Plenum Press, New York. Bioacoustics Research Program (2017) Raven Pro: Interactive Sound Analysis Software. Cornell Laboratory of Ornithology, Ithaca (NY). Breiman L (2001) Random forests. Machine Learning 45:5–32. Breiman L, Friedman JH, Olshen RA, Stone CJ (1984) Classification and Regression Trees. Chapman and Hall, London. Briefer E, McElligott AG (2011) Indicators of age, body size and sex in goat kid calls revealed using the source-filter theory. Applied Animal Behaviour Science 133:175–185. Briefer EF (2012) Vocal expression of emotions in mammals: Mechanisms of production and evidence. Journal of Zoology 288:1–20. Briefer EF, Maigrot AL, Mandel R, Freymond SB, Bachmann I, Hillmann E (2015) Segregation of information about emotional arousal and valence in horse whinnies. Scientific Reports 4:1–11. Brown M, Corkeron P (1995) Pod characteristics of migrating humpback whales (Megaptera novaeangliae) off the East Australian coast. Behaviour 132:163–179. Cardoso GC (2013) Using frequency ratios to study vocal communication. Animal Behaviour 85:1529–1532. Cato DH (1976) Ambient sea noise in waters near Australia. The Journal of the Acoustical Society of America 60:320–328. Cato DH, McCauley RD (2002) Australian research in ambient sea noise. Acoustics Australia 30:13–20. Chittleborough RG (1959) Australian marking of humpback whales. Novsk Hvalfangsttid 48:47– 55. Chittleborough RG (1965) Dynamics of two populations of the humpback whale, Megaptera

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

novaeangliae (Borowski). Australian Journal of Marine and Freshwater Research 16:33–128. Chmelnitsky EG, Ferguson SH (2012) Beluga whale, Delphinapterus leucas, vocalizations from the Churchill River, Manitoba, Canada. The Journal of the Acoustical Society of America 131:4821–4835. Clapham PJ (1993) Social organization of humpback whales on a North Atlantic feeding ground. Zoological Symposium 66:131–145. Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mammal Review 26:27–49. Clapham PJ, Palsboll PJ, Mattila DK, Vasquez O (1992) Composition and dynamics of humpback whale competitive groups in the West Indies. Behaviour 122:182–194. Corkeron PJ, Brown M, Slade RW, Bryden MM (1994) Humpback whales, megaptera novaeangliae (Cetacea: Balaenopteridae), in Hervey Bay, Queensland. Wildlife Research 21:293–305. De’ath G, Fabricius KE (2000) Classification and Regression Trees: A powerful yet simple technique for ecological data analysis. Ecology 81:3178–3192. Dunlop RA (2016) The effect of vessel noise on humpback whale, Megaptera novaeangliae, communication behaviour. Animal Behaviour 111:13–21. Dunlop RA (2018) The communication space of humpback whale social sounds in wind-dominated noise. The Journal of the Acoustical Society of America 144:540–551. Dunlop RA, Cato DH, Noad MJ (2008) Non-song acoustic communication in migrating humpback whales (Megaptera novaeangliae). Marine Mammal Science 24:613–629. Dunlop RA, Cato DH, Noad MJ (2010) Your attention please: increasing ambient noise levels elicits a change in communication behaviour in humpback whales (Megaptera novaeangliae). Proceedings of the Royal Society B: Biological Sciences 277:2521–2529. Dunlop RA, Noad MJ, Cato DH (2013) Modification of humpback whale social sound repertoire and vocal source levels with increased noise. Proceedings of Meetings on Acoustics 19:1–6. Dunlop RA, Noad MJ, Cato DH, Stokes DM (2007) The social vocalization repertoire of east Australian migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 122:2893–2905. Dunn JC (1973) A fuzzy relative of the ISODATA process and its use in detecting compact well- separated clusters. Journal of Cybernetics 3:32–57. Erbe C, Reichmuth C, Cunningham K, Lucke K, Dooling R (2016) Communication masking in marine mammals: A review and research strategy. Marine Pollution Bulletin 103:15–38.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Evans CS (1997) Referential Signals. Perspectives in Ethology 12:99–143. Ey E, Fischer J (2009) The “acoustic adaptation hypothesis”—A review of the evidence from birds, anurans and mammals. Bioacoustics 19:21–48. Ferraro MB, Giordani P (2015) A toolbox for fuzzy clustering using the R programming language. Fuzzy Sets and Systems 279:1–16. Fischer J, Wadewitz P, Hammerschmidt K (2017) Structural variability and communicative complexity in acoustic communication. Animal Behaviour 134:229–237. Ford JKB (1989) Acoustic behaviour of resident killer whales (Orcinus orca) off Vancouver Island, British Columbia. Canadian Journal of Zoology 67:727–745. Fournet ME, Szabo A, Mellinger DK (2015) Repertoire and classification of non-song calls in Southeast Alaskan humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 137:1–10. Fournet MEH, Gabriele CM, Culp DC, Sharpe F, Mellinger DK, Klinck H (2018a) Some things never change: multi-decadal stability in humpback whale calling repertoire on Southeast Alaskan foraging grounds. Scientific Reports 8:1–13. Fournet MEH, Jacobsen L, Gabriele CM, Mellinger DK, Klinck H (2018b) More of the same: allopatric humpback whale populations share acoustic repertoire. PeerJ 6:e5365. Freeberg TM, Dunbar RIM, Ord TJ (2012) Social complexity as a proximate and ultimate factor in communicative complexity. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1785–1801. Garland EC, Castellote M, Berchok CL (2015) Beluga whale (Delphinapterus leucas) vocalizations and call classification from the eastern Beaufort Sea population. The Journal of the Acoustical Society of America 137:3054–3067. Girola E, Noad MJ, Dunlop RA, Cato DH (2019) Source levels of humpback whales decrease with frequency suggesting an air-filled resonator is used in sound production. The Journal of the Acoustical Society of America 145:869–880. Green S, Marler P (1979) The Analysis of Animal Communication. In: Social Behavior and Communication. Plenum Press, New York (NY), pp 73–158. Hammerschmidt K, Fischer J (2008) Constraints in primate vocal production. In: Oller DK, Griebel U (eds) Evolution of Communicative Flexibility: Complexity, Creativity, and Adaptability in Human and Animal Communication. MIT Press, Cambridge, pp 93–119. Hastie T, Tibshirani R, Friedman J (2008) The elements of statistical learning: Data mining, inference, and prediction, Second. Springer.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Hauser MD (1996) The evolution of communication. MIT Press, Cambridge, MA. Jansen DAWAM, Cant MA, Manser MB (2012) Segmental concatenation of individual signatures and context cues in banded mongoose (Mungos mungo) close calls. BMC Biology 10:1–10. Johnson MP, Tyack PL (2003) A digital acoustic recording tag for measuring the response of wild marine mammals to sound. IEEE Journal of Oceanic Engineering 28:3–12. Kaufman L, Rousseeuw PJ (1990) Finding groups in data: An introduction to cluster analysis. Wiley, New York. Kavanagh AS, Noad MJ, Blomberg SP, Goldizen AW, Kniest E, Cato DH, Dunlop RA (2017) Factors driving the variability in diving and movement behavior of migrating humpback whales (Megaptera novaeangliae): Implications for anthropogenic disturbance studies. Marine Mammal Science 33:413–439. Keenan S, Lemasson A, Zuberbühler K (2013) Graded or discrete? A quantitative analysis of Campbell’s monkey alarm calls. Animal Behaviour 85:109–118. Leong KM, Ortolani A, Burks KD, Mellen JD, Savage A (2003) Quantifying acoustic and temporal characteristics of vocalizations for a group of captive african elephants Loxodonta africana. Bioacoustics 13:213–231. Liaw A, Wiener M (2002) Classification and Regression by randomForest. R News 2:18–22. Macedonia JM, Evans CS (1993) Variation among mammalian alarm call systems and the problem of meaning in animal signals. Ethology 93:177–197. Manser MB (2013) Semantic communication in vervet monkeys and other animals. Animal Behaviour 86:491–496. Manser MB (2010) The generation of functionally referential and motivational vocal signals in mammals. In: Brudzynski SM (ed) Handbook of Mammalian Vocalization - an integrative neuroscience approach. Academic Press, London (UK), pp 477–486. Manser MB, Jansen DAWAM, Graw B, Hollén LI, Bousquet CAH, Furrer RD, le Roux A (2014) Vocal complexity in meerkats and other mongoose species. Advances in the Study of Behavior 46:281–310. Marler P (1977) The structure of animal communication sounds. In: Bullock T, Evans E (eds) Recognition of complex acoustic signals. Dahlem Konferenzen, Berlin, pp 17–35. Marler P (1961) The logical analysis of animal communication. Journal of Theoretical Biology 1:295–317. Marler P (1976) Social organization, communication, and graded signals: The chimpanzee and the gorilla. In: Bateson PPG, Hinde RA (eds) Growing Points in Ethology. Cambridge University

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Press, Oxford (UK), pp 239–277. Marler P (1975) On the origin of speech from animal sounds. In: Kavanagh JF, Cutting J (eds) The Role of Speech in Language. MIT Press, Cambridge (MA), pp 11–37. Marler P, Evans CS, Hauser MD (1992) Animal signals: Motivational, referential, or both? In: Papousek H, Jurgens U, Papousek M (eds) Nonverbal vocal communication: comparative and developmental approaches. Cambridge University Press, Cambridge, UK, pp 66–86. MathWorks (2018) Matlab. The MathWorks, Inc., Natick, MA. May-Collado LJ, Agnarsson I, Wartzok D (2007) Phylogenetic review of tonal sound production in whales in relation to sociality. BMC Evolutionary Biology 7:1–20. McGrath N, Dunlop R, Dwyer C, Burman O, Phillips CJC (2017) Hens vary their vocal repertoire and structure when anticipating different types of reward. Animal Behaviour 130:79–96. Melendez K V., Jones DL, Feng AS (2006) Classification of communication signals of the little brown bat. The Journal of the Acoustical Society of America 120:1095–1102. Morton ES (1977) On the occurrence and significance of motivation-structural rules in some bird and mammal sounds. The American Naturalist 111:855–869. Ouattara K, Lemasson A, Zuberbuhler K (2009) Campbell’s monkeys use affixation to alter call meaning. PloS one 4:1–7. Owings D, Morton ES (1998) Animal Vocal Communication: A New Approach. Cambridge University Press, Cambridge (UK). Parks SE, Cusano DA, Stimpert AK, Weinrich MT, Friedlaender AS, Wiley DN (2014) Evidence for acoustic communication among bottom foraging humpback whales. Scientific Reports 4:7508. Peckre L, Kappeler PM, Fichtel C (2019) Clarifying and expanding the social complexity hypothesis for communicative complexity. Behavioral Ecology and Sociobiology 73:1–19. Premack D (1975) On the origin of language. In: Gazzaniga MS, Blakemore CB (eds) Handbook of Psychobiology. Academic Press, New York, pp 591–605. Price T, Wadewitz P, Cheney D, Seyfarth R, Hammerschmidt K, Fischer J (2015) Vervets revisited: A quantitative analysis of alarm call structure and context specificity. Scientific Reports 5:1– 11. R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Rehn N, Teichert S, Thomsen F (2007) Structural and temporal emission patterns of variable pulsed calls in free-ranging killer whales (Orcinus orca). Behaviour 144:307–329.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Rekdahl ML, Dunlop RA, Noad MJ, Goldizen AW (2013) Temporal stability and change in the social call repertoire of migrating humpback whales. The Journal of the Acoustical Society of America 133:1785–1795. Rekdahl ML, Tisch C, Cerchio S, Rosenbaum H (2017) Common nonsong social calls of humpback whales (Megaptera novaeangliae) recorded off northern Angola, southern Africa. Marine Mammal Science 33:365–375. Richardson WJ, Greene, Jr. CR, Malme CI, Thomson DH (1995) Marine Mammals and Noise. Academic Press, San Diego (CA). Risch D, Clark CW, Dugan PJ, Popescu M, Siebert U, Van Parijs SM (2013) Minke whale acoustic behavior and multi-year seasonal and diel vocalization patterns in Massachusetts Bay, USA. Marine Ecology Progress Series 489:279–295. Ruspini EH (1969) A new approach to clustering. Information and Control 15:22–32. Seger KD (2016) Ambient acoustic environments and cetacean signals, baseline studies from humpback whale and gray whale breeding grounds. University of California, San Diego. Seyfarth RM, Cheney DL (2003) Meaning and emotion in animal vocalizations. Annals New York Academy of Sciences 1000:32–55. Seyfarth RM, Cheney DL, Marler P (1980) Vervet monkey alarm calls: Semantic communication in a free-ranging primate. Animal Behaviour 28:1070–1094. Sharpe DL, Castellote M, Wade PR, Cornick LA (2019) Call types of Bigg’s killer whales (Orcinus orca) in western Alaska: Using vocal dialects to assess population structure. Bioacoustics 28:74–99. Silber GK (1986) The relationship of social vocalizations to surface behavior and aggression in the Hawaiian humpback whale (Megaptera novaeangliae). Can. J. Zool. 64:2075–2080 Soltis J, Leighty KA, Wesolek CM, Savage A (2009) The expression of affect in African elephant (Loxodonta africana) rumble vocalizations. Journal of Comparative Psychology 123:222–225. Soltis J, Leong K, Savage A (2005) African elephant vocal communication II: Rumble variation reflects the individual identity and emotional state of callers. Animal Behaviour 70:589–599. Stimpert AK, Au WWL, Parks SE, Hurst T, Wiley DN (2011) Common humpback whale (Megaptera novaeangliae) sound types for passive acoustic monitoring. The Journal of the Acoustical Society of America 129:476–482. Stimpert AK, Lammers MO, Pack AA, Au WWL (2020) Variations in received levels on a sound and movement tag on a singing humpback whale: Implications for caller identification. The Journal of the Acoustical Society of America 147:3684–3690.

Chapter 3: Quantifying the number of call types in a complex vocal repertoire using fuzzy clustering

Taylor AM, Reby D (2010) The contribution of source-filter theory to mammal vocal communication research. Journal of Zoology 280:221–236. Therneau TM, Atkinson EJ (2018) rpart: Recursive partioning and regression trees. R Foundation for Statistical Computing, Vienna, Austria. Thinh VN, Hallam C, Roos C, Hammerschmidt K (2011) Concordance between vocal and genetic diversity in crested gibbons. BMC Evolutionary Biology 11:1–9. Titze IR (1994) Principles of Voice Production. Prentice Hall, Englewood Cliffs (NJ). Tyack PL, Whitehead H (1983) Male competition in large groups of wintering humpback whales. Behaviour 83:132–154. van Hooff J, Preuschoft S (2003) Laughter and smiling: the intertwining of nature and culture. In: de Waal F, Tyack P (eds) Animal social complexity: intelligence, culture and individualized societies. Harvard University Press, Cambridge, MA, pp 260–287. Van Opzeeland IC, Van Parijs SM (2004) Individuality in harp seal, Phoca groenlandica, pup vocalizations. Animal Behaviour 68:1115–1123. Wadewitz P, Fischer J, Hammerschmidt K, Battaglia D, Sennhenn-Reulen H (2016) DoTC: Distribution of Typicality Coefficients. R package version 0.2. Wadewitz P, Hammerschmidt K, Battaglia D, Witt A, Wolf F, Fischer J (2015) Characterizing Vocal Repertoires — Hard vs . Soft Classification Approaches. PLoS ONE 10:1–16. Webster TA, Dawson SM, Rayment WJ, Parks SE, Van Parijs SM (2016) Quantitative analysis of the acoustic repertoire of southern right whales in New Zealand. The Journal of the Acoustical Society of America 140:322–333. Wheeler BC, Fischer J (2012) Functionally referential signals: A promising paradigm whose time has passed. Evolutionary Anthropology 21:195–205. Whitehead H (1983) Structure and stability of humpback whale groups off Newfoundland. Can J Zool 61:1391–1397. Wood JD, McCowan B, Langbauer Jr WR, Viljoen JJ, Hart LA (2005) Classification of african elephant loxodonta Africana rumbles using acoustic parameters and cluster analysis. Bioacoustics 15:143–161. Yin S, McCowan B (2004) Barking in domestic dogs: Context specificity and individual identification. Animal Behaviour 68:343–355. Zadeh LA (1965) Fuzzy sets. Information and Control 8:338–353.

Chapter 4

Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

4.1 Abstract

The ‘social complexity hypothesis for communicative complexity’ (SCHCC) states that animals living in complex social environments demonstrate a high level of communicative complexity. Standard metrics used to define a species’ communicative complexity are the number of call types within their repertoire and the use of graded calls that have few or no boundaries between call types. Graded calls are thought to be particularly useful for social species given that they can contain flexible information related to a variety of signaller attributes, including motivation, arousal, and/or dominance status. Humpback whales are not thought to have high social complexity, yet they use a large repertoire of calls that includes graded call types. This does not appear to fit with the concepts outlined in the SCHCC. Here, we investigated the call use of humpback whales during interactions of variable social complexity to determine how they are mediated vocally. A cluster analysis revealed four social states, from simple female-calf dyads, to highly social groups of multiple male escorts. Results show that graded calls are more common overall than discrete calls, however the ‘unstable’ social state used a larger proportion of discrete calls compared to the other states. The number of different call types was highest during the ‘competitive’ social state. Combined with the high proportion of graded calls, this indicates an increased benefit or greater need to communicate a wider variety of potential information in this state. Our results provide support for the idea that multiple-escort groups engaging in competitive behaviour are socially complex, and show that these interactions correlate with higher communicative complexity. The results also show that larger groups may not be necessarily more complex, and group behaviour is the most important indication of group complexity. This ultimately supports the SCHCC in a novel way.

Keywords: communicative complexity, graded repertoire, humpback whale, social complexity, SCHCC

88 Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

4.2 Introduction

There is a direct link between the level of social complexity and the level of informational detail that an animal needs to communicate (Monticelli and Ades 2013). Lamarck (1809) and Darwin (1872) proposed that social animals have a greater need for, and dependence on, within- species communication compared to more solitary animals. Highly social species should have a rich and diverse communication system with which to convey a wider range of information, such as the identity, behaviour, and motivation of groups and individuals (Marler 1976; Waser 1982; Freeberg et al. 2012). This concept is known as the ‘social complexity hypothesis for communicative complexity’ (SCHCC) (Freeberg et al. 2012), meaning animals living in a complex social environment have developed greater communicative complexity for relaying important information and facilitating social interactions (Freeberg 2006; Pika 2017; Peckre et al. 2019). The most widely applicable metric used to quantify communicative complexity is the number of signals or displays in a species’ repertoire (Oller and Griebel 2008; Freeberg et al. 2012; Pika 2017; Peckre et al. 2019). A socially complex species tends to use a larger number of signal types compared to a species with a simpler social structure [e.g. North American wrens, family Troglodytidae (Kroodsma 1977); non-human primates (McComb and Semple 2005; Maestripieri 2007; Delbarco-Trillo et al. 2011; Bouchet et al. 2013); seals, family Phocidae (Stirling and Thomas 2003); sciurid rodents, family Sciuridae (Blumstein and Armitage 1997; Pollard and Blumstein 2012); lizards (Ord and Garcia-Porta 2012)]. In the acoustic domain, which is the most heavily researched signal modality (Peckre et al. 2019), this manifests as the number of call types. However, while a wide repertoire of call types is considered evidence of communicative complexity, repertoire size may be difficult to establish in species with considerable variability, or gradation, between calls (Green and Marler 1979; Hammerschmidt and Fischer 1998; Freeberg et al. 2012; Manser et al. 2014; Garland et al. 2015). Thus repertoire size alone may not be the best indicator of communicative complexity (Fischer et al. 2017b) In contrast to stereotyped discrete signals, graded signals exist along a continuum, with few or no clear boundaries between signal types (Marler 1977; Keenan et al. 2013; Peckre et al. 2019). Gradation may function to convey subtle changes about the signaller, including motivation, emotional or physiological arousal, valence, and/or intent (Marler 1967, 1976; Morton 1977; Green and Marler 1979; Owings and Morton 1998; Manser 2013). However, potential information could also be derived from the call type itself, as graded calls can be perceived as discrete categories (Fischer 2006). Further, some signals can contain ‘multiple messages’ related to both fixed and flexible information (Gerhardt 1992). Graded signals therefore have the potential to contain more

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale information than discrete signals (Manser 2013; Peckre et al. 2019). A graded communication system would be more likely to develop in a socially-complex species with frequent social interactions (Freeberg et al. 2012), potentially as a way to engage in sophisticated affiliative interactions (Rebout et al. 2020). Gradation in a repertoire and the presence of graded signals is considered an additional metric of communicative complexity (Freeberg et al. 2012; Fischer et al. 2017b; Peckre et al. 2019). Humpback whales (Megaptera novaeangliae) are an interesting species for the study of social and communicative complexity as they do not seem to be a traditional fit for the SCHCC. They have a complex and diverse communication system, including their songs, one of the most complex acoustic signals in animal behaviour (Payne and McVay 1971; Payne et al. 1984). However, humpback whales also have a wide and highly variable repertoire of non-song ‘calls’ (Silber 1986; Thompson et al. 1986; Clark 1990; Dunlop et al. 2007, 2008; Stimpert et al. 2011; Rekdahl et al. 2013, 2017; Fournet et al. 2015, this thesis Chapters 2 and 3). The repertoire is likely to contain between eight and 16 stable call types (i.e. present across years) (Dunlop et al. 2007; Stimpert et al. 2011; Rekdahl et al. 2013, 2017; Fournet et al. 2015). In addition to these, studies in eastern Australian humpbacks in particular indicate that other social calls may be less stable, including ‘song unit social sounds’ (Dunlop et al. 2007; Rekdahl et al. 2013) and other ‘inconsistent’ calls (Rekdahl et al. 2013). Including these calls increases the repertoire of east Australian humpbacks to between 34 and 46 call types in any given year. This repertoire size is large compared to other baleen whales, which have relatively few (2-8) well described call types [e.g. blue whale, Balaenoptera musculus (McDonald et al. 2001); fin whale, Balaenoptera physalus (Watkins et al. 1987; Širović et al. 2015); southern right whale, Eubalaena australis (Clark 1982); North Atlantic right whale, Eubalaena glacialis (Parks and Tyack 2005; Parks et al. 2011); gray whale, Eschrichtius robustus (Cummings et al. 1968; Burnham et al. 2018)]. Further, the call repertoire of humpback whales includes both discrete, context-specific calls (Jurasz and Jurasz 1979; D’Vincent et al. 1985; Cerchio and Dahlheim 2001; Parks et al. 2014; Fournet et al. 2018), as well as graded calls which likely contain flexible information related to motivation or arousal (Dunlop et al. 2008; Dunlop 2017). Overall, the large number of call types, the apparently variable use of some but obligate use of others, and the mix of discrete and graded signals, provides evidence for a high level of communicative complexity. However, this does not match their described social complexity. Humpback whales have an unstable, and supposedly simple, social structure (Berta and Sumich 1999; May-Collado et al. 2007). This is in large part because they do not live in permanent

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale or stable groups, unlike the complex societies of many toothed whales or primates, for example (Clapham 1993, 1996). Rather, social groups are typically small and unstable (Whitehead 1983; Baker and Herman 1984; Mobley, Jr. and Herman 1985; Clapham et al. 1992; Clapham 1993, 1996, 2000; Mattila et al. 1994; Valsecchi et al. 2002; Ramp et al. 2010). There is no evidence for permanent dominance sorting (Clapham et al. 1992; Clapham 1996). This is presumably because of low encounter rates with known and specific individuals due to wide-ranging movement patterns and large population sizes (Clapham 1996, 2009; Palsbøll et al. 1997; Clapham and Zerbini 2015; Noad et al. 2019). However, individuals do frequently interact, albeit without forming long-term bonds and little evidence for the role of kinship in associations (Clapham et al. 1992; Clapham 1993, 1996, but see Weinrich 1991). This occurs most notably during cooperative feeding (D’Vincent et al. 1985; Clapham 1993; Sharpe 2001; Wiley et al. 2011; Parks et al. 2014) and during presumed breeding interactions (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Clapham et al. 1992; Clapham 1996; Felix and Novillo 2015). Social interactions associated with breeding behaviour in humpback whales can be broadly categorized as a male 'escorting' a female, and larger competitive groups, where males physically compete with each other, presumably for access to a female (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Clapham et al. 1992; Clapham 1996; Felix and Novillo 2015). The complexity of the social interactions are variable, with larger competitive groups likely representing an extreme due to the number of potential interactions as well as the diversity of relationships between group members. Competitive groups typically consist of multiple males competing for the position of primary escort to a single female. While this competitive behaviour is undoubtedly related to breeding, no mating has been observed in these groups (Clapham 2000; Pack et al. 2002). Further, all male groups have been reported, which potentially function in dominance sorting (Clapham et al. 1992; Brown and Corkeron 1995), particularly as secondary escorts also engage in agonistic interactions amongst themselves (Herman et al. 2007). Further, although most of the behaviour during competitive groups is agonistic, there are reports of male humpback whales working cooperatively during competitive behaviour (Clapham et al. 1992; Felix and Novillo 2015). These observations indicate a broader function for these groups than simply breeding, and complex social relationships between the animals involved (Felix and Novillo 2015). Although social complexity is traditionally defined by group size (Freeberg et al. 2012; Silk et al. 2013; Bergman and Beehner 2015; Fischer et al. 2017a; Kappeler 2019; Peckre et al. 2019), the lack of stable groups and the absence of a permanent dominance structure have been cited as additional metrics because they create a more fluid social dynamic and necessitate more frequent

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale social interactions (Freeberg et al. 2012; Peckre et al. 2019). This metric may apply to the breeding behaviour of humpback whales, where intense agonistic interactions occur. As such, this could correlate with the complexity of the humpback whale communication system. As a preliminary test to see if the group behaviour of humpback whales fit with the SCHCC, we quantified their call complexity in groups with varying degrees of social interaction, used here as a proxy for potential differences in group social complexity. Since graded calls are proposed to be highly informative regarding the motivational state or arousal of the signaller (Marler 1967, 1976; Green and Marler 1979; Owings and Morton 1998; Manser 2013), it was expected that graded calls would be used significantly more often in groups engaging in social interactions. Specifically, it is hypothesised that in these groups, 1) the number of discrete and graded call types detected (i.e. repertoire size) will be higher, as well as the use of graded calls in particular; and 2) graded calls will have a higher level of acoustic variability in groups engaging in social interactions (i.e. more male escorts), presumably to relay flexible information. The results of this study will provide a basis for making inter-species comparisons between social and communicative complexity, which can ultimately advance our understanding of the evolution of communication in many of these species.

4.3 Methods

Data on east Australian humpback whales were collected in September and October 2010, 2011, 2014, and 2017 off the eastern coast of Australia during their southward migration from breeding grounds in the Great Barrier Reef, to feeding grounds in the Antarctic. Humpback whales on their southward migration continue to display typical breeding behaviours such as singing, males joining and escorting females, and the formation of competitive groups (albeit smaller and with less energy than on the breeding grounds proper; Corkeron et al. 1994). Acoustic data were collected using two types of non-invasive digital recording tags attached to adult females accompanied by calves: DTAGS (Johnson and Tyack 2003) in 2010, 2011, and 2014, and Acousonde 3B tags (Greeneridge Sciences, Inc., http://www.acousonde.com) in 2017. Both tag types are equipped with a hydrophone to record the acoustic environment around the tagged animal. DTAGS were programmed to sample 16-bit audio at either 48 kHz (2011) or 96 kHz (2010, 2014) with a 400 Hz Butterworth high-pass filter to reduce low frequency flow noise. Acousondes were programmed to sample 16-bit audio at 25.8 kHz. The first ten minutes of all tag deployments were excluded from any analyses to allow the animal to return to ‘normal’ (pre-tagging) behaviour (Williamson et al. 2016).

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

4.3.1 Social states To determine the social context of calls, we first quantitatively identified the social states of the whales. The tagged lactating females were focal followed for the duration of the tag deployment using focal sampling methods (Altmann 1974). Focal follows were divided into ten-minute time bins, and nine variables were assessed in each bin. Eight of these were obtained from the boat-based follow data: number of escorts for the majority (> 50%) of observations; predominant group interaction (i.e. stable or splitting/joining of animals); number of blows; number of fluking dives started per whale; a count of high energetic surface behaviours (breaching, fluke slapping, pectoral fin slapping, head lunging) per whale; a count of low energetic surface behaviours (spy hopping, pectoral fin waving) per whale; average speed (km/hr) the whales travelled from the first observation in each bin to the last; and average course variation (degrees) of the group from the first observation to the last. The last variable, proportion of time the tagged animal spent at the surface, was measured from the dive record of the tagged animal. As per (Kavanagh 2014), all time bins were run through an unsupervised cluster analysis to objectively determine the social states based on the chosen variables. Due to the mixed data (both categorical and numeric), the partitioning around medoids clustering algorithm was used. This is a robust method which minimises the sum of dissimilarities of observations (Kaufman and Rousseeuw 1990; Maechler et al. 2018). First, the pairwise distances between all of the data points were computed to obtain the sum of dissimilarities using the Gower coefficient (Gower 1971; Maechler et al. 2018). The resulting dissimilarity matrix was used to run the cluster analysis. To determine the optimal number of clusters (social states) in the data set, multiple iterations were run. The final clustering scheme used the number of clusters that yielded the highest silhouette width, which is an internal validation metric representing how similar objects are within- and between clusters (Kaufman and Rousseeuw 1990). Analyses were conducted in R (R Core Team 2018) using the cluster package (Maechler et al. 2018). Behavioural focal follow data from 2017 were collected from the same observers as previous years, however the methodology was different. Therefore, while behavioural data included the timing of splits and joins, and the number of animals in the group, no counts were available for the number of each behaviour. As such, these data were excluded from the cluster analysis. However, using the results of the cluster analysis as a proxy, most time bins from 2017 were able to be assigned to a social state. Those which were unable to be assigned to a state were excluded from further analyses.

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

4.3.2 Acoustic behaviour Spectrograms of recordings were browsed visually and aurally in Raven Pro 1.5 (Bioacoustics Research Program 2017) using a Hann window, Fast Fourier Transform size of either 4096 or 2048 samples (depending on the sampling rate), and 50% overlap. All humpback whale vocal non-song signals (‘calls’) were marked and extracted for further analysis. A clip of surrounding background noise before or after the associated call was also extracted. The noise clip was at least 0.5 seconds in duration and covered a bandwidth identical to the associated call clip. A custom Matlab script (MathWorks 2018) removed this noise energy from the call by subtracting the spectrum of the noise clips from the corresponding calls (Girola et al. 2019). The calls extracted from the DTAGS were corrected for the high-pass filter, but no correction was needed for the calls from the Acousonde as it did not have a similar filter (see Chapter 2 for details). The same custom script measured three acoustic features from the de-noised call clips which convey motivation and arousal in other species (Marler et al. 1992; Dunlop 2017; McGrath et al.

2017; Mandl et al. 2019): the duration (Dur) of the signal; peak frequency (FP); and inter-quartile bandwidth (FIQ) (Girola et al. 2019; Chapters 2 and 3). Additionally, aggregate entropy (AgEnt) was measured in Raven from the original (not de-noised) sound clips as this measurement was not included in the Matlab script (Girola et al. 2019). This resulted in a total of four acoustic parameters. Frequency measurements (FP and FIQ) were converted to a logarithmic scale to account for the mammalian perception of pitch (Richardson et al. 1995; Cardoso 2013). To increase the confidence that calls originated from the focal group, a signal-to-noise ratio (SNR) threshold was implemented. A simulated call (a 0.5 second signal with a shape between 50 Hz and 1 kHz) was created in Matlab to determine the point at which frequency measurements were no longer reliable (see Chapter 2 for details). This threshold was calculated to be -3 dB (Girola et al. 2019). While a SNR of 20 dB may be necessary to assign calls to a focal animal, calls from the group were the focus of the present study (Stimpert et al. 2020). Therefore, a SNR cut-off of 5 dB was selected to ensure only the highest quality calls were included from the focal group while retaining enough calls for a relatively high sample size for most call types.

4.3.2.1 Call classification

Chapter 3 of this thesis assessed a combination of techniques, including fuzzy clustering, for classifying the repertoire of humpback whales. The results indicated a total of 15 call types, six of which were relatively discrete (D) and nine which were more graded (G): ‘low entropy snort/knock’

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

(G), ‘wup/low frequency eeaw’ (G), ‘grumble/long snort’ (G), ‘discrete snort’ (D), ‘broadband bop’ (G), ‘high frequency squeak’ (G), ‘squeak/high frequency eeaw’ (G), ‘low entropy bop’ (G), ‘high frequency bop’ (G), ‘high entropy bop’ (G), ‘meow’ (D), ‘paired croaks’ (D), ‘spiccato’ (D), ‘thwop’ (D), and ‘wop’ (D). Spectrograms of the 15 sound types can be found in Appendix 3, Figure A3.1, and the four acoustic measurements in Table 4.1. After classification, each individual call was assigned to the social state in which it was detected. The number of different call types, the use of discrete and graded calls, and changes in the four acoustic parameters were then quantified between social states.

Table 4.1 Mean ± SD of the acoustic measurements for each call type. D: discrete, G: graded. N (% Total Inter-quartile Aggregate Call Type Duration Peak Frequency Calls) Bandwidth Entropy Low entropy 219 (11.7%) 0.18 ± 0.13 138.8 ± 39.9 106.0 ± 62.8 4.14 ± 0.69 snort/knock (G) Wup/low frequency 293 (15.7%) 0.26 ± 0.20 345.7 ± 110.8 181.3 ± 144.0 4.98 ± 0.73 eeaw (G) Grumble/long snort 185 (9.9%) 0.32 ± 0.31 215.2 ± 64.0 141.7 ± 109.4 4.63 ± 0.74 (G) Discrete snort (D) 119 (6.4%) 0.22 ± 0.29 121.0 ± 59.4 113.0 ± 90.1 4.31 ± 0.79

Broadband bop (G) 91 (4.9%) 0.18 ± 0.13 823.9 ± 286.8 497.8 ± 341.8 5.57 ± 0.60 High frequency 82 (4.4%) 0.20 ± 0.14 2336.7 ± 735.4 537.1 ± 430.4 6.03 ± 1.01 squeak (G) Squeak/high 95 (5.1%) 0.35 ± 0.27 1158.9 ± 521.8 868.5 ± 531.5 6.10 ± 0.93 frequency eeaw (G) Low entropy bop 237 (12.7%) 0.14 ± 0.08 417.0 ± 135.0 106.7 ± 61.5 3.74 ± 0.69 (G) High frequency bop 113 (6.0%) 0.14 ± 0.07 990.5 ± 351.5 155.4 ± 77.2 4.24 ± 0.74 (G) High entropy bop 162 (8.6%) 0.15 ± 0.08 496.4 ± 142.5 294.0 ± 205.5 5.08 ± 0.66 (G) Meow (D) 141 (7.5%) 0.24 ± 0.10 583.7 ± 506.1 286.6 ± 377.1 5.53 ± 1.02

Paired Croaks (D) 43 (2.3%) 0.40 ± 0.05 114.6 ± 50.4 74.9 ± 30.3 3.64 ± 0.90

Spiccato (D) 40 (2.1%) 1.78 ± 1.61 223.2 ± 176.1 150.9 ± 118.0 4.47 ± 0.83

Thwop (D) 16 (0.8%) 0.68 ± 0.24 132.2 ± 90.4 101.8 ± 53.7 4.30 ± 0.71

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Wop (D) 35 (1.9%) 0.79 ± 0.64 154.9 ± 150.9 125.7 ± 112.6 4.66 ± 0.63

4.3.2.2 The use of discrete and graded call types

To investigate changes in the proportion of each call type and category (discrete or graded) within the different social states (hypothesis 1), we ran generalised linear mixed models (GLMMs) with binomial error distributions (Crawley 2013). All models were run in R using base packages, with post-hoc analyses using emmeans (Lenth 2018). GLMMs were chosen to compensate for the non-constant variance and strictly bounded data associated with proportion data. Separate models were run for each call type and category, with social state as the explanatory variable and the internally calculated proportion of calls detected within that state as the response variable. Post-hoc analyses compared these proportions to ultimately determine whether a call type was used more often in certain social states than others.

4.3.2.3 Call variability between states

To test whether the acoustic parameters of graded calls are more variable during social contexts (hypothesis 2), we calculated the coefficient of variation (CV) test statistic. The CV is simply a standardised measure of spread, and is calculated as the standard deviation divided by the mean (Pearson 1896; Krishnamoorthy and Lee 2014). It is a useful metric to compare variation in samples with different units of measure and/or very different means, and has been used in a variety of studies to investigate call complexity (Lemasson and Hausberger 2011; Bouchet et al. 2013; Peckre et al. 2019). To compare CVs across social states, a modified signed-likelihood ratio test (MSLRT) for equality of CVs was run for each call type and both call categories, with separate tests for each of the four call parameters (Krishnamoorthy and Lee 2014; Marwick and Krishnamoorthy 2019). Compared to the frequently used asymptotic test for the equality of CV (Feltz and Miller 1996), the MSLRT has lower type I error rates and is better able to deal with unequal sample sizes. Tests were run in R using the package cvequality (Marwick and Krishnamoorthy 2019).

4.4 Results

4.4.1 Social states A total of 24 focal follows were conducted on tagged lactating female humpback whales including more than 54 hours of data (Appendix 3, Table A3.1). The results of the cluster analysis

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale suggested a four-cluster solution corresponding to four states. Group composition was primarily responsible for how clusters split, providing similar results to those in Chapter 2, followed by group stability (splitting/joining of group members). The first state (‘FC’) was comprised of female-calf only groups, constituting most of the focal follows (189 of 311 (61%) time bins analysed, Table 4.2). Female-calf pairs typically exhibited slower swim speeds, high course variation (i.e. ‘milling’), and a large number of surface active behaviours, particularly from the calf. The second state (‘FCE’) was characterised by the female and calf being escorted by one other whale (presumably male), relatively high blow rates, and low course variation. The third state (‘unstable’) was comprised of female-calf dyads with one or more escorts, and was characterised by frequent splitting and joining of group members. These groups typically had high travel speeds, minimal course variation, and relatively few surface behaviours. The final state (‘competitive’) was comprised of a female-calf dyad with multiple escorts and had a very high number of active surface behaviours, particularly those considered aggressive (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Felix and Novillo 2015). (Table 4.2). The four social states had varying degrees of social interaction. FC groups were not considered to be particularly social given the absence of social interactions with conspecifics and the lack of membership change (splitting and joining). The remaining three states were presumed to have higher degrees of social interaction due to the increased number of animals within the group and diversity of interactions that are reported to occur.

Table 4.2 Results of the social state cluster analysis, indicating the mean ± SD of each variable and the proposed social state for each cluster. FC: female-calf pair, FCE: female-calf-escort group, FCME: female-calf-multiple escort group. Social State ‘FC’ ‘FCE’ ‘Unstable’ ‘Competitive’ Predominant Group FC FCE FCE-FCME FCME Composition Group Interaction Stable Stable Unstable Stable Average Blows 7.47 ± 3.28 7.21 ± 0.28 6.90 ± 0.89 6.65 ± 0.54 (per whale/bin) Fluking Dives Started 0.59 ± 0.63 0.41 ± 0.04 0.60 ± 0.10 0.66 ± 0.09 (per whale/bin) Aggressive Behaviours 0.34 ± 1.67 0.26 ± 0.08 0.17 ± 0.12 0.89 ± 0.26 (per whale/bin)

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Active Behaviours 0.47 ± 0.72 0.15 ± 0.04 0.05 ± 0.03 0.56 ± 0.15 (per whale/bin) Proportion of Time at Surface (tagged 0.27 ± 0.26 0.23 ± 0.03 0.35 ± 0.09 0.39 ± 0.05 animal) Average Speed 3.56 ± 1.85 3.85 ± 0.17 5.38 ± 0.58 3.42 ± 0.18 (km/hour) Average Course 41.52 ± 56.01 28.04 ± 3.75 27.16 ± 8.07 28.76 ± 5.59 Variation (degrees) Number of Bins 189 87 10 25

Calls 132 558 202 979

4.4.2 Acoustic behaviour

4.4.2.1 Call type and category usage

A total of 1,871 calls were selected for acoustic analysis (Appendix 3, Table A3.1). There were three times more graded calls detected than discrete calls (n = 1,477 and 394, respectively), and all social states used a higher proportion of graded calls than discrete calls (Table 4.3). However, the use of both discrete and graded calls differed markedly between the four social states (Figure 4.1, Figure 4.2). Contrary to expectations, the proportion of graded calls used within the repertoire was significantly higher in the ‘FC’ state (the unaccompanied female-calf dyad) compared with the remaining three states (‘FCE’, estimate = 1.88±0.47, z ratio = 4.009, p = 0.0003; ‘unstable’, estimate = 2.12±0.48, z ratio = 4.386, p < 0.0001; ‘competitive’, estimate = 2.03±0.46, z ratio = 4.384, p < 0.0001; Table 4.4). The opposite was true for discrete calls, which were used in significantly higher proportions in the ‘FCE’, ‘unstable’, and ‘competitive’ states (Table 4.3, Table 4.4). However, several discrete call types showed some context specificity. ‘Spiccatos’ were only detected in the ‘competitive’ social state, while ‘discrete snorts’ were only detected in the ‘FCE’ and ‘competitive’ states (Table 4.3). The inclusion of these discrete call types meant that the vocal repertoire was largest in groups with more social interactions (Figure 4.1). Surprisingly however, the number of calls in the repertoire was similar for the ‘FC’ and ‘unstable’ social states, despite the difference in the presumed number and diversity of social interactions (Table 4.3).

Table 4.3 Proportion ± SE of each call type and category that comprise the four social states. Note that columns, not rows, add up to 1.0. D: discrete, G: graded. Social State

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Call Category/ FC FCE Unstable Competitive Type Discrete Calls 0.04 ± 0.02 0.20 ± 0.02 0.25 ± 0.03 0.23 ± 0.01

Graded Calls 0.96 ± 0.02 0.80 ± 0.02 0.75 ± 0.03 0.77 ± 0.01

Low entropy snort/knock (G) 0.14 ± 0.03 0.13 ± 0.01 0.04 ± 0.01 0.12 ± 0.01

Wup/low frequency eeaw (G) 0.26 ± 0.04 0.14 ± 0.01 0.16 ± 0.03 0.15 ± 0.01

Grumble/long snort (G) 0.14 ± 0.03 0.08 ± 0.01 0.13 ± 0.02 0.09 ± 0.01

Discrete snort (D) 0 0.10 ± 0.01 0 0.07 ± 0.01

Broadband bop (G) 0.06 ± 0.02 0.04 ± 0.01 0.05 ± 0.01 0.05 ± 0.01

High frequency squeak (G) 0.02 ± 0.01 0.05 ± 0.01 0.08 ± 0.02 0.04 ± 0.01

Squeak/high frequency eeaw (G) 0.02 ± 0.01 0.05 ± 0.01 0.11 ± 0.02 0.04 ± 0.01

Low entropy bop (G) 0.13 ± 0.03 0.14 ± 0.01 0.08 ± 0.02 0.13 ± 0.01

High frequency bop (G) 0.08 ± 0.02 0.08 ± 0.01 0.04 ± 0.01 0.05 ± 0.01

High entropy bop (G) 0.11 ± 0.03 0.08 ± 0.01 0.06 ± 0.02 0.09 ± 0.01

Meow (D) 0.01 ± 0.01 0.07 ± 0.01 0.22 ± 0.03 0.06 ± 0.01

Paired Croaks (D) 0 0.01 ± 0.003 0.01 ± 0.003 0.04 ± 0.01

Spiccato (D) 0 0 0 0.04 ± 0.01

Thwop (D) 0.01 ± 0.01 0.02 ± 0.005 0 0.01 ± 0.002

Wop (D) 0.02 ± 0.01 0.01 ± 0.005 0.02 ± 0.01 0.02 ± 0.005

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Figure 4.1 Composition of the discrete call type repertoire for the four social states, indicating the proportions of each call type.

Graded call types were used in all four social states, however certain call types were used more frequently in some states than others (Figure 4.2, Table 4.3). This included ‘high frequency squeak’ and ‘squeak/high frequency eeaw’, which were used most often in the ‘unstable’ social state. This was significant when compared with the ‘competitive’ state (‘high frequency squeak’, estimate =-0.88 ± 0.31, z ratio = -2.88, p = 0.0178; ‘squeak/high frequency eeaw’, estimate = -0.98 ± 0.27, z ratio = -3.57, p = 0.0016; Table 4.4). For ‘squeak/high frequency eeaw’, this result was also statistically significant compared to the ‘FCE’ (estimate = -0.88 ± 0.30, z ratio = -2.93, p = 0.0161) and ‘FC’ states (estimate = -1.66 ± 0.63, z ratio = -2.65, p = 0.0359) (Table 4.4). ‘Wup/low frequency eeaw’, which are presumed calf calls (Indeck et al. 2020), comprised the largest proportion of calls in the ‘FC’ state (Figure 4.2, Table 4.3). This result was significant compared to the ‘FCE’ (estimate = 0.74 ± 0.23, z ratio = 3.19, p = 0.0072) and ‘competitive’ states (estimate =

100

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

0.67 ± 0.22, z ratio = 3.06, p = 0.0112; Table 4.4). This suggests that calves tend to use graded signals.

Figure 4.2 Composition of the graded call type repertoire for the four social states, indicating the proportions of each call type.

Table 4.4 Results of the generalised linear mixed models indicating the social states with statistically significant differences in the proportion of any call type detected. A negative estimate and z ratio indicate the estimate is lower in the first intensity level listed, and an asterisk indicates statistical significance at the p < 0.05 level. Note: only significant results are included in the table, therefore not all social state contrasts or results are listed. D: discrete, G: graded. Social State Contrasts FC vs. FC vs. FCE vs. FCE vs. Competitive Call Type FC vs. FCE Unstable Competitive Unstable Competitive vs. Unstable -1.88 ± 0.47 -2.12 ± 0.48 -2.03 ± 0.46 Discrete z = -4.009 z = -4.386 z = -4.384

101

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

p = 0.0003 p < 0.0001 p < 0.0001 Low entropy 1.41 ± 0.44 1.33 ± 0.38 1.19 ± 0.38 snort/knock z = 3.211 z = 3.476 z = 3.186 (G) p = 0.0061 p = 0.0024 p = 0.0069 Wup/low 0.74 ± 0.23 0.67 ± 0.22 frequency z = 3.189 z = 3.057 eeaw (G) p = 0.0072 p = 0.0112 High -0.88 ± 0.31 frequency z = -2.879 squeak (G) p = 0.0178 Squeak/high -1.66 ± 0.63 -0.88 ± 0.30 -0.98 ± 0.27 frequency z = -2.650 z = -2.925 z = -3.565 eeaw (G) p = 0.0359 p = 0.0161 p = 0.0016 -3.63 ± 1.02 -1.34 ± 0.24 -1.55 ± 0.22 Meow (D) z = -3.562 z = -5.649 z = -7.122 p = 0.0015 p < 0.0001 p < 0.0001 -2.04 ± 0.60 Paired z = -3.388 croaks (D) p = 0.0025

4.4.2.2 Call variability between states

Duration and aggregate entropy were the only call parameters with significant differences in the coefficient of variation (CV) between discrete and graded call types in the four social states. For these parameters, the lowest CV, being the least amount of within-call-type variability, was always lowest in the ‘FC’ state for both discrete and graded call types (Table 4.5). This was true even though the ‘FC’ state contained the most graded calls of all the social states, indicating there is substantial variability in duration and aggregate entropy in graded calls produced in the other three social states. There were no significant differences in within-call variability in peak frequency or inter-quartile frequency across social states (Table 4.5).

Table 4.5 Comparison of the coefficient of variation (CV) in discrete and graded calls for the four acoustic parameters across each social state. A significant result (p < 0.05) from the modified signed-likelihood ratio test (MSLRT) is indicated by a * next to the parameter, with the highest and lowest CVs indicated in bold. Social State Call Type Parameter Test Statistic FC FCE Unstable Competitive MSLRT=15.33 Discrete Duration 54.27 76.47 60.65 153.69 p=0.0016*

102

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Peak MSLRT=5.21 96.34 139.27 72.42 140.24 frequency p=0.1571 Inter-quartile MSLRT=3.59 34.09 134.38 150.39 147.46 frequency p=0.3093 Aggregate MSLRT=17.87 2.75 23.31 19.76 24.35 Entropy p=0.0005* MSLRT=34.80 Duration 56.54 105.16 76.27 90.21 p<0.0001* Peak MSLRT=3.45 84.95 99.69 106.18 104.05 frequency p=0.3276 Graded Inter-quartile MSLRT=2.87 118.02 117.03 142.33 113.99 frequency p=0.4127 Aggregate MSLRT=57.09 12.24 20.98 22.47 22.81 Entropy p<0.0001*

Within the 15 individual call types, there were few significant differences in CVs across social states (Appendix 3, Table A3.2). Of the significant results, aggregate entropy and duration were the most consistent, and were lowest in the state with the fewest presumed social interactions (i.e. the ‘FCE’ state if that call type was not detected in the ‘FC’ state). The other two acoustic features did not demonstrate any consistent trends in within-call variability across states (Appendix 3, Table A3.2). These results indicate that the acoustic features of calls were not significantly variable between social states for most call types.

4.5 Discussion

The social complexity hypothesis for communicative complexity (SCHCC) proposes that social species require a complex vocal repertoire with which to mediate social interactions. The humpback whale does not initially appear to fit with this hypothesis, given its ephemeral social structure and apparent lack of social bonds, yet large and complex social communication repertoire. However, social complexity can also be evidenced by an egalitarian society with no clear dominance structure and frequent role reversals, particularly during agonistic interactions (Freeberg et al. 2012). This model may better describe humpback whale social organisation. If these social interactions are indeed complex, the SCHCC predicts that a complex vocal repertoire should be required to mediate these interactions. Here, we found that humpback whales, whilst forming temporary competitive groups, are likely to use an increasingly complex communication repertoire. These groups are presumed to be high in social complexity because of large group size and the diversity of relationships that occur, and this assumption was reflected in the results. Competitive

103

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale groups used an increased number of different call types, particularly ‘discrete’ call types. Further, while the parameters of most call types did not significantly vary between social states, the variability that did occur was present in both ‘discrete’ and ‘graded’ calls, and typically highest in competitive groups. Our results indicate that humpback whales use a more variable repertoire, both in the number of call types and the variability within call types, in social states with a high level of presumed social complexity. Discrete calls are likely to convey more precise, unambiguous information than graded signals (Green and Marler 1979). This can include fixed information about the sender that remains stable or changes slowly, such as age-class, sex, individual identity, and to a limited extent, body size (although see Fitch and Hauser 2002). Communicating some of these features vocally might allow individuals to assess conspecifics remotely, potentially removing the need for physical confrontation (Zahavi 1982). For example, the roars of red deer stags (Cervus elaphus) contain information regarding sex, age, and fighting ability, and other stags use this information to remotely assess their competitors (Clutton-Brock and Albon 1979; Reby and McComb 2003; Reby et al. 2005). Here, we found the proportion of discrete calls was significantly higher in the ‘FCE’, ‘competitive’, and ‘unstable’ states. These states were characterised by the presence of one (‘FCE’) or multiple (‘competitive’ and ‘unstable’) escorts, as well as splitting and/or joining of escorts (‘unstable’). In the present study, the high proportion of discrete calls in social contexts with at least one escort could signify the importance of conveying specific information like body size to potential competitors or mates. While the specific information content of humpback calls is unknown, the discrete and stereotyped ‘upcalls’ of North Atlantic right whales contain information related to age, body size, and individuality (McCordic et al. 2016). It is likely that humpback calls contain similar information, although playback studies would be required in order to determine whether this information is actually used by these animals. There was also some level of context specificity in discrete calls. Of the six discrete call types, four were never heard in the ‘low social complexity’ state, meaning they were only detected in the presence of male escorts. The ‘spiccato’ in particular was only detected in the ‘competitive’ state. This call type, while discrete, demonstrates some variability in duration (Chapter 2). This is similar to the discrete ‘type one loud calls’ of male pig-tailed langurs (Simias concolor), which are proposed to function in mediating intergroup interactions between males and vary only in the duration of the call (Tenaza 1989). Future research should look for a correlation between fixed information and acoustic features in some of the most common discrete calls. Again, playback

104

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale studies will be needed to determine if and how this information is actually processed by conspecifics (Fitch and Hauser 1995; Fischer 1998). Interestingly, like the ‘FCE’ state, only three (of six) discrete call types were heard in the ‘unstable’ state (Table 4.3, Figure 4.1). This result could indicate that there are similar underlying motivations or arousal levels in both states, although this seems unlikely given the disparity in behaviour. Alternatively, it could indicate that, contrary to expectation, the ‘unstable’ social state is not high in social complexity despite a larger number of potential social partners. Further support for this comes from the absence of one of the most common discrete call types (‘discrete snorts’) in the unstable state. This call type is short, low frequency, and narrow-band (Chapter 3), however despite being classified as discrete, there is significant variability between the two states in which it was detected (Appendix 3, Table A3.2). Although the exact function of ‘discrete snorts’ is unknown, based on the acoustic features of this call type and its variability, it likely conveys multiple messages, similar to the ‘snorts’ of male rock hyraxes (Procavia capensis) (Koren and Geffen 2009; Demartsev et al. 2016; Weissman et al. 2019). Such a call type would potentially be most beneficial during complex social interactions, making its absence in the ‘unstable’ social state noteworthy. An increasingly complex social structure usually also requires some level of gradation to communication flexible information, such as the signaller’s motivation (Peckre et al. 2019). This was not well supported here, at least with the acoustic parameters and social states chosen, as only a few call types demonstrated any significant differences in signal structure across social states. However, in support of hypothesis 2, the state with the lowest coefficient of variation (CV) for aggregate entropy and duration was consistently the least socially complex state (Table 4.5, Appendix 3, Table A3.2). These two acoustic features were most important in classifying call types in Chapter 3, and so their importance here is not surprising. It can be assumed that these features are two of the more variable features of humpback whale calls and can be readily altered to convey different information, especially given that duration is an acoustic feature that is relatively easy to adjust (Janik and Slater 1997; Seyfarth and Cheney 2010). In contrast, the state with the highest CV was often either the ‘competitive’ or ‘unstable’ social states, at least for aggregate entropy and duration (Table 4.5, Appendix 3, Table A3.2). It is possible that individual variation is responsible for some of the observed increases in CVs in these states simply because of the greater number of individuals present. Alternatively, or perhaps in addition, this increased variability could be the result of a larger number of call types detected in some of the more social states (Table 4.3, Figure 4.1, Figure 4.2). However, if these were the

105

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale predominant drivers for higher variability, it would be expected that CV values for all parameters would increase in a linear or predictable way with group membership, a trend that was not observed. As such, while individual variation and/or diversity of call types may account for some of the variability in call parameters, they are likely not the only drivers. The relationship between the complexity of the social context and both peak frequency and bandwidth was less clear, and the social state with the highest and lowest CV was inconsistent. Overall, the results of the CV analysis make sense given the context specificity of some call types and the findings of Chapter 2, which indicated that humpbacks in higher arousal contexts added additional call types to their repertoire rather than adjusting the features of calls. Further research incorporating a larger selection of acoustic features is needed, as well as integrating additional behavioural and motivational data. For example, since gradation conveys information about the motivational state or arousal of the signaller, it is possible that graded calls are specific to motivational contexts (e.g. aggressive, affiliative, appeasement) rather than discrete behavioural states (Hammerschmidt and Fischer 1998), as in some terrestrial species [e.g. birds (Yasukawa 1978; Morton 1982; Waas 1991; Ręk et al. 2011); anurans (Wells and Schwartz 1984; Schwartz 1989; Wagner 1989; Grafe 1995; Reichert and Gerhardt 2013); mammals (Begg 1975; Schehka et al. 2007; Bastian and Schmidt 2008; Gadziola et al. 2012; Fedurek et al. 2015; Walter and Schnitzler 2017)]. Graded calls did not necessarily demonstrate more acoustic variability than discrete calls (Table 4.5, Appendix 3, Table A3.2). This suggests that while some call types can be considered relatively discrete, all humpback calls contain some level of variation (Marler 1977, 1978, 1984; Marler et al. 1992). Such flexibility would allow humpbacks to convey an enormous amount of information in their calls. This concept has been well described in the alarm calls of some terrestrial species, which contain information on predator type as well as motivational state or arousal (Beynon and Rasa 1989; Manser 2001; Kokolakis et al. 2010). Further, discrete call types are documented to be somewhat variable in killer whales, with duration being one of the primary acoustic features that is modified (Ford 1989). The results of the repertoire analysis in Chapter 3 support the idea of a high level of overall gradation in the repertoire, even within more stereotyped calls. Despite the consistent presence of variation, some calls do exhibit more variability than others. ‘Low entropy snort/knock’ and ‘wup/low frequency eeaw’, both graded calls, had significant variation in aggregate entropy and duration while other calls exhibited no variation in any parameter between social states (e.g. ‘high entropy bop’ and ‘high frequency bop’). The discrete ‘meow’ had significant differences in all four acoustic parameters, with consistently lower variation in the ‘FCE’

106

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale state with the exception of peak frequency. Combined with the results of the previous chapters, it is likely that while all calls may have the potential to convey flexible information, some call types are more likely to be used in this manner than others. While there were no context specific graded calls, some trends in their use did appear. There was a significantly higher proportion of graded calls used in the ‘FC’ state compared to the other three social states (Table 4.3, Table 4.4). This was unexpected considering the assumption that graded calls convey a high level of motivational or arousal information and are thus more useful in highly social contexts consisting of complex within-group interactions. However, although female- calf only groups have low call rates overall (Videsen et al. 2017; Indeck et al. 2020), a large proportion of the calls they do produce can be attributed to the calf (Zoidis et al. 2008; Indeck et al. 2020, Chapter 2). Calls from humpback calves are often similar to those produced by adults but shorter in duration and less structured, i.e. more acoustically variable (Zoidis et al. 2008; Indeck et al. 2020). Infant and juvenile calls of other species are observed to contain higher levels of some types of non-linear phenomena (NLP), which can cause an otherwise stereotyped discrete sound to look and sound qualitatively different (Fitch et al. 2002). Based on the presence of NLP in songs of male humpback whales (Cazau et al. 2016) and the infant and juvenile calls of related species [North Atlantic right whales (Parks and Tyack 2005; McCordic et al. 2016; Root-Gutteridge et al. 2018); sperm whales, Physeter macrocephalus (Watkins et al. 1988)], it would not be surprising to find evidence of NLP in humpback whale calves. If so, this could explain the high proportion of graded calls present in the ‘FC’. In conclusion, we have provided evidence that humpback whales have a complex communication system that becomes increasingly complex as within-group social interactions increase. Overall, the SCHCC holds true for humpback whales, albeit not in the traditional way. This an important step in formally testing the SCHCC, although future work will be needed to test alternative hypotheses involving additional selection pressures (e.g. predation, habitat, and resources) (Peckre et al. 2019). The results here can advance our understanding of how humpback whales use vocal signals and provide a basis for comparing the social and communicative complexity of humpback whales with other species.

4.6 References

107

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Bastian A, Schmidt S (2008) Affect cues in vocalizations of the bat, Megaderma lyra, during agonistic interactions. The Journal of the Acoustical Society of America 124:598–608. Begg RJ (1975) The agonistic vocalizations of Rattus villasissimus. Australian Journal of Zoology 23:597–614. Bergman TJ, Beehner JC (2015) Measuring social complexity. Animal Behaviour 103:203–209. Berta A, Sumich JL (1999) Marine mammals: Evolutionary biology. Academic Press, San Diego. Beynon P, Rasa OAE (1989) Do dwarf mongooses have a language-Warning vocalizations transmit complex information. South African Journal of Science 85:447–450. Bioacoustics Research Program (2017) Raven Pro: Interactive Sound Analysis Software. Cornell Laboratory of Ornithology, Ithaca (NY). Blumstein DT, Armitage KB (1997) Alarm calling in yellow-bellied marmots: I. The meaning of situationally variable alarm calls. Animal Behaviour 53:143–171. Bouchet H, Blois-Heulin C, Lemasson A (2013) Social complexity parallels vocal complexity: a comparison of three non-human primate species. Frontiers in Psychology 4:1–15. Brown M, Corkeron P (1995) Pod characteristics of migrating humpback whales (Megaptera novaeangliae) off the East Australian coast. Behaviour 132:163–179. Burnham R, Duffus D, Mouy X (2018) Gray whale (Eschrictius robustus) call types recorded during migration off the west coast of Vancouver Island. Frontiers in Marine Science 5:1–11. Cardoso GC (2013) Using frequency ratios to study vocal communication. Animal Behaviour 85:1529–1532. Cazau D, Adam O, Aubin T, Laitman JT, Reidenberg JS (2016) A study of vocal nonlinearities in humpback whale songs: From production mechanisms to acoustic analysis. Scientific Reports 6:1–12. Cerchio S, Dahlheim M (2001) Variation in feeding vocalizations of humpback whales Megaptera novaeangliae from southeast Alaska. Bioacoustics 11:277–295. Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mammal Review 26:27–49. Clapham PJ (1993) Social organization of humpback whales on a North Atlantic feeding ground. Zoological Symposium 66:131–145. Clapham PJ (2000) The humpback whale: seasonal feeding and breeding in a baleen whale. In: Mann J, Connor R, Tyack PL, Whitehead H (eds) Cetacean Societies: Field Studies of Dolphins and Whales. University of Chicago Press, Chicago, pp 173–196. Clapham PJ (2009) Humpback Whale, Megaptera novaeangliae. In: Perrin WF, Wursig B,

108

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 582–585. Clapham PJ, Palsboll PJ, Mattila DK, Vasquez O (1992) Composition and dynamics of humpback whale competitive groups in the West Indies. Behaviour 122:182–194. Clapham PJ, Zerbini AN (2015) Are social aggregation and temporary immigration driving high rates of increase in some Southern Hemisphere humpback whale populations? Marine Biology 162:625–634. Clark CW (1982) The acoustic repertoire of the Southern right whale, a quantitative analysis. Animal Behaviour 30:1060–1071. Clark CW (1990) Acoustic behavior of mysticete whales. In: Thomas J, Kastelein R (eds) Sensory Abilities of Cetaceans. Plenum Press, New York, pp 571–583. Clutton-Brock TH, Albon SD (1979) The roaring of red deer and the evolution of honest advertisement. Behaviour 69:145–170. Corkeron PJ, Brown M, Slade RW, Bryden MM (1994) Humpback whales, megaptera novaeangliae (Cetacea: Balaenopteridae), in Hervey Bay, Queensland. Wildlife Research 21:293–305. Crawley MJ (2013) The R Book, 2nd edn. John Wiley & Sons, Ltd., Chichester (UK). Cummings WC, Thompson PO, Cook R (1968) Underwater sounds of migrating gray whales, Eschrichtius glaucus (Cope). The Journal of the Acoustical Society of America 44:1278–1281. D’Vincent CG, Nilson RM, Hanna RE (1985) Vocalizations and coordinated feeding behavior of the humpback whale in southeastern Alaska. Scientific Report of the Whales Research Institute 36:41–47. Darwin C (1872) The Expression of the Emotions in Man and Animals. John Murray, London. Delbarco-Trillo J, Burkert BA, Goodwin TE, Drea CM (2011) Night and day: The comparative study of strepsirrhine primates reveals socioecological and phylogenetic patterns in olfactory signals. Journal of Evolutionary Biology 24:82–98. Demartsev V, Bar Ziv E, Shani U, Goll Y, Koren L, Geffen E (2016) Harsh vocal elements affect counter-singing dynamics in male rock hyrax. Behavioral Ecology 27:1397–1404. Dunlop RA (2017) Potential motivational information encoded within humpback whale non-song vocal sounds. The Journal of the Acoustical Society of America 141:2204–2213. Dunlop RA, Cato DH, Noad MJ (2008) Non-song acoustic communication in migrating humpback whales (Megaptera novaeangliae). Marine Mammal Science 24:613–629. Dunlop RA, Noad MJ, Cato DH, Stokes DM (2007) The social vocalization repertoire of east

109

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Australian migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 122:2893–2905. Fedurek P, Slocombe KE, Zuberbühler K (2015) Chimpanzees communicate to two different audiences during aggressive interactions. Animal Behaviour 110:21–28. Felix F, Novillo J (2015) Structure and dynamics of humpback whales competitive groups in Ecuador. Animal Behavior and Cognition 2:56–70. Feltz CJ, Miller GE (1996) An asymptotic test for the equality of coefficients of variation from k populations. Statistics in Medicine 15:647–658. Fischer J (2006) Categorical perception in animals. In: Brown K (ed) Encyclopedia of language and linguistics, 2nd edn. Elsevier Ltd, London, pp 248–251. Fischer J (1998) Barbary macaques categorize shrill barks into two call types. Animal Behaviour 55:799–807. Fischer J, Farnworth MS, Sennhenn-Reulen H, Hammerschmidt K (2017a) Quantifying social complexity. Animal Behaviour 130:57–66. Fischer J, Wadewitz P, Hammerschmidt K (2017b) Structural variability and communicative complexity in acoustic communication. Animal Behaviour 134:229–237. Fitch WT, Hauser MD (1995) Vocal production in nonhuman primates: acoustics, physiology and functional constraints on “honest” advertising. American Journal of Primatology 37:191–219. Fitch WT, Hauser MD (2002) Unpacking " honesty ": Vertebrate vocal production and the evolution of acoustic signals. In: Simmons A, Fay RR, Popper AN (eds) Acoustic Communication. Springer, New York (NY), pp 1–44. Fitch WT, Neubauer J, Herzel H (2002) Calls out of chaos: The adaptive significance of nonlinear phenomena in mammalian vocal production. Animal Behaviour 63:407–418. Ford JKB (1989) Acoustic behaviour of resident killer whales (Orcinus orca) off Vancouver Island, British Columbia. Canadian Journal of Zoology 67:727–745. Fournet ME, Szabo A, Mellinger DK (2015) Repertoire and classification of non-song calls in Southeast Alaskan humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 137:1–10. Fournet MEH, Gabriele CM, Sharpe F, Straley JM, Szabo A (2018) Feeding calls produced by solitary humpback whales. Marine Mammal Science 34:851–865. Freeberg T (2006) Social complexity can drive vocal complexity: Group size and information in chickadee calls in Carolina chickadees. The Journal of the Acoustical Society of America 17:557–561.

110

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Freeberg TM, Dunbar RIM, Ord TJ (2012) Social complexity as a proximate and ultimate factor in communicative complexity. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1785–1801. Gadziola MA, Grimsley JMS, Faure PA, Wenstrup JJ (2012) Social vocalizations of big brown bats vary with behavioral context. PLoS ONE 7:e44550. Garland EC, Castellote M, Berchok CL (2015) Beluga whale (Delphinapterus leucas) vocalizations and call classification from the eastern Beaufort Sea population. The Journal of the Acoustical Society of America 137:3054–3067. Gerhardt HC (1992) Multiple messages in acoustic signals. Seminars in Neuroscience 4:391–400. Girola E, Noad MJ, Dunlop RA, Cato DH (2019) Source levels of humpback whales decrease with frequency suggesting an air-filled resonator is used in sound production. The Journal of the Acoustical Society of America 145:869–880. Gower JC (1971) A general coefficient of similarity and some of its properties. Biometrics 27:857– 874. Grafe TU (1995) Graded aggressive calls in the African painted reed frog Hyperolius marmoratus (Hyperoliidae). Ethology 101:67–81. Green S, Marler P (1979) The Analysis of Animal Communication. In: Social Behavior and Communication. Plenum Press, New York (NY), pp 73–158. Hammerschmidt K, Fischer J (1998) The vocal repertoire of Barbary macaques: A quantitative analysis of a graded signal system. Ethology 104:203–216. Herman EYK, Herman LM, Pack AA, Marshall G, Shepard CM, Bakhtiari M (2007) When whales collide: CRITTERCAM offers insight into the competitive behavior of humpback whales on their Hawaiian wintering grounds. Marine Technology Society 41:35–43. Indeck KL, Girola E, Torterotot M, Noad MJ, Dunlop RA (2020) Adult female-calf acoustic communication signals in migrating east Australian humpback whales. Bioacoustics Janik VM, Slater PJB (1997) Vocal learning in mammals. Advances in the Study of Behavior 26:59–99. Jurasz CM, Jurasz VP (1979) Feeding modes of the humpback whale, Megaptera noavaengliae, in southeast Alaska. Scientific Report of the Whales Research Institute 31:69–83. Kappeler PM (2019) A framework for studying social complexity. Behavioral Ecology and Sociobiology 73:1–14. Kaufman L, Rousseeuw PJ (1990) Finding groups in data: An introduction to cluster analysis. Wiley, New York.

111

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Kavanagh AS (2014) The behaviour of humpback whales: an analysis of the social and environmental context variables affecting their behaviour on migration. The University of Queensland. Keenan S, Lemasson A, Zuberbühler K (2013) Graded or discrete? A quantitative analysis of Campbell’s monkey alarm calls. Animal Behaviour 85:109–118. Kokolakis A, Smith CL, Evans CS (2010) Aerial alarm calling by male fowl (Gallus gallus) reveals subtle new mechanisms of risk management. Animal Behaviour 79:1373–1380. Koren L, Geffen E (2009) Complex call in male rock hyrax (Procavia capensis): A multi- information distributing channel. Behavioral Ecology and Sociobiology 63:581–590. Krishnamoorthy K, Lee M (2014) Improved tests for the equality of normal coefficients of variation. Computational Statistics 29:215–232. Kroodsma DE (1977) Correlates of song organization among North American wrens. The American Naturalist 111:995–1008. Lamarck JB (1809) Philosophie Zoologique. Hafner Publishing, New York. Lemasson A, Hausberger M (2011) Acoustic variability and social significance of calls in female Campbell’s monkeys (Cercopithecus campbelli campbelli). The Journal of the Acoustical Society of America 129:3341–3352. Lenth R V. (2018) emmeans: estimated marginal means, aka least-squares means. R Core Team. Maechler M, Rousseeuw PJ, Struyf A, Hubert M, Hornik K (2018) cluster: Cluster Analysis Basics and Extensions. R package version 2.0.7-1. Maestripieri D (2007) Gestural communication in three species of macaques (Macaca mulatta, M. nemestrina, M. arctoides): Use of signals in relation to dominance and social context. In: Liebal K, Muller C, Pika S (eds) Gestural Communication in Nonhuman Primates. John Benjamins, Amsterdam, pp 51–66. Mandl I, Schwitzer C, Holderied M (2019) Sahamalaza sportive lemur, Lepilemur sahamalaza, vocal communication: Call use, context and gradation. Folia Primatologica 90:336–360. Manser MB (2001) The acoustic structure of suricates’ alarm calls varies with predator type and the level of response urgency. Proceedings of the Royal Society B: Biological Sciences 268:2315– 2324. Manser MB (2013) Semantic communication in vervet monkeys and other animals. Animal Behaviour 86:491–496. Manser MB, Jansen DAWAM, Graw B, Hollén LI, Bousquet CAH, Furrer RD, le Roux A (2014) Vocal complexity in meerkats and other mongoose species. Advances in the Study of Behavior

112

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

46:281–310. Marler P (1976) Social organization, communication, and graded signals: The chimpanzee and the gorilla. In: Bateson PPG, Hinde RA (eds) Growing Points in Ethology. Cambridge University Press, Oxford (UK), pp 239–277. Marler P (1977) The structure of animal communication sounds. In: Bullock T, Evans E (eds) Recognition of complex acoustic signals. Dahlem Konferenzen, Berlin, pp 17–35. Marler P (1967) Animal Communication Signals. Science 157:769–774. Marler P (1978) Affective and symbolic meaning: Some zoosemiotic speculations. In: Sebeok T (ed) Sight, sound and sense. Indiana University Press, Bloomington, pp 113–123. Marler P (1984) Animal communication: Affect or cognition? In: Scherer K, Ekman P (eds) Approaches to emotion. Erlbaum, Hillsdale, N.J., pp 345–365. Marler P, Evans CS, Hauser MD (1992) Animal signals: Motivational, referential, or both? In: Papousek H, Jurgens U, Papousek M (eds) Nonverbal vocal communication: comparative and developmental approaches. Cambridge University Press, Cambridge, UK, pp 66–86. Marwick B, Krishnamoorthy K (2019) cvequality: Tests for the Equality of Coefficients of Variation from Multiple Groups. R Core Team. MathWorks (2018) Matlab. The MathWorks, Inc., Natick, MA. Mattila DK, Clapham PJ, Vasquez O, Bowman RS (1994) Occurrence, population composition, and habitat use of humpback whales in Samana Bay, Dominican Republic. Canadian Journal of Zoology 72:1898–1907. May-Collado LJ, Agnarsson I, Wartzok D (2007) Phylogenetic review of tonal sound production in whales in relation to sociality. BMC Evolutionary Biology 7:1–20. McComb K, Semple S (2005) Coevolution of vocal communication and sociality in primates. Biology Letters 1:381–385. McCordic JA, Root-Gutteridge H, Cusano DA, Denes SL, Parks SE (2016) Calls of North Atlantic right whales Eubalaena glacialis contain information on individual identity and age class. Endangered Species Research 30:157–169. McDonald MA, Calambokidis J, Teranishi AM, Hildebrand JA (2001) The acoustic calls of blue whales off California with gender data. The Journal of the Acoustical Society of America 109:1728–1735. McGrath N, Dunlop R, Dwyer C, Burman O, Phillips CJC (2017) Hens vary their vocal repertoire and structure when anticipating different types of reward. Animal Behaviour 130:79–96. Mobley, Jr. JR, Herman LM (1985) Transience of social affiliations among humpback whales on

113

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

the Hawaiian wintering grounds. Canadian Journal of Zoology 63:762–772. Monticelli PF, Ades C (2013) The rich acoustic repertoire of a precocious rodent, the wild cavy Cavia aperea. Bioacoustics 22:49–66. Morton ES (1977) On the occurrence and significance of motivation-structural rules in some bird and mammal sounds. The American Naturalist 111:855–869. Morton ES (1982) Grading, discreteness, redundancy, and motivation-structural rules. In: Kroodsma DE, Miller MH (eds) Acoustic communication in birds. Academic Press, New York (NY), pp 183–212. Noad MJ, Kniest E, Dunlop RA (2019) Boom to bust? Implications for the continued rapid growth of the eastern Australian humpback whale population despite recovery. Population Ecology 61:198–209. Oller DK, Griebel U (eds) (2008) Evolution of communicative flexibility: Complexity, creativity, and adaptability in human and animal communication. MIT Press, Cambridge, MA. Ord TJ, Garcia-Porta J (2012) Is sociality required for the evolution of communicative complexity? Evidence weighed against alternative hypotheses in diverse taxonomic groups. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1811–1828. Owings D, Morton ES (1998) Animal Vocal Communication: A New Approach. Cambridge University Press, Cambridge (UK). Pack AA, Herman LM, Craig AS, Spitz SS, Deakos MH (2002) Penis extrusions by humpback whales (Megaptera novaeangliae). Aquatic Mammals 28:131–146. Palsbøll PJ, Allen J, Bérubé M, et al (1997) Genetic tagging of humpback whales. Nature 388:767– 769. Parks SE, Cusano DA, Stimpert AK, Weinrich MT, Friedlaender AS, Wiley DN (2014) Evidence for acoustic communication among bottom foraging humpback whales. Scientific Reports 4:7508. Parks SE, Searby A, Célérier A, Johnson MP, Nowacek DP, Tyack PL (2011) Sound production behavior of individual North Atlantic right whales: implications for passive acoustic monitoring. Endangered Species Research 15:63–76. Parks SE, Tyack PL (2005) Sound production by North Atlantic right whales (Eubalaena glacialis) in surface active groups. The Journal of the Acoustical Society of America 117:3297–3306. Payne K, Tyack PL, Payne R (1984) Progressive changes in the songs of humpback whales (Megaptera novaeangliae): A detailed analysis of two seasons in Hawaii. In: Payne R (ed) Communication and Behavior of Whales. Westview Press, Boulder, pp 9–57.

114

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Payne RS, McVay S (1971) Songs of humpback whales. Science 173:585–597. Pearson K (1896) Mathematical contributions to the theory of evolution. III. Regression, heredity, and panmixia. Philosophical Transactions of the Royal Society of London Series A 187:253– 318. Peckre L, Kappeler PM, Fichtel C (2019) Clarifying and expanding the social complexity hypothesis for communicative complexity. Behavioral Ecology and Sociobiology 73:1–19. Pika S (2017) Unpeeling the layers of communicative complexity. Animal Behaviour 134:223–227. Pollard KA, Blumstein DT (2012) Evolving communicative complexity: Insights from rodents and beyond. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1869– 1878. R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Ramp C, Hagen W, Palsbøll P, Bérubé M, Sears R (2010) Age-related multi-year associations in female humpback whales (Megaptera novaeangliae). Behavioral Ecology and Sociobiology 64:1563–1576. Rebout N, De Marco A, Lone JC, et al (2020) Tolerant and intolerant macaques show different levels of structural complexity in their vocal communication: Vocal complexity in macaques. Proceedings of the Royal Society B: Biological Sciences 287: Reby D, McComb K (2003) Anatomical constraints generate honesty: Acoustic cues to age and weight in the roars of red deer stags. Animal Behaviour 65:519–530. Reby D, McComb K, Cargnelutti B, Darwin C, Fitch WT, Clutton-Brock T (2005) Red deer stags use formants as assessment cues during intrasexual agonistic interactions. Proceedings of the Royal Society B: Biological Sciences 272:941–947. Reichert MS, Gerhardt HC (2013) Gray tree frogs, Hyla versicolor, give lower-frequency aggressive calls in more escalated contests. Behavioral Ecology and Sociobiology 67:795–804. Ręk P, Osiejuk TS, Budka M (2011) Functionally similar acoustic signals in the corncrake (Crex crex) transmit information about different states of the sender during aggressive interactions. Hormones and Behavior 60:706–712. Rekdahl ML, Dunlop RA, Noad MJ, Goldizen AW (2013) Temporal stability and change in the social call repertoire of migrating humpback whales. The Journal of the Acoustical Society of America 133:1785–1795. Rekdahl ML, Tisch C, Cerchio S, Rosenbaum H (2017) Common nonsong social calls of humpback whales (Megaptera novaeangliae) recorded off northern Angola, southern Africa. Marine

115

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Mammal Science 33:365–375. Richardson WJ, Greene, Jr. CR, Malme CI, Thomson DH (1995) Marine Mammals and Noise. Academic Press, San Diego (CA). Root-Gutteridge H, Cusano DA, Shiu Y, Nowacek DP, Van Parijs SM, Parks SE (2018) A lifetime of changing calls: North Atlantic right whales, Eubalaena glacialis, refine call production as they age. Animal Behaviour 137:21–34. Schehka S, Esser KH, Zimmermann E (2007) Acoustical expression of arousal in conflict situations in tree shrews (Tupaia belangeri). Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 193:845–852. Schwartz JJ (1989) Graded aggressive calls of the spring peeper, Pseudacris crucifer. Herpetologica 45:172–181. Seyfarth RM, Cheney DL (2010) Production, usage, and comprehension in animal vocalizations. Brain and Language 115:92–100. Sharpe FA (2001) Social foraging of the southeast Alaskan humpback whale, Megaptera novaeangliae. PhD Thesis. Simon Fraser University, Ann Arbor, MI. Silber GK (1986) The relationship of social vocalizations to surface behavior and aggression in the Hawaiian humpback whale (Megaptera novaeangliae). Can. J. Zool. 64:2075–2080 Silk J, Cheney D, Seyfarth R (2013) A practical guide to the study of social relationships. Evolutionary Anthropology 22:213–225. Širović A, Rice A, Chou E, Hildebrand JA, Wiggins SM, Roch MA (2015) Seven years of blue and fin whale call abundance in the Southern California Bight. Endangered Species Research 28:61–76. Stimpert AK, Au WWL, Parks SE, Hurst T, Wiley DN (2011) Common humpback whale (Megaptera novaeangliae) sound types for passive acoustic monitoring. The Journal of the Acoustical Society of America 129:476–482. Stimpert AK, Lammers MO, Pack AA, Au WWL (2020) Variations in received levels on a sound and movement tag on a singing humpback whale: Implications for caller identification. The Journal of the Acoustical Society of America 147:3684–3690. Stirling I, Thomas JA (2003) Relationships between underwater vocalizations and mating systems in phocid seals. Aquatic Mammals 29:227–246. Tenaza RR (1989) Intergroup calls of male pig-tailed langurs (Simias concolor). Primates 30:199– 206. Thompson PO, Cummings WC, Ha SJ (1986) Sounds, source levels, and associated behavior of

116

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

humpback whales, Southeast Alaska. The Journal of the Acoustical Society of America 80:735–740. Tyack PL, Whitehead H (1983) Male competition in large groups of wintering humpback whales. Behaviour 83:132–154. Valsecchi E, Hale P, Corkeron P, Amos W (2002) Social structure in migrating humpback whales (Megaptera novaeangliae). Molecular Ecology 11:507–518. Videsen SKA, Bejder L, Johnson M, Madsen PT (2017) High suckling rates and acoustic crypsis of humpback whale neonates maximise potential for mother–calf energy transfer. Functional Ecology 31:1561–1573. Waas JR (1991) Do little blue penguins signal their intentions during aggressive interactions with strangers? Animal Behaviour 41:375–382. Wagner WE (1989) Graded aggressive signals in Blanchard’s cricket frog: Vocal responses to opponent proximity and size. Animal Behaviour 38:1025–1038. Walter MH, Schnitzler HU (2017) Spectral call features provide information about the aggression level of greater mouse-eared bats (Myotis myotis) during agonistic interactions. Bioacoustics 28:1–25. Waser PM (1982) The evolution of male loud calls among mangabeys and baboons. In: Snowdon CT, Brown CH, Petersen MR (eds) Primate Communication. Cambridge University Press, Cambridge (UK), pp 117–143. Watkins WA, Moore KE, Clark CW, Dahlheim M (1988) The sounds of sperm whale calves. In: Nachtigall PE, Moore PWB (eds) NATO ASI Science Series A: Life Sciences: Vol. 156. Animal Sonar. Springer, Boston (MA), pp 99–107. Watkins WA, Tyack P, Moore KE, Bird JE (1987) The 20-Hz signals of finback whales (Balaenoptera physalus). Journal of the Acoustical Society of America 82:1901–1912. Weinrich MT (1991) Stable social associations among humpback whales (Megaptera novaeangliae) in the southern Gulf of Maine. Canadian Journal of Zoology 69:3012–3019. Weissman YA, Demartsev V, Ilany A, Barocas A, Bar-Ziv E, Shnitzer I, Geffen E, Koren L (2019) Acoustic stability in hyrax snorts: Vocal tightrope-walkers or wrathful verbal assailants? Behavioral Ecology 30:223–230. Wells KD, Schwartz JJ (1984) Vocal Communication in a Neotropical Treefrog, Hyla ebraccata: Aggressive Calls. Behaviour 91:128–145. Whitehead H (1983) Structure and stability of humpback whale groups off Newfoundland. Can J Zool 61:1391–1397.

117

Chapter 4: Support for a link between the social complexity hypothesis and communicative complexity in the humpback whale

Wiley D, Ware C, Bocconcelli A, Cholewiak D, Friedlaender A, Thompson M, Weinrich M (2011) Underwater components of humpback whale bubble-net feeding behaviour. Behaviour 148:575–602. Yasukawa K (1978) Aggressive tendencies and levels of a graded display: Factor analysis of response to song playback in the redwinged blackbird (Agelaius phoeniceus). Behavioral Biology 23:446–459. Zahavi A (1982) The pattern of vocal signals and the information they convey. Behaviour 80:1–8. Zoidis AM, Smultea MA, Frankel AS, Hopkins JL, Day A, McFarland AS, Whitt AD, Fertl D (2008) Vocalizations produced by humpback whale (Megaptera novaeangliae) calves recorded in Hawaii. The Journal of the Acoustical Society of America 123:1737–1746.

118

Chapter 5

The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

5.1 Abstract

Intraspecific conflict can be costly, therefore many species engage in ritualised contests composed of several stages. Each stage is typically characterised by different levels of aggression, arousal, and physical conflict. During these different levels of ‘intensity’, animals benefit from communicating information related to fighting ability, level of aggression, and intent. This information is most frequently encoded in the acoustic features of calls, where both fixed (i.e. age, sex, and body size) and flexible (i.e. motivation or arousal) information are conveyed by using different classes of calls. Those that contain fixed information are generally considered ‘discrete’ or stereotyped, while calls that convey flexible information are more ‘graded’, existing along an acoustic continuum. During intraspecific conflict, the use of these calls, and the information they convey, likely play different roles depending on factors like intensity level. Intraspecific conflict commonly results from socially complex breeding interactions involving competition amongst males for access to breeding females. Here, we categorised the behaviour of humpback whales in competitive groups into three mutually exclusive stages: ‘low intensity’ groups, characterised by fast travel speeds and few surface behaviours; ‘moderate intensity’ groups, characterised by slower speeds and more surface behaviours; and ‘high intensity’ groups, characterised by a high frequency of aggressive behaviours and body contact. As predicted from terrestrial species, call rates, and the use of graded call types, increased linearly with intensity. Discrete calls were infrequent compared to graded calls, particularly during the highest intensity level. However some discrete call types, particularly ‘discrete snorts’ and ‘song unit social sounds’, were still used, indicating their potential importance. These results indicate that in ‘low intensity’ groups, fixed information may be more important in order to allow individuals to remotely assess their competitors and weigh the risks of engaging in conflict. In contrast, flexible information may be more important in ‘moderate’ and ‘high intensity’ groups as males continue to assess the motivation and intent (i.e. fight or flight) of competitors. Overall, the acoustic and surface behaviour of humpback whales follows a similar pattern to many terrestrial species during male-male conflict, where evolution may have selected for the use of acoustic signals to mediate intraspecific contests.

119 Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Keywords: competitive group, graded calls, humpback whale, intraspecific conflict, vocalisations

5.2 Introduction

Intraspecific conflict arises when critical resources are limited, such as food, territory, or access to breeding opportunities (Campagna 2009; Bradbury and Vehrencamp 2011; Hardy and Briffa 2013). Arguably the most common source of agonistic interaction involves the latter, particularly competition between males for access to reproductive females (Campagna 2009). Conflict can be costly, requiring high energy expenditure and possibly resulting in injury or death (Campagna 2009). In order to prevent serious injury, some species employ the strategy of ‘ritualised fighting’, where competition escalates in successive stages that provide information on the contestants’ fitness, fighting ability, or motivation (Maynard Smith and Price 1973; Maynard Smith 1974). Individuals benefit from conveying this information continuously to facilitate decisions on whether to retreat or to engage. Males that produce signals indicative of strong body condition and large body size, therefore, should be involved in fewer conflicts and injuries because inferior opponents can avoid or disengage from combative situations they will likely lose (Maynard-Smith and Harper 2003). If competitors choose to proceed and aggression escalates, signalling behaviour often reflects this escalation (Bradbury and Vehrencamp 2011; Hof and Podos 2013). Animal vocalisations (‘calls’) are frequently used in agonistic breeding interactions as a way for individuals to remotely assess their opponents and avoid conflicts they are unlikely to win (Zahavi 1982). Acoustic cues related to fitness or fighting ability are typically correlated with fixed attributes which do not change over time or change slowly, including sex, body size, age class, or individual identity (Marler 1961, 1977; Green and Marler 1979). These calls tend to be highly stereotyped (‘discrete’) in that the call structure has little variability in acoustic features between- and within-contexts in order to reliably encode these traits. For example, the discrete ‘groans’ of fallow deer (Dama dama) are displays produced during the breeding season to convey information on body size (Vannoni and McElligott 2008; Charlton and Reby 2011). As large body size in these animals is generally associated with higher rank and mating success (McElligott et al. 2001), these acoustic features can be used by potential competitors to assess the odds of successfully winning an agonistic encounter (McElligott and Hayden 1999). Red deer stags (Cervus elaphus), another species in which males defend harems, engage in ‘roaring contests’ during the breeding season (Clutton-Brock and Albon 1979; Reby et al. 2005). Males use the acoustic features of roars to remotely assess the fighting ability of their opponents. If neither male withdraws, the rate of roaring

120

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale increases. Males also move closer together to signal fitness using visual displays. If males are evenly matched, or neither backs down, the interaction may then escalate to physical combat (Clutton-Brock and Albon 1979). During conflicts, it is also beneficial to convey information regarding intent (i.e. willingness to fight, disengage, or not engage) or level of aggression (Morton 1982; Enquist 1985). This information is considered flexible, and is related to internal factors such as physiological or motivational state, or external factors such as social context (Marler 1961, 1975, 1976; Morton 1977; Marler et al. 1992; Hauser 1996; Manser 2010). Unlike discrete calls, those that contain flexible information tend to be highly variable, or ‘graded’. It is this gradation that provides listener’s with information on the subtle variations in the signaller’s internal attributes at the time of the call (Marler 1961, 1976; Morton 1977, 1982; Owings and Morton 1998; Briefer 2012). As escalation progresses beyond threats and displays, it may become increasingly more important to communicate flexible information (i.e. intent) rather than fixed information (i.e. body size or condition), especially considering that smaller animals with higher motivation are sometimes able to dominate larger opponents (Wagner 1989; Kotiaho et al. 1999; Hofmann and Schildberger 2001). For example, changes in the dominant frequency and temporal features of a graded call in cricket frogs (Acris crepitans) provides accurate information regarding the intent of an individual, or how willing it is to progress in a conflict, independent of its body size. Males that attacked an opponent produced longer duration calls with more pulses per call than those that tolerated an opponent. Additionally, males that fled an opponent significantly lowered the dominant frequency of their call, while those that attacked lowered this frequency even further (Burmeister et al. 2002). The frequency of intraspecific conflict and its intensity are partially dependent on the complexity of the social system (Campagna 2009). Species that live in dense societies, and have a polygamous mating system, have more opportunities and motives to engage in conflict, particularly during the breeding season. Most baleen whales (i.e. the filter-feeding whales) have a relatively simple social system, where they live primarily solitary lives (May-Collado et al. 2007). There is also a tendency towards mating strategies that do not include overt aggressive male competition for mates. While some baleen species do engage in agonistic or competitive behaviours associated with breeding, the level and intensity of aggression is lower in species which engage primarily in sperm competition [e.g. North Atlantic right whales, Eubalaena glacialis (Kraus and Hatch 2001; Parks 2003; Parks and Tyack 2005; Parks et al. 2007); southern right whales, Eubalaena australis (Clark 1983, 1990; Payne and Dorsey 1983); bowhead whales, Eubalaena mysticetus (Würsig et al. 1993; Rugh and Shelden 2009)]. Humpback whales, (Megaptera novaeangliae), do not utilise sperm

121

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale competition and instead engage heavily in intraspecific competition (Brownell, Jr. and Ralls 1986; Clapham 1996; Mesnick and Ralls 2009). This complex competitive behaviour results in the formation of large assemblages termed ‘competitive groups’ (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Mattila et al. 1989; Clapham et al. 1992; Clapham 1996; Pack et al. 1998; Darling and Bérubé 2001; Herman et al. 2007; Felix and Novillo 2015). Competitive groups appear to function in intrasexual competition between males for access to a female (Tyack and Whitehead 1983). There is a definitive structure to groups, with multiple male escorts centred around a nuclear female (Baker and Herman 1984; Clapham et al. 1992; Brown and Corkeron 1995). The escort that maintains the closest position to the female is the principal, or primary, escort. Primary escorts are challenged by other escorts and will defend their close proximity to the female. In large and active groups, the composition and dynamic changes often, with principal escorts and secondary escorts changing positions and roles frequently (Tyack and Whitehead 1983; Clapham et al. 1992).These groups can vary in intensity, progressing from low to high levels of aggression and arousal (Baker and Herman 1984). Usually, all males within the group behave in a similar way, therefore ‘intensity’ can be classified at a group level. Low intensity (i.e. low aggression and arousal) groups are characterised by animals which have no direct physical contact and instead rely on displays and chasing behaviour (Darling 2001). Other ‘non- contact’ agonistic display behaviours include blowing streams of bubbles, jaw clapping, and extending the throat pleats. Moderate intensity levels are indicated by more ‘intermediate’ levels of aggression, with ‘head lunging’ one of the most common behaviours observed (Baker and Herman 1984). In contrast, higher intensity competitive groups tend to move more erratically and have elevated respiration rates (Tyack and Whitehead 1983; Silber 1986; Clapham et al. 2008). They also exhibit more aggressive behaviours, which can include ‘body thrashes’, ‘tail lashes’, collisions, and minor injuries (Tyack 1981; Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Darling 2001). Behaviour in competitive groups, as in other social interactions of humpback whales, is mediated through the use of calls (Silber 1986; Cerchio and Dahlheim 2001; Dunlop et al. 2008; Parks et al. 2014). Humpbacks have the most variable, complex, and well-studied vocal repertoire of any of the large whales (Edds-Walton 1997). Calls are produced by all age and sex classes (Winn et al. 1979; Zoidis et al. 2008), and in all habitats [e.g. breeding grounds (Tyack and Whitehead 1983; Silber 1986), feeding grounds (Jurasz and Jurasz 1979; D’Vincent et al. 1985; Thompson et al. 1986; Stimpert et al. 2007, 2011; Parks et al. 2014), during migration (Dunlop et al. 2007, 2008)]. The number of calls within the repertoire is highly variable, depending on the population,

122

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale habitat area, and behavioural context (Dunlop et al. 2007; Stimpert et al. 2008; Fournet et al. 2015; Rekdahl et al. 2017). In previous chapters, we have quantitatively shown that the acoustic repertoire of humpback calls includes both discrete and graded call types. We have also shown that they use both types of calls in different ways. Discrete calls are used more predominately in the presence of escorts, situations where conveying information on body size and fitness may be more important. In contrast, graded signals appear to be more versatile and are used in all behavioural contexts. Silber (1986) reported the presence of both discrete, stereotyped sounds as well as graded, variable sounds in competitive groups. However, no analysis beyond general call rate was included. Dunlop (2017) provided evidence that multiple escort groups use calls that are indicative of higher arousal and aggression, however competitive groups in their study area were rare. To date no further research has been conducted into the potential information content of acoustic signals in competitive groups. Here, we collected dedicated acoustic and behavioural data from competitive groups of humpback whales in order to test the hypothesis that discrete and graded calls perform different functions during agonistic interactions. Following trends evident in terrestrial species like red deer, we hypothesised that discrete calls will be used more often in groups with lower levels of perceived group aggression (i.e. during agonistic displays) where it is more important to convey fixed information on body size and fitness. In contrast, graded calls will increase with the perceived level of aggression (i.e. overt aggression) in order to convey flexible information regarding intent and motivation to escalate or continue conflict. Additionally, call rates will increase linearly with group size and intensity, like during the ‘roaring contests’ of red deer. The results from this study can be ultimately used to increase our understanding of the use of vocalisations, and specifically discrete and graded calls, in agonistic situations in an animal that engages heavily in male competition during the breeding season. Further, this will provide a basis for future comparisons between baleen whale species with drastically different mating strategies, social systems, and vocal repertoires.

5.3 Methods

Behavioural and acoustic data were collected on competitive groups of humpback whales on the breeding grounds in the Great Barrier Reef (Figure 5.1). Effort concentrated around the Whitsunday Island group which has high densities of humpback whales during the breeding season (Smith et al. 2012). Data collection was conducted in four consecutive years between July and September, 2016-2019. Data were collected from 6-7m rigid-hulled inflatable boats on days with winds less than 15 knots and a sea state less than Beaufort 4. Competitive groups were located

123

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale opportunistically, and were defined as three or more adults demonstrating surface active or chasing behaviour (Tyack and Whitehead 1983; Clapham et al. 1992).

5.3.1 Behavioural data collection

After sighting a competitive group, a behavioural focal follow was initiated using continuous focal animal sampling methods (Altmann 1974). Recorded data included the number of animals in the group, the social composition of the group (i.e. general age class of group members determined by body size), a rough estimate of group speed based on the vessel speed, and the

124

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale frequency of occurrence of specified behaviours. These behaviours were selected based on previous research on humpback competitive groups in other areas (Tyack and Whitehead 1983; Baker and Herman 1984; Silber 1986; Mattila et al. 1989; Clapham et al. 1992; Clapham 1996; Pack et al. 1998; Darling 2001; Darling and Bérubé 2001; Herman et al. 2007; Felix and Novillo 2015) and formed the behavioural ethogram for the study (Table 5.1). An intensity scale was then established based on the estimated speed of the group, an estimate of the number of breaths per whale during each surfacing (given that animals performing high intensity behaviours tend to have elevated respiration rates, Helweg and Herman 1994), and the presence and frequency of behaviours deemed aggressive or highly aggressive from the ethogram. The designation of an aggressive or highly aggressive group was primarily based on whether or not there was direct physical contact, or perceived attempted physical contact, between group members. For example, a ‘tail slap’ is a behaviour in which the fluke is raised out of the water and forcibly slapped against either the surface (aggressive) or another whale (highly aggressive) (Tyack 1981; Tyack and Whitehead 1983). Each follow was then assigned an intensity level. If the behaviour of the group changed during the follow, a new intensity level was assigned to the group with no break in data collection. Humpback whales can be individually identified using distinct markings on the tail (flukes) and the shape of the dorsal fin (Katona and Whitehead 1981). Identification photographs were therefore taken of all animals in the group during the focal follow to corroborate the number of animals present and determine if individuals maintained consistent roles (e.g. one animal consistently leading, the presumed primary and secondary escorts based on behaviour and position).

Body slam ** The collision of two or more whales. Leap in which the entire, or part of, the whale body (up to the tail Breach ** stock) exits the water. The whale twists in the air and lands on its dorsal or lateral side. Includes half breaches and other variations. Bubble streaming * Blowing bubble streams underwater. The head is raised out of the water and slapped against the water’s Chin/Head Slap * surface (aggressive) or another whale (body contact - highly or ** aggressive) Open/Distended The distension of the ventral grooves or pleats to make the animal Pleats * appear bigger in size.

125

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Energetic forward motion with a forward lunge of the head, with less Head Lunge ** than 40% of the body leaving the water with an angle to the water < 45o. Jaw Clap * Forceful opening and closing of the mouth. The left or right pectoral or both pectorals are raised out of the water Pec Slap * or ** and forcibly slapped against the water’s surface (aggressive) or another whale (body contact - highly aggressive). Roll Surface or underwater roll in any direction or plane. Includes belly up. The throwing of the entire fluke and peduncle in a lateral motion out of Peduncle the water (aggressive) or at/on another whale (body contact - highly Throw/Rear body aggressive). No initial lifting from the water as in a peduncle or slap thrash * or ** tail, just a single scything motion. The fluke is raised out of the water and forcibly slapped against the Tail Slap/Lobtail * water’s surface (aggressive) or another whale (body contact - highly or ** aggressive) Tail Slash/Flick * Movement of tail in a sideways motion through water (aggressive) or or ** at/on another whale (body contact - highly aggressive) Tonal Blow accompanied by a loud vocalisation, usually low frequency. Blow/Trumpet * Underwater blow * A forceful, audible release of breath underwater

5.3.2 Acoustic data collection Acoustic recordings were collected using a Zoom H4n Pro Handy Recorder (Zoom North America, 44.1 kHz sampling rate, 16 bit) and an HTI-96-MIN dip hydrophone dropped over the side of the boat (High-Tech, Inc., flat (±1 dB) 2 Hz–30 kHz sensitivity, nominal -164 dB re 1V/μPa). The engine was shut down during recordings to minimise background noise. Although humpback whale calls are reported to have an estimated active space of up to 4 km in wind- dominated noise (Dunlop 2018a), and up 2.5 km in vessel noise (Dunlop 2018b), the acoustic environment in the study area was dominated by additional biotic noise (i.e. snapping shrimp, humpback whale song chorusing). Therefore, data were only collected when whales were within 400 m of the boat in order to ensure that all calls from the group were detected. Any whales within this distance were either involved in the competitive group or alone. As lone humpback whales rarely vocalise (Silber 1986, personal observation), it is unlikely that calls were detected from animals outside the focal group. Due to the high speeds and often erratic behaviour of competitive groups, the duration of acoustic recordings was limited to short periods (average duration 0:06:15, range one to 10 recordings per follow).

126

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

The nearly constant amount of background song from singing males in the area meant that most calls detected from competitive groups had overlapping song units of varying amplitude. This precluded any analysis of acoustic features such as peak and centre frequency, aggregate entropy, and bandwidth (as used in Chapters 2, 3, and 4). The overlapping song also prevented any automated classification techniques, like those used in the previous chapters. However, as the focus was on the use of discrete and graded calls specifically, call type assignment could be based on aural and visual assessment of the spectrograms. Calls were assigned to a call type if they matched one of the discrete call types classified in chapter 3 and assigned as graded if they did not. In addition, individual song units are sometimes used as social calls in this population (Dunlop et al. 2007, 2008; Rekdahl et al. 2013). Although song units were rare, they were detected in some of the competitive groups and classified as discrete sounds. This resulted in seven discrete call types: ‘meow’, ‘paired croaks’, ‘discrete snort’, ‘song unit’, ‘spiccato’, ‘thwop’, and ‘wop’. Spectrograms of the discrete call types and exemplar graded calls are in Figure 5.2.

127

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

5.3.3 Statistical analysis To assess the differences in the communicative behaviour within groups, call rates and the proportion of call types were used as the response variables with group intensity level as the predictor variable. Call rates were standardised for varying group sizes by dividing the number of

128

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale calls by the total number of animals. Then, rates were compared using generalised linear mixed models (GLMMs) with a negative binomial error distribution to account for zero-inflated data. A log offset was applied to the fixed effect of ‘time’ to provide rates (calls per time of deployment) rather than counts (number of calls), and group ID was included as a random effect. Models were run in R Studio (R Core Team 2018) using the package glmmTMB (Brooks et al. 2017). Next, a GLMM with a binomial error distribution for proportions was used to compare the use of discrete and graded calls between the intensity levels. The internally calculated proportion of discrete versus graded calls was the response variable, with group ID added as a random effect. Additionally, separate models were run for each of the seven discrete call types. However, due to the low effect size for group ID for some call types, no random effects were included and general linear models (GLMs) were used for the discrete call type models. Post-hoc analyses for all models were run using the emmeans package (Lenth 2018) with the ‘mvt’ method.

5.4 Results

A total of 43 competitive groups were observed and recorded in 2016 (n = 5), 2017 (n = 12), 2018 (n = 12), and 2019 (n = 14) for over 20 hours and 38 minutes of acoustic and behavioural data (Appendix 4, Table A4.1). In order to maintain a workable distance of < 400 m (see Methods), observations were broken up into multiple recordings for a total of 198 recordings. A large proportion of sightings of competitive groups (>75%) were to the north east of the major island groups, particularly between Whitsunday Island and Bait Reef (Figure 5.1; Appendix 4, Table A4.1). As per previous studies, the number of whales in competitive groups was highly variable (average 5.8 ± 2.4, range 3-16).

5.4.1 Intensity level Three intensity levels were qualitatively determined from the behavioural focal follow data (Table 5.2). Level one (‘low intensity’) was characterised by the fastest swim speeds (10+ kts) and few course changes (total acoustic recording time 05:18:57 from 67 recordings). The average group size was 5.7 animals (range 3-9) and it often appeared as if one animal was consistently leading (as identified using dorsal fin and fluke identification markings and shape), with the remaining animals following behind or ‘chasing’. Animals spent little time at the surface, typically coming up only to take a few breaths before resubmerging.

129

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Table 5.2 Intensity scale developed for competitive groups based on the behaviours outlined in the ethogram and observations of speed and breathing rates. Intensity Est. Avg. Key Behaviours Level Speed (kts) Fast travel in a steady direction, long down times, ~ 3 1 blows/surfacing; chasing behaviour, often with a consistent 10+ (Low) animal leading; limited presence of aggressive behaviours like head lunges, pec slaps, tail slaps (1-2 displays per surfacing). Slower travel in no clear direction, with shorter down times and 2 longer surface times, ~ 4 blows/surfacing; increased presence of 5-10 (Moderate) aggressive behaviours, but limited or no heightened aggressive behaviours. Even slower travel, although with similar down times and surface times as 2; increased presence of aggressive behaviours (more 3 < 5 than 10 displays per surfacing); addition of heightened aggressive (High) behaviours like direct body contact and breaches; evidence of blood on tubercles and dorsal fins. Level two (‘moderate intensity’) was the most commonly observed (total recording time 11:59:50 from 108 recordings). This level was associated with slower speeds (< 10 kts), more time spent at the surface, and a more erratic course (i.e. more course changes). The average group size was 5.4 (range 3-11) and animals used more surface-active behaviours (e.g. ‘flipper slapping’ and ‘tail slapping’) compared with intensity level one. Many of these behaviours were identified as ‘aggressive’, but not ‘highly aggressive’, because of the lack of direct body contact (Table 5.1, Table 5.2). Level three (‘high intensity’) groups were relatively rare (total recording time 03:19:16 from 23 recordings). While this level was similar in speed and time spent at the surface to intensity level two, it was characterised by a higher frequency of surface activity. The average group size for level three was 8.9 (range 3-16) and animals performed ‘high aggressive’ behaviours such as ‘breaching’, ‘tail slashing’, and ‘chin’ or ‘head slaps’ on other group members (Table 5.1, Table 5.2). Animals during these follows would periodically surface with blood on their dorsal fins, indicative of this direct physical contact. A calf was only ever observed in a group with this intensity level for one recording (about 11 minutes).

5.4.2 Acoustic behaviour A total of 5,493 calls were detected over the study period: 970 in intensity level one, 3,546 in intensity level two, and 977 in intensity level three. As expected, call rates (per whale per hour)

130

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale increased linearly with intensity level (Figure 5.3). The lowest call rates were in intensity level one (4.5 ± 1.5), increasing in intensity level two (11.5 ± 3.5), and highest in intensity level three (20.0 ± 8.1). Though there was an increase in call rate per whale between intensity two and three, these results were not significantly different, possibly due to differences in sample size (estimate -0.56 ± 0.33, t ratio = -1.708, p = 0.2022). The significant difference between intensity level one and two (estimate -0.93 ± 0.27, t ratio = -3.44, p = 0.0030), and one and three (estimate -1.49 ± 0.42, t ratio = -3.584, p = 0.0020) suggests there was a significant increase in the need for individuals to communicate motivation or intent in more moderate and high intensity levels.

Figure 5.3 Average call rate (calls per whale per hour) within each intensity level.

Of the total calls, 1,136 were classified as discrete calls and 4,357 as graded calls. As with call rate, the proportion of graded calls used with the group’s repertoire significantly increased with intensity level (Figure 5.4, Table 5.3). The highest proportion of graded calls was from groups in intensity level three (0.93 ± 0.02), while the highest proportion of discrete calls was recorded from groups in intensity level one (0.40 ± 0.07). This indicates that discrete calls, and the information they convey, may be more useful in lower intensity competitive groups where non-contact threats and displays are more common than overt contact aggression. In contrast, here it seems that graded calls are used more in escalated contests where contact aggression between whales is common.

131

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Figure 5.4 Proportions of discrete and graded call types detected in the three intensity levels.

Table 5.3 Results of the generalised linear mixed models with the probability of detecting discrete and graded calls in each intensity level in the first three columns, and the pairwise contrasts (comparisons) in the last three columns. An asterisk indicates statistical significance at the p < 0.05 level. Call Type Low Moderate High Low-Mod Low-High Mod-High (prob ± SE) (prob ± SE) (prob ± SE) (est. ± SE) (est. ± SE) (est. ± SE) 1.22±0.21 2.21±0.30 0.99±0.21 Discrete 0.40±0.07 0.17±0.04 0.07±0.02 z ratio=5.90 z ratio=7.36 z ratio=4.60 Graded 0.60±0.07 0.83±0.04 0.93±0.02 p < 0.0001* p < 0.0001* p < 0.0001* All seven of the analysed discrete call types were detected in competitive groups, however the use of these calls differed between the three intensity levels (Figure 5.5, Table 5.4). Low intensity groups used a significantly higher proportion of ‘paired croaks’ (0.19 ± 0.02), which are repeated pairs of low-frequency pulses heard only in groups containing one or more escorts (thesis

132

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Chapters 2 and 4). These sequences are correlated with an increase in arousal and social complexity, although their exact function is unknown (thesis Chapters 2 and 4). ‘Wops’ were also detected more in low intensity groups (0.56 ± 0.03). However ‘wops’ comprised a large proportion of the repertoire of discrete calls in all three intensity levels (≥ 0.40). This is not surprising given this call type is one of the most common calls in the repertoire (Dunlop et al. 2007, 2008). High intensity groups only used four of the seven detected discrete call types, with no ‘paired croaks’, ‘spiccatos’, or ‘thwops’ detected. They used a significantly higher proportion of ‘discrete snorts’ (0.14 ± 0.02) compared to low intensity groups. This call type is a short, low- frequency sound heard only in groups containing one or more escort (thesis Chapter 4), compared to more graded ‘snorts’, which are commonly heard across a range of contexts (Dunlop et al. 2008; Indeck et al. 2020, thesis Chapters 2 and 4). High intensity groups also used a significantly higher proportion of ‘song unit social sounds’ (0.18 ± 0.03). These results suggest that, although graded calls are used more often as conflict escalates, ‘discrete snorts’ and ‘song unit social sounds’ may convey fixed information that is particularly important during moderate and high intensity interactions.

133

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Figure 5.5 Proportions of discrete call types detected in the three intensity levels.

Table 5.4 Results of the generalised linear models, with the model calculated proportion of each call type in each intensity level in the first three columns, and the pairwise comparisons in the last three columns. A negative estimate and z ratio indicate the estimate is lower in the first intensity level listed, and an asterisk indicates statistical significance at the p < 0.05 level. Call Low-Mod Low-High Mod-High Low Moderate High Type (est. ± SE) (est. ± SE) (est. ± SE) 0.37±0.38 1.44±0.77 1.07±0.75 0.04± 0.03± 0.01± Meow z ratio=0.986 z ratio=1.87 z ratio=1.43 0.01 0.006 0.007 p=0.5728 p=0.1385 p=0.3126 1.53±0.24 Paired 0.19± 0.05± NA z ratio=6.43 NA NA croaks 0.02 0.01 p<0.0001* -1.11±0.19 -1.44±0.23 -0.33±0.17 Discrete 0.14± 0.32± 0.40± z ratio=-5.79 z ratio=-6.38 z ratio=-1.96 snort 0.02 0.02 0.03 p<0.0001* p<0.0001* p=0.1189

134

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

-1.92±0.37 -2.03±0.40 -0.11±0.21 Song 0.03± 0.16± 0.18± z ratio=-5.12 z ratio=-5.02 z ratio=-0.53 unit 0.01 0.01 0.03 p<0.0001* p<0.0001* p=0.8526 1.24±0.65 0.02± 0.01± Spiccato NA z ratio=1.91 NA NA 0.01 0.003 p=0.1102 -0.19±0.44=5 0.02± 0.03± Thwop NA z ratio=-0.42 NA NA 0.01 0.01 p=0.8959 0.64±0.14 0.60±0.19 -0.04±0.17 0.56± 0.40± 0.41± Wop z ratio=1.49 z ratio=3.21 z ratio=-0.25 0.03 0.02 0.04 p<0.0001* p=0.0037* p=0.9664

5.5 Discussion

In terrestrial animals, intraspecific conflict often progresses from low-intensity threats and displays to escalated conflicts. Each successive stage of the conflict provides an opportunity for animals to assess their opponents and make decisions on whether to engage, often using acoustic signals. Here, we have shown that humpback whales behave in a similar way during competitive behaviour. As the intensity level of these competitive interactions increase, from low-level non- contact displays, to high-level overt aggression, the calling behaviour of these whales also changes. Lower intensity groups used fewer overtly aggressive behaviours, had lower call rates per whale, but used a higher proportion of discrete calls such as ‘wops’ and ‘paired croaks’ within their repertoire. In other species, discrete call types typically contain fixed information related to features like body size which may be important information to convey to opponents to avoid costly conflict. In contrast, higher intensity groups used more aggressive behaviours, had significantly higher call rates per animal, and used significantly more graded calls within their repertoire. Graded calls may provide more information on the intent of the caller, or their willingness to engage or continue to conflict. While no conclusion can be made about the intent of the caller, these results show that in humpback whales, the use of graded and discrete calls and call rates are clearly correlated with the level of intensity. These findings provide a basis for investigating what information content is actively conveyed in these situations. Ultimately this will increase our understanding of the use of vocal calls during male-male competition in mammals. In this study, low intensity competitive groups (intensity level one) were described as exhibiting little surface activity, and few overtly aggressive behaviours. Coupled with the low call rates and higher relative use of discrete calls, we propose that communication during low intensity

135

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale competitive groups potentially functions as a way for males to assess each other remotely without resorting to physical contact. This is further supported by the relatively fast speeds of these groups, which would make visual displays less functional. There is ample evidence in terrestrial species to suggest that discrete calls function to convey information to potential competitors (Zahavi 1982), such as sex, age class, and body size (see introduction). It is unknown whether the discrete calls of humpbacks contain information that is related to any specific features, however this is seen in the discrete ‘upcall’ of another baleen whale, the North Atlantic right whale. The ‘upcall’ is the most common call type for this species, thought to be used as a contact call, and contains information on age class, sex, and individual identity (Clark 1990; McCordic et al. 2016). Similarly, the ‘wop’, commonly detected here, comprised the largest proportion of discrete calls across all three intensity levels (0.56 of discrete calls in level one, 0.40 in level two, and 0.41 in level three). It is also a call type that is used across many behavioural contexts (Dunlop et al. 2008, thesis Chapter 4) and populations (Dunlop et al. 2007; Stimpert et al. 2011; Fournet et al. 2015; Rekdahl et al. 2017), and is proposed to potentially function as an intra- and/or inter-group contact call (Dunlop et al. 2008). It is therefore a good candidate for a call type which could contain fixed information, including individual identity, especially if it functions to maintain contact between individuals. Although the proportion of discrete calls decreased significantly as intensity increased (0.40, 0.17, and 0.07), discrete call types still appear to play a role in the higher intensity groups. ‘Discrete snorts’ and ‘song unit social sounds’, in particular, comprised a large proportion of the repertoire in intensity level three (0.40 and 0.18). With the escalated contests likely occurring in these high intensity groups, these calls may provide information that can be used by competitors to decide whether or not to continue the conflict. ‘Discrete snorts’ are low frequency, unmodulated, and short in duration (thesis Chapter 3). This call type was found to be produced only in female-calf and single escort groups and competitive groups (thesis Chapter 4). Despite being a relatively discrete and stereotyped call, especially compared to more graded ‘snorts’ described elsewhere (Dunlop et al. 2008, Dunlop 2017, Indeck et al. 2020, thesis Chapters 3 and 4), it had significantly more variability in duration, peak frequency, and bandwidth in single escort groups than the more social behavioural states (thesis Chapter 4). Combined with the results presented here, it is possible that ‘discrete snorts’ contain multiple types of information that is relevant to dominance and/or body condition. For example, the ‘snorts’ of male rock hyraxes (Procavia capensis) and the ‘rattles’ of male barn swallows (Hirundo rustica) are also low frequency pulsed sounds that are thought to function in conveying information regarding dominance status and/or testosterone levels, primarily to other males (Galeotti et al. 1997; Koren and Geffen 2009; Demartsev et al. 2016; Weissman et al.

136

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

2019). For both species, these sound types are also part of song, aimed primarily at females, and so these calls likely serve multiple functions. The ‘discrete snorts’ of humpback whales could function in a similar way, conveying information related to temporary dominance status or current hormonal levels. ‘Song unit social sounds’ also comprised a large proportion of the calls in high intensity groups compared to low and moderate intensity groups. These calls are detected most often in lone males and groups of multiple animals, and are proposed to be a signal used only by males (Dunlop et al. 2008). Combined with one of the presumed functions of song to be reproductive advertisement, this suggests that, as with the rock hyrax and barn swallow, certain aspects of song likely contain information that is aimed primarily at females, while others contain information more relevant to other males (Tyack 1981; Darling and Bérubé 2001; Mercado et al. 2005; Herman 2017; Murray et al. 2018). As competition escalates into higher intensity and overt aggression, like in intensity level three groups, using certain ‘song unit social sounds’ may be a way to broadcast that information to other males in the immediate group. The fact that this call type has a lower source level when used in social contexts than when produced in song (Dunlop et al. 2013), and is produced at a higher rate when males join groups of multiple adults, provides further evidence that they are an intragroup signal aimed at other males (Dunlop and Noad 2016). Moderate and high intensity groups (intensity level two and three) were significantly slower than intensity level one groups, and displayed more surface active behaviour. This included an increased number of aggressive behaviours like ‘tail slashes’, ‘breaches’, and direct body contact. Open wounds and blood were also sometimes observed, further indications of body contact that occurred subsurface. There was a linear increase in the number of calls per hour and proportion of graded calls from intensity level one. Here, we propose that whales are progressing to using more conspicuous displays, both visual and acoustic, and using a larger proportion of graded calls, signals that are potentially more indicative of motivation or intent (Morton 1982; Enquist 1985). As per Silber (1986), these vocalisations may be used in conjunction with visual threats to convey aggression level more effectively than using only one signal modality (Smith 1977). This is also seen in some seals and sea lions during agonistic interactions, where graded calls that convey level of threat and/or intensity are associated with visual displays (Insley et al. 2003). Unfortunately, in the present study, the continuous background song precluded automated measurement of any acoustic features of the calls, or any quantitative classification of call types. Future studies should therefore attempt to do this in order to determine if there is a difference in any acoustic features between the three intensity levels, and thus which specific features convey

137

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale important information. It would also be beneficial in order to identify potential call types that, although not detected in previous chapters, are relatively discrete. Incorporating data from other breeding grounds where population sizes are smaller (i.e. Tonga or New Caledonia) may help as there is less background chorusing from singing males (J. Allen, pers. comm.), potentially as a result of lower numbers on these breeding grounds (Constantine et al. 2012). In the current study, we were also limited to making assumptions about the behaviour of the animals based on surface observations. This might not be an accurate representation of their behaviour below, especially as humpbacks in competitive groups are known to use the entire water column (Herman et al. 2007). However, we carefully determined intensity levels that were qualitatively discrete and mutually exclusive, providing what is likely a conservative view of the variety of intensity levels observed in competitive groups. Incorporating underwater video will help to validate the correlation between surface and underwater behaviour, as well as their relationship with calling behaviour. Overall, we have provided evidence that humpback whales follow similar trends to other species which engage heavily in male competition during the breeding season. As theories regarding male-male competition predict, humpback whale competitive groups progressed from low intensity displays to higher intensity contests, but with escalated contests being relatively uncommon. Further, we have shown that humpback whales use acoustic signals in concordance with visual displays during this progression. Calls in the low intensity levels likely function to convey more fixed information, while calls in the higher intensity levels likely function to convey flexible information on motivation and arousal. To our knowledge, this study represents the first analysis on the potential function of discrete and graded calls in a species which engages in male competition during the breeding season. Future research can build from this to make comparisons between animals with different mating strategies and degrees of social complexity, and perhaps shed light on the evolution of communicative behaviour during intraspecific conflict.

5.6 References

Altmann J (1974) Observational study of behavior: sampling methods. Behaviour 49:227–267. Baker CS, Herman LM (1984) Aggressive behavior between humpback whales (Megaptera novaeangliae) wintering in Hawaiian waters. Canadian Journal of Zoology 62:1922–1937. Bradbury JW, Vehrencamp SL (2011) Principles of Animal Communication, 2nd edn. Sinauer Associates, Inc., Sunderland, MA. Briefer EF (2012) Vocal expression of emotions in mammals: Mechanisms of production and evidence. Journal of Zoology 288:1–20.

138

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Brooks ME, Kristensen K, van Bentham K, Magnusson A, Berg CW, Nielsen A, Skaug HJ, Maechler M, Bolker B (2017) Modeling zero-inflated count data with glmmTMB. Ecological Modelling 1–14. Brown M, Corkeron P (1995) Pod characteristics of migrating humpback whales (Megaptera novaeangliae) off the East Australian coast. Behaviour 132:163–179. Brownell, Jr. RL, Ralls K (1986) Potential for sperm competition in baleen whales. Report of the International Whaling Commission Special Is:97–112. Burmeister SS, Ophir AG, Ryan MJ, Wilczynski W (2002) Information transfer during cricket frog contests. Animal Behaviour 64:715–725. Campagna C (2009) Aggressive Behavior, Intraspecific. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 18–24. Cerchio S, Dahlheim M (2001) Variation in feeding vocalizations of humpback whales Megaptera novaeangliae from southeast Alaska. Bioacoustics 11:277–295. Charlton BD, Reby D (2011) Context-related acoustic variation in male fallow deer (Dama dama) groans. PLoS ONE 6:e21066. Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mammal Review 26:27–49. Clapham PJ, Mattila DK, Palsbøll PJ (2008) High-latitude-area composition of humpback whale competitive groups in Samana Bay: Further evidence for panmixis in the North Atlantic population. Canadian Journal of Zoology 71:1065–1066. Clapham PJ, Palsboll PJ, Mattila DK, Vasquez O (1992) Composition and dynamics of humpback whale competitive groups in the West Indies. Behaviour 122:182–194. Clark CW (1983) Acoustic communication and behavior of the southern right whale (Eubaleana australis). In: Payne RS (ed) Communication and Behavior of Whales. Westview Press, Boulder, CO, pp 163–198. Clark CW (1990) Acoustic behavior of mysticete whales. In: Thomas J, Kastelein R (eds) Sensory Abilities of Cetaceans. Plenum Press, New York, pp 571–583. Clutton-Brock TH, Albon SD (1979) The roaring of red deer and the evolution of honest advertisement. Behaviour 69:145–170. Constantine R, Jackson JA, Steel D, et al (2012) Abundance of humpback whales in Oceania using photo-identification and microsatellite genotyping. Marine Ecology Progress Series 453:249– 261. D’Vincent CG, Nilson RM, Hanna RE (1985) Vocalizations and coordinated feeding behavior of

139

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

the humpback whale in southeastern Alaska. Scientific Report of the Whales Research Institute 36:41–47. Darling JD (2001) Characterization of Behavior of Humpback Whales in Hawaiian Waters.Honolulu, HI. Darling JD, Bérubé M (2001) Interactions of singing humpback whales with other males. Marine Mammal Science 17:570–584. Demartsev V, Bar Ziv E, Shani U, Goll Y, Koren L, Geffen E (2016) Harsh vocal elements affect counter-singing dynamics in male rock hyrax. Behavioral Ecology 27:1397–1404. Dunlop RA (2018a) The communication space of humpback whale social sounds in wind- dominated noise. The Journal of the Acoustical Society of America 144:540–551. Dunlop RA (2018b) The communication space of humpback whale social sounds in vessel noise. Proceedings of Meetings on Acoustics 35:1–16. Dunlop RA, Cato DH, Noad MJ (2008) Non-song acoustic communication in migrating humpback whales (Megaptera novaeangliae). Marine Mammal Science 24:613–629. Dunlop RA, Cato DH, Noad MJ, Stokes DM (2013) Source levels of social sounds in migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 134:706–714. Dunlop RA, Noad MJ (2016) The “risky” business of singing: Tactical use of song during joining by male humpback whales. Behavioral Ecology and Sociobiology 70:2149–2160. Dunlop RA, Noad MJ, Cato DH, Stokes DM (2007) The social vocalization repertoire of east Australian migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 122:2893–2905. Edds-Walton PL (1997) Acoustic communication signals of mysticete whales. Bioacoustics 8:47– 60. Enquist M (1985) Communication during aggressive interactions with particular reference to variation in choice of behaviour. Animal Behaviour 33:1152–1161. Felix F, Novillo J (2015) Structure and dynamics of humpback whales competitive groups in Ecuador. Animal Behavior and Cognition 2:56–70. Fournet ME, Szabo A, Mellinger DK (2015) Repertoire and classification of non-song calls in Southeast Alaskan humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 137:1–10. Galeotti P, Saino N, Sacchi R, MØller AP (1997) Song correlates with social context, testosterone and body condition in male barn swallows. Animal Behaviour 53:687–700.

140

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Green S, Marler P (1979) The Analysis of Animal Communication. In: Social Behavior and Communication. Plenum Press, New York (NY), pp 73–158. Hardy ICW, Briffa M (2013) Animal Contests. Cambridge University Press, Cambridge, UK. Hauser MD (1996) The evolution of communication. MIT Press, Cambridge, MA. Helweg DA, Herman LM (1994) Diurnal patterns of behaviour and group membership of humpback whales (Megaptera novaeangliae) wintering in Hawaiian waters. Ethology 98:298– 311. Herman EYK, Herman LM, Pack AA, Marshall G, Shepard CM, Bakhtiari M (2007) When whales collide: CRITTERCAM offers insight into the competitive behavior of humpback whales on their Hawaiian wintering grounds. Marine Technology Society 41:35–43. Herman LM (2017) The multiple functions of male song within the humpback whale (Megaptera novaeangliae) mating system: review, evaluation, and synthesis. Biological Reviews 92:1795– 1818. Hof D, Podos J (2013) Escalation of aggressive vocal signals: A sequential playback study. Proceedings of the Royal Society B: Biological Sciences 280:1–8. Hofmann HA, Schildberger K (2001) Assessment of strength and willingness to fight during aggressive encounters in crickets. Animal Behaviour 62:337–348. Indeck KL, Girola E, Torterotot M, Noad MJ, Dunlop RA (2020) Adult female-calf acoustic communication signals in migrating east Australian humpback whales. Bioacoustics Insley SJ, Phillips A V., Charrier I (2003) A review of social recognition in pinnipeds. Aquatic Mammals 29:181–201. Jurasz CM, Jurasz VP (1979) Feeding modes of the humpback whale, Megaptera noavaengliae, in southeast Alaska. Scientific Report of the Whales Research Institute 31:69–83. Katona SK, Whitehead HP (1981) Identifying humpback whales using their natural markings. Polar Record 20:439–444. Koren L, Geffen E (2009) Complex call in male rock hyrax (Procavia capensis): A multi- information distributing channel. Behavioral Ecology and Sociobiology 63:581–590. Kotiaho JS, Alatalo R V., Mappes J, Parri S (1999) Honesty of agonistic signalling and effects of size and motivation asymmetry in contests. Acta Ethologica 2:13–21. Kraus SD, Hatch JJ (2001) Mating strategies in the North Atlantic right whale. Journal of Cetacean Research and Management Special Is:237–244. Lenth R V. (2018) emmeans: estimated marginal means, aka least-squares means. R Core Team. Manser MB (2010) The generation of functionally referential and motivational vocal signals in

141

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

mammals. In: Brudzynski SM (ed) Handbook of Mammalian Vocalization - an integrative neuroscience approach. Academic Press, London (UK), pp 477–486. Marler P (1977) The structure of animal communication sounds. In: Bullock T, Evans E (eds) Recognition of complex acoustic signals. Dahlem Konferenzen, Berlin, pp 17–35. Marler P (1961) The logical analysis of animal communication. Journal of Theoretical Biology 1:295–317. Marler P (1976) Social organization, communication, and graded signals: The chimpanzee and the gorilla. In: Bateson PPG, Hinde RA (eds) Growing Points in Ethology. Cambridge University Press, Oxford (UK), pp 239–277. Marler P (1975) On the origin of speech from animal sounds. In: Kavanagh JF, Cutting J (eds) The Role of Speech in Language. MIT Press, Cambridge (MA), pp 11–37. Marler P, Evans CS, Hauser MD (1992) Animal signals: Motivational, referential, or both? In: Papousek H, Jurgens U, Papousek M (eds) Nonverbal vocal communication: comparative and developmental approaches. Cambridge University Press, Cambridge, UK, pp 66–86. Mattila DK, Clapham PJ, Katona SK, Stone GS (1989) Population composition of humpback whales, Megaptera novaeangliae, on Silver Bank, 1984. Canadian Journal of Zoology 67:281– 285. May-Collado LJ, Agnarsson I, Wartzok D (2007) Phylogenetic review of tonal sound production in whales in relation to sociality. BMC Evolutionary Biology 7:1–20. Maynard-Smith J, Harper D (2003) Animal Signals. Oxford University Press, Oxford, UK. Maynard Smith J (1974) The theory of games and the evolution of animal conflicts. Journal of Theoretical Biology 47:209–221. Maynard Smith J, Price GR (1973) The logic of animal conflict. Nature 246:15–18. McCordic JA, Root-Gutteridge H, Cusano DA, Denes SL, Parks SE (2016) Calls of North Atlantic right whales Eubalaena glacialis contain information on individual identity and age class. Endangered Species Research 30:157–169. McElligott AG, Gammell MP, Harty HC, Paini DR, Murphy DT, Walsh JT, Hayden TJ (2001) Sexual size dimorphism in fallow deer (Dama dama): Do larger, heavier males gain greater mating success? Behavioral Ecology and Sociobiology 49:266–272. McElligott AG, Hayden TJ (1999) Context-related vocalization rates of fallow bucks, Dama dama. Animal Behaviour 58:1095–1104. Mercado E, Herman LM, Pack AA (2005) Stereotypical sound patterns in humpback whale songs: Usage and function. Aquatic Mammals 29:37–52.

142

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Mesnick SL, Ralls K (2009) Mating Systems. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 712–719. Morton ES (1977) On the occurrence and significance of motivation-structural rules in some bird and mammal sounds. The American Naturalist 111:855–869. Morton ES (1982) Grading, discreteness, redundancy, and motivation-structural rules. In: Kroodsma DE, Miller MH (eds) Acoustic communication in birds. Academic Press, New York (NY), pp 183–212. Murray A, Dunlop RA, Noad MJ, Goldizen AW (2018) Stereotypic and complex phrase types provide structural evidence for a multi-message display in humpback whales ( Megaptera novaeangliae ). The Journal of the Acoustical Society of America 143:980–994. Owings D, Morton ES (1998) Animal Vocal Communication: A New Approach. Cambridge University Press, Cambridge (UK). Pack AA, Salden DR, Ferrari MJ, Glockner-Ferrari DA, Herman LM, Stubbs HA, Straley JM (1998) Male humpback whale dies in competitive group. Marine Mammal Science 14:861– 873. Parks SE (2003) Response of North Atlantic right whales (Eubalaena Glacialis) to playback of calls recorded from surface active groups in both the North and South Atlantic. Marine Mammal Science 19:563–580. Parks SE, Brown MW, Conger LA, Hamilton PK, Knowlton AR, Kraus SD, Slay CK, Tyack PL (2007) Occurrence, composition, and potential functions of North Atlantic right whale (Eubalaena glacialis) surface active groups. Marine Mammal Science 23:868–887. Parks SE, Cusano DA, Stimpert AK, Weinrich MT, Friedlaender AS, Wiley DN (2014) Evidence for acoustic communication among bottom foraging humpback whales. Scientific Reports 4:7508. Parks SE, Tyack PL (2005) Sound production by North Atlantic right whales (Eubalaena glacialis) in surface active groups. The Journal of the Acoustical Society of America 117:3297–3306. Payne RS, Dorsey EM (1983) Sexual dimorphism and aggressive use of callosities in right whales (Eubalaena australis). In: Payne RS (ed) Communication and Behavior of Whales. Westview Press, Boulder, CO, pp 295–329. R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Reby D, McComb K, Cargnelutti B, Darwin C, Fitch WT, Clutton-Brock T (2005) Red deer stags use formants as assessment cues during intrasexual agonistic interactions. Proceedings of the

143

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Royal Society B: Biological Sciences 272:941–947. Rekdahl ML, Dunlop RA, Noad MJ, Goldizen AW (2013) Temporal stability and change in the social call repertoire of migrating humpback whales. The Journal of the Acoustical Society of America 133:1785–1795. Rekdahl ML, Tisch C, Cerchio S, Rosenbaum H (2017) Common nonsong social calls of humpback whales (Megaptera novaeangliae) recorded off northern Angola, southern Africa. Marine Mammal Science 33:365–375. Rugh DJ, Shelden KEW (2009) Bowhead whale: Balaena mysticetus. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 131–133. Silber GK (1986) The relationship of social vocalizations to surface behavior and aggression in the Hawaiian humpback whale (Megaptera novaeangliae). Can. J. Zool. 64:2075–2080 Smith JN, Grantham HS, Gales N, Double MC, Noad MJ, Paton D (2012) Identification of humpback whale breeding and calving habitat in the Great Barrier Reef. Marine Ecology Progress Series 447:259–272. Smith WJ (1977) The behavior of communicating, an ethological approach. Harvard University Press, Cambridge, MA. Stimpert AK, Au WWL, Parks SE, Hurst T, Wiley DN (2011) Common humpback whale (Megaptera novaeangliae) sound types for passive acoustic monitoring. The Journal of the Acoustical Society of America 129:476–482. Stimpert AK, Au WWL, Wiley DN, Mattila DK (2008) Contextual sound production by tagged humpback whales (Megaptera novaeangliae) on a feeding and breeding ground. In: Program Abstracts from the 156th Meeting of the Acoustical Society of America. Miami, FL, p 2484. Stimpert AK, Wiley DN, Au WWL, Johnson MP, Arsenault R (2007) ‘ Megapclicks ’: acoustic click trains and buzzes produced during night-time foraging of humpback whales (Megaptera novaeangliae). Biology Letters 1–4. Thompson PO, Cummings WC, Ha SJ (1986) Sounds, source levels, and associated behavior of humpback whales, Southeast Alaska. The Journal of the Acoustical Society of America 80:735–740. Tyack PL (1981) Interactions between singing Hawaiian humpback whales and conspecifics nearby. Behavioral Ecology 8:105–116. Tyack PL, Whitehead H (1983) Male competition in large groups of wintering humpback whales. Behaviour 83:132–154.

144

Chapter 5: The differential use of discrete and graded calls during intraspecific conflict in the humpback whale

Vannoni E, McElligott AG (2008) Low frequency groans indicate larger and more dominant fallow deer (Dama dama) males. PLoS ONE 3:e3113. Wagner WE (1989) Graded aggressive signals in Blanchard’s cricket frog: Vocal responses to opponent proximity and size. Animal Behaviour 38:1025–1038. Weissman YA, Demartsev V, Ilany A, Barocas A, Bar-Ziv E, Shnitzer I, Geffen E, Koren L (2019) Acoustic stability in hyrax snorts: Vocal tightrope-walkers or wrathful verbal assailants? Behavioral Ecology 30:223–230. Winn HE, Beamish P, Perkins PJ (1979) Sounds of two entrapped humpback whales (Megaptera novaeangliae) in Newfoundland. Marine Biology 55:151–155. Würsig B, Guerrero J, Silber GK (1993) Social and sexual behavior of bowhead whales in fall in the Western Arctic: A re‐examination of seasonal trends. Marine Mammal Science 9:103–115. Zahavi A (1982) The pattern of vocal signals and the information they convey. Behaviour 80:1–8. Zoidis AM, Smultea MA, Frankel AS, Hopkins JL, Day A, McFarland AS, Whitt AD, Fertl D (2008) Vocalizations produced by humpback whale (Megaptera novaeangliae) calves recorded in Hawaii. The Journal of the Acoustical Society of America 123:1737–1746.

145

Chapter 6

Discussion

6.1 Introduction

Social animals typically have more frequent and diverse interactions compared with less social species, which may interact only rarely with conspecifics (Freeberg et al. 2012; Bergman and Beehner 2015; Fischer et al. 2017a; Allen 2019; Kappeler et al. 2019). This difference in social complexity is traditionally defined using metrics that focus on stable groups, however species which live in more fluid societies are naturally more difficult to assess. In these animals, it may be more appropriate to consider specific social interactions, such as cooperative and competitive behaviour. Species which engage in group competition are rare. These conflict interactions could be considered complex due to a greater number of relationships (e.g. larger group size), frequent changes in group dynamics (e.g. splitting and joining of males), and the diversity of potential relationships (e.g. multiple levels of temporary dominance hierarchies) (Freeberg et al. 2012; Allen 2019; Kappeler 2019; Peckre et al. 2019). Social interactions in marine mammals are often mediated using acoustic signals, which can provide potential information to conspecifics regarding the signaller’s internal motivational state and arousal level (Green and Marler 1979; Gerhardt 1992). This may be particularly useful for species that utilise male competition as a mating strategy. Acoustic signals in this context could be used as a way for a receiver to remotely assess the internal state of a competitor, including arousal level, motivational state, willingness to engage or continue engaging in overt physical conflict, dominance status, and/or reproductive status. As intraspecific conflict can be physically costly (Campagna 2009), these species can benefit from continuously conveying these attributes during competition (Zahavi 1982). By standard metrics, the social structure of the humpback whale (Megaptera novaeangliae) is considered to be relatively simple. Humpback whale groups do not have permanent members and there is limited evidence for stable associations between conspecifics beyond the mother and calf pair, which only lasts for around eight months (Whitehead 1983; Baker and Herman 1984; Mobley, Jr. and Herman 1985; Clapham et al. 1992; Clapham 1993, 1996, 2000; Mattila et al. 1994; Valsecchi et al. 2002; Ramp et al. 2010). In addition, humpback whale females raise offspring alone, with no evidence for kin recognition past weaning (Clapham 1993; Pomilla and Rosenbaum 2006). However, humpback whales do engage in highly social behaviour and interactions,

146 Chapter 6: Discussion particularly during the feeding and breeding seasons. These involve cooperation, potentially the formation of coalitions, and in at least some populations, repeated interactions with individuals that may span across years. According to the ‘social complexity hypothesis for communicative complexity’ (SCHCC), the relatively simple social system of humpback whales should correlate with a relatively simple communication system (Freeberg et al. 2012). However, humpback whales, especially whilst undertaking the aforementioned social interactions, have some of the most complex communication signals found in any non-human animal. There are several metrics and criteria used to study communicative complexity. Historically, the most prevalent was the number of call types in an acoustic repertoire (Freeberg et al. 2012; Fischer et al. 2017b). The acoustic repertoire of the humpback whale is larger than other baleen whales, with many other species producing only 2-8 well described call types [e.g. blue whale, Balaenoptera musculus (McDonald et al. 2001); fin whale, Balaenoptera physalus (Watkins et al. 1987; Širović et al. 2015); southern right whale, Eubalaena australis (Clark 1982); North Atlantic right whale, Eubalaena glacialis (Parks and Tyack 2005; Parks et al. 2011); bowhead whale, Eubalaena mysticetes (Wursig et al. 1985; Clark 1990); and gray whale, Eschrichtius robustus (Cummings et al. 1968; Burnham et al. 2018)]. Currently, other criteria are also used, including context-specificity of calls and the presence and proportion of acoustically stereotyped ‘discrete’ calls and more variable ‘graded’ calls. Previous research suggests that both discrete and graded call types are present in the humpback whale acoustic repertoire (Jurasz and Jurasz 1979; Dunlop et al. 2008; Fournet et al. 2015; Dunlop 2017). As with terrestrial animals, discrete calls in humpback whales likely convey ‘fixed’ information which stays relatively stable over time (e.g. body size), while graded calls likely convey ‘flexible’ information related to the internal state of the signaller (e.g. motivation or arousal) (Marler 1975; Green and Marler 1979). Overall, the acoustic repertoire of humpback whales appears to be complex as judged by all of the metrics and criteria for communicative complexity, although quantitative analyses, particularly of the discrete and graded call types, are lacking. Humpback whales, therefore, may fit the SCHCC as assessed by non-traditional metrics.

6.2 Summary of work performed

In humpback whales, although social behaviour is clearly mediated using vocal signals, how potential information is transferred (i.e. call types or call features) remains poorly understood. This is, in part, because of the focus in the literature on the number of call types in the repertoire, rather than the information potentially contained within them. The primary aim of this thesis was to assess how humpback whales communicate during social interactions, and how this might begin to explain

147

Chapter 6: Discussion the apparent communicative complexity observed in this species. To answer these research questions, social interactions relating to breeding behaviour were used given they range from simple (e.g. mother-calf dyads) to relatively complex (e.g. large competitive groups of multiple males competing for reproductive access to a female). Chapters 2-4 used data from the southward migration of the east Australian humpback whale population, during which they swim close to shore on a predictable path, providing a good opportunity for observation and tagging (Noad, M. J., Cato, D. H. & Stokes 2004; Cato et al. 2013). Animals on the southward migration still engage in breeding behaviours, albeit at a reduced level (Brown and Corkeron 1995; Dunlop and Noad 2016). The initial research conducted in Chapter 2 tested whether changes in call production were correlated with group membership (i.e. the addition of male escorts), and presumably to an increase in arousal and changes in motivation. As per terrestrial studies, this research focussed on changes to temporal features of call production, including individual call rates and the use of long bouts of calls, as well as an increase in the variety of call types used. To investigate the type of potential information that is most likely to be conveyed within the calls used in Chapter 2 (i.e. fixed or flexible), Chapter 3 modified an analytic tool that has only recently been applied to the study of animal acoustic communication: ‘fuzzy k-means clustering’ (FKM) (Wadewitz et al. 2015; Fischer et al. 2017b). The novel application of this analysis to the calls of humpback whales classified clusters of calls as ‘call types’. It further revealed which call types were acoustically discrete, with few differences in structure between contexts, or graded, existing along an acoustic continuum. The call types identified in Chapter 3 were compared across behavioural states of varying degrees of sociality in subsequent chapters (Chapters 4-5) to determine their potential function and information content. Chapter 4 focussed on vocal communication during various behavioural states that were variable in the number and diversity of social interactions. States were identified through a cluster analysis, which revealed one female-calf state (‘FC’), one female-calf-escort state (‘FCE’), and two states with a varying number of escorts (‘unstable’ groups and ‘competitive’ groups). Changes in the acoustic structure of calls and the use of specific call types were then related to the type of social interaction, with emphasis on the differential use of discrete and graded calls. Chapter 5 moved away from the moderately active low latitude migration to the high energy breeding grounds proper. It used data from the Great Barrier Reef breeding area to focus exclusively on ‘competitive’ groups because they are one of the most complex social behaviours observed in humpback whales. Here, competitive groups were split into behaviourally distinct ‘intensity levels’ defined by progressively increasing arousal and aggression. Changes to the use of discrete and graded calls were related to changes in the intensity level of the group (i.e. aggression

148

Chapter 6: Discussion and arousal), which progressively increased from more passive displays to highly aggressive fighting. Finally, in this chapter (Chapter 6), the significant findings are discussed, including the potential functions of certain call types, the complexity of the humpback whale repertoire, and implications in regards to the SCHCC.

6.3 The presumed function of humpback calls

Chapters 2, 4, and 5 provided insight into the context in which calls are produced. Chapter 3 determined that call types were either discrete or graded, thus providing an idea of the type of information that could be conveyed in each call type (i.e. fixed or flexible). Combined, these results present a unique and novel opportunity to speculate about the potential function of some humpback whale call types. While many are used across social contexts, some appear to be more important during particular interactions, including the assessment of opponents and presumed dominance sorting. For example, the ‘spiccato’ was highly context specific and only detected in competitive groups (Chapters 2 and 4). This call type was low in peak frequency and bandwidth and the longest in duration of any call type detected during migration (Chapters 2-5). While it displayed variability in duration, it was considered a discrete call due to its distinctiveness in aural and visual assessment (Chapters 2 and 3). Calls that are low in frequency and long in duration are often correlated with hostile or aggressive contexts (Morton 1977; Briefer 2012). Interestingly, the ‘spiccato’ was only ever heard in ‘low intensity level’ competitive groups (i.e. low arousal and aggression; Chapter 5). This use is similar to the ‘roars’ of red deer (Cervus elaphus), a discrete call type produced during the breeding season that contain information regarding the age and weight of the sender, which receivers can use to assess rivals (Clutton-Brock and Albon 1979; Reby and McComb 2003; Reby et al. 2005). Further, ‘roars’ are produced during the early stages of conflict, before escalation to physical fighting (Clutton-Brock and Albon 1979). Thus, it is possible the ‘spiccato’ functions in a similar way, allowing individuals to gain valuable information on opponents in the early stages of intraspecific agonistic conflict (Chapter 5). As in competitive groups, dominance sorting and the assessment of potential competitors is likely also occurring to some degree during the ‘unstable’ state (Chapter 4). Several call types were associated with this state significantly more than with other social states, including ‘meows’. This discrete call type is a short, mid-frequency, down-swept call with higher bandwidth and aggregate entropy than many of the other call types. Despite being classified as discrete, ‘meows’ demonstrated significant variability in acoustic features across the social states (Chapter 4). Interestingly, the ‘unstable’ state was also the social state with the highest variability, with the exception of peak frequency (Chapter 4). These results suggest that ‘meows’ may convey

149

Chapter 6: Discussion information about more than one attribute, similar to some terrestrial species (Taylor and Reby 2010). Calls that contain more than one informational cue are thought to be highly complex, with the dual information potentially serving as a either a ‘backup signal’ or to convey ‘multiple messages’ (Johnstone 1996; Hebets and Papaj 2005). This could include both fixed and flexible information simultaneously, such as dominance rank and age (e.g. baboons, Papio cynocephalus ursinus, Fischer et al. 2004) or individual identity and arousal (e.g. tree shrews, Tupaia belangeri, Schehka and Zimmermann 2009). In particular, individual identity is an important feature to communicate when establishing or maintaining contact (Kondo and Watanabe 2009), such as during changes in group membership. In humpback whales, there is evidence for individual distinctiveness in the ‘cry’ unit of some songs (Hafner et al. 1979), as well as in the ‘feeding cry’ (Cerchio and Dahlheim 2001). However, it is possible that individuality is encoded in multiple call types (e.g. rhesus monkeys, Macaca mulatta, Rendall et al. 1998). ‘Meows’ therefore could function in dominance sorting, containing identity information as well as flexible information related to the immediate arousal or motivation of the group (Chapters 4-5). ‘High frequency squeaks’ and ‘squeak/high frequency eeaws’ were also used in a significantly higher proportion in the ‘unstable’ state (Chapter 4). These call types were the highest in frequency and bandwidth of all call types (Chapter 3). Calls that are high in frequency and broadband are often associated with aversive (i.e. negative) contexts (Morton 1977). This concept is demonstrated in the graded ‘screams’ of chimpanzees (Pan troglodytes), which are produced during antagonistic situations and vary depending on whether they are produced by the aggressor or the victim (Slocombe and Zuberbühler 2007), as well as the intensity of the interaction (Slocombe et al. 2009). Calls that are high in frequency and tonal, or that sharply rise and fall in frequency (as seen in ‘eeaws’) are also associated with distress contexts (Lingle et al. 2012). The general ‘squeaks’ described in humpback whale female-calf pairs are proposed to function as distress calls (Indeck 2020), which supports the use of ‘high frequency squeaks’ and ‘squeak/high frequency eeaws’ as potential indicators of distress. The qualitative aural-visual assessment originally grouped ‘snorts’ as one call type (Chapters 2 and 3) as in previous studies on humpback whale acoustic repertoires (Dunlop et al. 2007, 2008; Rekdahl et al. 2013; Indeck et al. 2020). However, the FKM split this call primarily amongst three clusters (Chapter 3). One ‘snort’ cluster was found to be discrete (Chapter 3) and specific to the ‘FCE’ and the ‘competitive state’ (Chapter 4). ‘Discrete snorts’ were short, narrow- band calls with some of the lowest frequency measurements in the repertoire (Chapter 3). Despite being classified as discrete, this call type showed significant variability between the two states for duration, peak frequency, and bandwidth. For all three acoustic parameters, the variability was

150

Chapter 6: Discussion greatest in the ‘competitive’ state (Chapter 4). During competitive groups, it was found in all three intensity levels, but was used in a significantly higher proportion in moderate and high intensity groups (Chapter 5). These results suggest that, like ‘meows’, ‘discrete snorts’ may function to convey multiple messages. This could include temporary dominance and/or body condition, as in the ‘snorts’ of male rock hyraxes (Procavia capensis), which are thought to function in conveying information regarding dominance status and/or testosterone levels (Koren and Geffen 2009; Demartsev et al. 2016; Weissman et al. 2019). The remaining ‘snort’ clusters were quantitatively determined to be graded (‘low entropy snort/knocks’ and ‘grumble/long snorts’), which agrees with previous studies of humpback whale calls (Dunlop et al. 2007, 2008; Rekdahl et al. 2013; Indeck et al. 2020). They were used in a broader range of contexts, indicating a more general function. The graded ‘snorts’ described by Dunlop et al. (2008) were detected most often in groups of more than one animal and were proposed to function in regulating intra-group interactions. Both ‘low entropy snort/knocks’ and ‘grumble/long snorts’ were used across all four social states (Chapter 4). Their widespread use here supports the findings of Dunlop et al. (2008) that they are likely important in intra-group interactions. Alternatively, ‘low entropy snort/knocks’ may function more in female-calf communication. While used across all social contexts, this call type was used in significantly higher proportions in the ‘FC’ state (i.e. female-calf dyads, Chapter 4). ‘Knocks’ are thought to be a call type produced only by female-calf pairs (Chapter 2), as in the ‘pulses’ of North Atlantic right whale mother-calf pairs (Parks et al. 2019). Several other call types emerged as potentially important for female-calf communication, possibly specific to calves. ‘Wup/low frequency eeaws’ were used across all social contexts, but were also found significantly more often in the ‘low social complexity’ state (Chapter 4). Further, ‘wups’ are proposed to be a calf call in humpback whale female-calf pairs (Indeck et al. 2020). The broad use of ‘wup/low frequency eeaws’, combined with an increased use in female-calf only pairs (Chapter 4), supports their proposed function as calf calls. ‘Bops’ are also presumed to be produced predominately by calves (Indeck et al. 2020). In this thesis, the FKM split this call into four call types: ‘broadband bops’, ‘high entropy bops’, ‘high frequency bops’, and ‘low entropy bops’ (Chapter 3). The first three were all relatively high in entropy (Chapter 3), which is a feature of infant calls in birds (Redondo and Exposito 1990; Wein et al. 2019). While no differences were detected in the use of these call types between social states (Chapter 4), a calf was present at all times in the tag data, which supports their designation as a calf call (Chapters 2-4). In contrast, ‘low entropy bops’ were lower in both entropy and frequency (Chapter 3). This could indicate that this

151

Chapter 6: Discussion call type is not a ‘calf’ call as is proposed for the other three call types, and/or has a more broad function in female-calf communication due to its prevalence across contexts (Chapter 3).

6.4 Humpback whales have high communicative complexity

Humpback whales have a complex acoustic repertoire as defined by a large number of call types, variability between- and within-calls, and evidence of context-specificity. The call repertoire consists of between 34 and 46 call types in any given year (Dunlop et al. 2007; Rekdahl et al. 2013). Some calls are ‘stable’ and exist across years, while others are only present in the repertoire for one or two years. The number of ‘stable’ calls in the repertoire is smaller, between 12 and 13 call types (Dunlop et al. 2007; Rekdahl et al. 2013). This is partly due to the changing presence of ‘song unit social sounds’ which are typically only present in the same or previous year as the song in which they occur (Rekdahl et al. 2013). The results of the FKM indicated 15 call types (Chapter 3), with most (9 out of 15) individual calls (data points) considered to be highly variable between- and within-call types (i.e. ‘graded’). The presence of graded calls has been reported for several other baleen whales, specifically the North Atlantic right whale, southern right whale, and bowhead whale. They are all reported to have ‘complex calls’ in their repertoires, including both discrete and graded calls (Clark 1990). However, all have less than 8 well described call types (Clark 1983, 1990; Parks 2003a; Parks and Tyack 2005; Parks et al. 2007). While a quantitative analysis like that used in Chapter 3 is lacking, there is currently no evidence of a call repertoire that is as large and fluid as the humpback whale. While six call types were considered to be ‘discrete’, they occasionally exhibited significant variability in acoustic features, particularly during socially complex states (Chapters 2, 4). Combined with the overall skew towards gradation, this indicates that all calls in the repertoire contain some level of variation (Marler 1976; Morton 1982; Manser 2013) and have the potential to encode a large amount of information (Fischer et al. 2017b). These results also support previous research indicating the plasticity of humpback whale calls. During increased noise, humpback whales are able to adjust the frequency of their calls (Parks et al. 2016) as well as the amplitude of calls (Dunlop et al. 2014; Dunlop 2016), potentially as a way to increase the detectability over high background noise. Humpback whales can also adjust the amplitude and fundamental frequency of calls depending on the presence of bystanders, indicating vocal plasticity as the result of social influences as well as environmental (Dunlop 2016). Vocal plasticity in acoustic features is linked with vocal learning (Snowdon 2009), a concept that is proposed to increase the complexity of a repertoire (Peckre et al. 2019).

152

Chapter 6: Discussion

Two acoustic features in particular were consistently variable across social contexts, regardless of whether a call was discrete or graded: duration (i.e. the length of the sound) and aggregate entropy (i.e. the ‘disorder’ of the sound determined by frequency range) (Chapter 4). These results indicate that these features may be two of the most flexible in the humpback whale acoustic repertoire. Changes in the duration of animal vocalisations are often associated with complex changes in the internal state of the signaller. For example, changes in duration are typically related to changes in arousal (Briefer 2012), with longer calls produced in higher arousal situations (Chapter 2), and increased variability during complex social situations (Chapter 4). Duration has also been linked to aggressive motivational states, with longer calls produced in more hostile contexts (Morton 1977; Briefer 2012). The link between entropy and complexity is less clear, although other vocal ‘nonlinearities’ (e.g. biphonation, frequency jumps, deterministic chaos) may generate complex vocalisations (Fitch et al. 2002). These vocal phenomena may increase auditory impact on listeners by providing cues to fitness, mate quality, urgency, individuality, provide arousal, or age (Fitch et al. 2002; Riede et al. 2004; Briefer 2012; Root-Gutteridge et al. 2018). Therefore, it is likely that entropy is also related to complex information in calls. Changes to acoustic features can increase the informative value of a call. This can also increase with the level of context specificity of a call (Bradbury and Vehrencamp 2011; Fischer et al. 2017b). While most calls were flexible in their use, some calls were used in a much more limited capacity, particularly the ‘spiccato’ which was only detected in competitive groups (Chapters 2, 4). Context-specific calls have been observed in other studies of humpback whale communication, specifically during feeding, with different discrete call types correlated with different feeding modes and habitats (Jurasz and Jurasz 1979; Cerchio and Dahlheim 2001; Stimpert et al. 2007; Parks et al. 2014; Fournet et al. 2018). There is also context specificity as male song during the breeding season (Payne and McVay 1971). Context-specific vocalisations are not common in baleen whales, however there are exceptions, including the presence of songs [e.g. fin whale (Croll et al. 2002); blue whale (McDonald et al. 2001); bowhead whale (Stafford et al. 2008)]. The songs of fin and blue whales are simple repeated trains of pulses and lack the complexity of humpback whale song (McDonald et al. 2001; Croll et al. 2002). The songs of bowhead whales are more complex, displaying high diversity and changing over the course of hours or days, however they typically only consist of a single phrase (Stafford et al. 2008). Evidence of context specificity also occurs outside of song but may be limited to North Atlantic right whales, which produce ‘scream’ calls only during ‘surface active groups’ (SAGs), and to the ‘gunshot’ displays of both North Atlantic and southern right whales (Parks and Tyack 2005; Parks et al. 2005). Overall, the level of context specificity evident in the humpback repertoire, the presence of complex songs, the large

153

Chapter 6: Discussion number of call types, and the presence of graded calls that vary with behavioural state and presumed motivation and arousal, suggests that the humpback acoustic repertoire is one of the most complex of any baleen whale.

6.5 Humpbacks fit the SCHCC in a novel way

According to the SCHCC, humpback whales should not have such a complex vocal repertoire due to their mostly unstable and short-term associations (Clapham 1996). However, they do exhibit a wide range of complex social behaviours and interactions, particularly during breeding behaviours (Clapham 1996, Chapters 4-5). These range from single males singing as a display for nearby conspecifics, to simple social relationships between females and their calves to highly social competitive groups (Clapham 1996). These interactions are facilitated using the complex song described elsewhere and the results described here, with an increasing number of call types and acoustic variability as the number and diversity of social interactions increases (Chapters 2, 4-5). Judging by the distance these calls can be heard, their social network is expansive, with ranges of up to 5 km during intra-group interactions and extending to more than 10 km while singing (Dunlop et al. 2013). Therefore, it is likely that not only the overall sociality of a species, but also fine-scale social structure based on individual relationships contributes to the evolution of a complex communication system in animals. Overall, this suggests that humpback whales do fit the SCHCC, but in a novel way. The wide range of behaviours during breeding interactions appears to be uncommon in cetaceans. Some baleen species do engage in competitive behaviours in breeding contexts, most notably in the SAGs of North Atlantic right whales (Kraus and Hatch 2001; Parks 2003b; Parks and Tyack 2005; Parks et al. 2007), southern right whales (Clark 1983), and bowhead whales (Würsig et al. 1993). However, the level of aggression is lower, presumably the result of mating strategies related to sperm competition, a tactic which humpback whales do not utilise (Brownell, Jr. and Ralls 1986; Clapham 1996; Mesnick and Ralls 2009). Interestingly, the other species of rorquals do not engage in sperm competition either, however there is no evidence of male-male competition over females in these species (Notarbartolo-di-Sciara et al. 2003; Aguilar and García-Vernet 2009). The complex feeding behaviour of humpback whales is also in stark contrast with other baleen whales, and in fact other asocial carnivores in general. A variety of foraging tactics have been observed in this species, including coordinated or cooperative feeding (D’Vincent et al. 1985; Sharpe 2001; Wiley et al. 2011; Parks et al. 2014). While many, if not all, baleen whales congregate to certain foraging grounds, there is little evidence of cooperation in any of these species. Thus, in terms of social complexity, humpback whales appear to be an anomaly amongst baleen whales.

154

Chapter 6: Discussion

While the results here help to elucidate the link between social and communicative complexity in humpbacks, it remains unclear if they are truly an asocial species which displays complex social behaviour or are in fact a social species which simply does not fulfil the current metrics of a complex social system. Regardless, how and why such behavioural social complexities evolved in this species but no other baleen whales remains to be answered. Further research dedicated to the evolution of sociality will be required in order to answer this question.

6.6 Limitations and future directions

One of the most pivotal features of the fuzzy clustering method was the ability to isolate different call types that aural-visual analyses could not distinguish (e.g. ‘snorts’ and ‘bops’). This highlights one of the benefits of using quantitative classification for graded repertoires (Chapter 3). Further, it provided a new method of classifying the acoustic repertoire of the humpback whale, whilst also distinguishing between calls that were ‘discrete’ and ‘graded’ (Chapter 3). Since this analysis can account for and quantify gradation in a repertoire, this technique should be the focus for further classification analyses in this species. Despite this improvement over the techniques typically applied to humpback call classification (e.g. DFA, k-means clustering), some input from the analyst (i.e. the determination of cut-offs for discrete calls) may always be necessary for data sets with graded call types (Wadewitz et al. 2015). Variability does not automatically exclude a call type from being classified as discrete (van Hooff and Preuschoft 2003), and purely quantitative techniques might not be able to distinguish subtle variations in discrete call types from substantial variations in graded call types. Improvements to the FKM for humpback whales would be the incorporation of additional acoustic features, particularly as data sets with a higher number of acoustic features lead to better clustering results (Wadewitz et al. 2015). Furthermore, pulse repetition rate and the number of inflection points have been important in the classification of humpbacks and other cetacean species (Rekdahl et al. 2013, 2017; Garland et al. 2015). Future research should aim to determine which set of features is most important for classification and set this as a standard across research groups (Seger 2016). This is a critical step in order to compare across studies, populations, geographical areas, and eventually species. While determining which acoustic features are most important for classification is key, these parameters may not be those most important for receivers. In other words, while clustering techniques may provide an idea of the number of call types in the data (as clusters), this does not mean that they hold any biological significance (Hauser 1996). This is particularly true for the graded call types because some species perceive continuous signals as distinct and meaningful categories (Fischer 2006). To fully understand the communication system of this species, it is

155

Chapter 6: Discussion necessary to also consider the way signal variation affects the behaviour of receivers (Freeberg et al. 2012; Fischer 2013; Fischer et al. 2017b). This variation can be both within a signal (i.e. changes to acoustic features) and between signals (i.e. changes to call types). To accomplish these goals, response studies will be needed involving playback experiments, whereby receiver responses to broadcast vocalisations are recorded (Seyfarth and Cheney 2003; Fischer 2013). Varying between- and within-call features would provide an opportunity to determine how this affects the responses of receivers. While the results of this thesis have provided some idea of which calls likely contain fixed and/or flexible information, it will be necessary to conduct a detailed analysis of the link between the acoustic features of these calls and the physical and physiological attributes of the signaller (Fischer 2013). This presents considerable challenges for studies involving a marine species. Humpback whales spend much of their lives submerged, and most of the assumptions made about their behaviour are based on limited surface observations (Chapter 5). Further, it is currently not possible to reliably identify the individual calling in a group of free-swimming cetaceans. While animal-borne archival tags like those used here provide information on the movement behaviour of the tagged individual and the immediate acoustic environment, there is no way to discriminate calls produced by the focal animal calling or a conspecific nearby. Tags equipped with video and acoustic recorders (e.g. CATS Cam, www.cats.is) could pave the way for future research of this nature, allowing researchers to observe the context of calls, as well as the proximity of other individuals. Another important consideration for future research is to incorporate calls from other social contexts. The east Australian population of humpback whales is reported to have several stable call types that were not detected here (e.g. ‘cry’ and ‘yap’) (Dunlop et al. 2007; Rekdahl et al. 2013). One explanation for this is that some social contexts were not included in the present study (e.g. single animals, female-escort pairs with no calf). Although it is possible these calls are no longer part of the call repertoire (Rekdahl et al. 2013), it is more likely that the full repertoire simply was not captured. Incorporating additional social and behavioural contexts will add further insight into the social complexity of this species. A true test of the SCHCC remains to be conducted. This is partly because of the need for a quantitative analysis of the social complexity in humpback whales (Fischer et al. 2017a). In addition, alternative hypotheses as to how the complex communication system evolved also need to be explored, including selection pressures related to habitat, predation, and species recognition (Freeberg et al. 2012; Fischer et al. 2017b). For example, habitat could be expected to be a strong selection pressure for humpback calls as they are primarily limited to vocal communication. As such, their signals would suffer the same selection pressures as species in dense

156

Chapter 6: Discussion forest, where discrete calls are used more for long-distance and between group communication while graded calls are more useful in close contact (Ey and Fischer 2009). However habitat is thought to be more linked to limitations on signal design rather than promoting signal diversity, as is predation pressure (Freeberg et al. 2012; Fischer et al. 2017b). Species recognition is another alternative hypothesis, and one that might be more applicable to explain the communicative complexity of humpback whales. In an area where multiple species of similar behaviour, ecology, and/or appearance cohabitate, one of the most viable solutions for species recognition might be to increase signal complexity (Freeberg et al. 2012). Although modifying existing signal characteristics is arguably a simpler pathway, the range of possible modifications that will reduce overlap with heterospecific signals is physiologically constrained and limited (Freeberg et al. 2012). This concept could also apply to increasing the number of call types. Right whales are sympatric with humpback whales, in both the Northern and Southern hemispheres (Department of the Environment 2020a, b; Hayes et al. 2020), and both species can produce calls that are similar enough to confound passive acoustic monitoring techniques (Munger et al. 2005; Mellinger et al. 2007). It is possible that this overlap contributed to a more diverse repertoire in humpback whales, although why it should not in right whales is unclear. Further research should investigate this and other ‘null hypotheses’ in regards to the SCHCC in humpback whales, as well as baleen whales in general.

6.7 Concluding remarks

The results of this thesis have provided an informative view of the production side of humpback communication, and under what circumstances fixed and flexible information is expected to be present in calls. This research represents an important step towards understanding the function of calls in humpback whales, however it also has broader implications for the fields of animal behaviour and communication. The application of a novel classification technique can form the basis for comparisons between other baleen whale species to investigate the evolution of complex communication systems. Further, the use of calls during intraspecific conflict is similar to strategies used in terrestrial animals during breeding behaviour, with discrete calls used during initial displays, and higher call rates used as conflict progresses. This is a link which can be investigated in regards to the evolution of mating systems and strategies. Lastly, this research demonstrates that the relationship between social and communicative complexity is not straight forward. Adjustments to the definitions of both may need to be considered for animals that are highly mobile and socialise over a range of geographic scales.

157

Chapter 6: Discussion

6.8 References

Aguilar A, García-Vernet R (2009) Fin Whale. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 368–371. Allen JA (2019) Community through culture: From insects to whales. How social learning and culture manifest across diverse animal communities. BioEssays 41:1–8. Baker CS, Herman LM (1984) Aggressive behavior between humpback whales (Megaptera novaeangliae) wintering in Hawaiian waters. Canadian Journal of Zoology 62:1922–1937. Bergman TJ, Beehner JC (2015) Measuring social complexity. Animal Behaviour 103:203–209. Bradbury JW, Vehrencamp SL (2011) Principles of Animal Communication, 2nd edn. Sinauer Associates, Inc., Sunderland, MA. Briefer EF (2012) Vocal expression of emotions in mammals: Mechanisms of production and evidence. Journal of Zoology 288:1–20. Brown M, Corkeron P (1995) Pod characteristics of migrating humpback whales (Megaptera novaeangliae) off the East Australian coast. Behaviour 132:163–179. Brownell, Jr. RL, Ralls K (1986) Potential for sperm competition in baleen whales. Report of the International Whaling Commission Special Is:97–112. Burnham R, Duffus D, Mouy X (2018) Gray whale (Eschrictius robustus) call types recorded during migration off the west coast of Vancouver Island. Frontiers in Marine Science 5:1–11. Campagna C (2009) Aggressive Behavior, Intraspecific. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 18–24. Cato DH, Noad MJ, Dunlop RA, et al (2013) A study of the behavioural response of whales to the noise of seismic air guns, design, methods and progress. Acoustics Australia 41:88–97. Cerchio S, Dahlheim M (2001) Variation in feeding vocalizations of humpback whales Megaptera novaeangliae from southeast Alaska. Bioacoustics 11:277–295. Clapham PJ (1993) Social organization of humpback whales on a North Atlantic feeding ground. Zoological Symposium 66:131–145. Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mammal Review 26:27–49. Clapham PJ (2000) The humpback whale: seasonal feeding and breeding in a baleen whale. In: Mann J, Connor R, Tyack PL, Whitehead H (eds) Cetacean Societies: Field Studies of Dolphins and Whales. University of Chicago Press, Chicago, pp 173–196. Clapham PJ, Palsboll PJ, Mattila DK, Vasquez O (1992) Composition and dynamics of humpback whale competitive groups in the West Indies. Behaviour 122:182–194. Clark CW (1982) The acoustic repertoire of the Southern right whale, a quantitative analysis.

158

Chapter 6: Discussion

Animal Behaviour 30:1060–1071. Clark CW (1983) Acoustic communication and behavior of the southern right whale (Eubaleana australis). In: Payne RS (ed) Communication and Behavior of Whales. Westview Press, Boulder, CO, pp 163–198. Clark CW (1990) Acoustic behavior of mysticete whales. In: Thomas J, Kastelein R (eds) Sensory Abilities of Cetaceans. Plenum Press, New York, pp 571–583. Clutton-Brock TH, Albon SD (1979) The roaring of red deer and the evolution of honest advertisement. Behaviour 69:145–170. Croll DA, Clark CW, Acevedo A, Tershy B, Flores S, Gedamke J, Urban J (2002) Only male fin whales sing loud songs. Nature 417:809. Cummings WC, Thompson PO, Cook R (1968) Underwater sounds of migrating gray whales, Eschrichtius glaucus (Cope). The Journal of the Acoustical Society of America 44:1278–1281. D’Vincent CG, Nilson RM, Hanna RE (1985) Vocalizations and coordinated feeding behavior of the humpback whale in southeastern Alaska. Scientific Report of the Whales Research Institute 36:41–47. Demartsev V, Bar Ziv E, Shani U, Goll Y, Koren L, Geffen E (2016) Harsh vocal elements affect counter-singing dynamics in male rock hyrax. Behavioral Ecology 27:1397–1404. Department of the Environment (2020a) Megaptera novaeangliae. In: Species Profile Threat. Database. http://www.environment.gov.au/sprat. Accessed 21 Nov 2020 Department of the Environment (2020b) Eubalaena australis. In: Species Profile Threat. Database Dunlop RA (2017) Potential motivational information encoded within humpback whale non-song vocal sounds. The Journal of the Acoustical Society of America 141:2204–2213. Dunlop RA (2016) Changes in vocal parameters with social context in humpback whales: considering the effect of bystanders. Behavioral Ecology and Sociobiology 70:857–870. Dunlop RA, Cato DH, Noad MJ (2008) Non-song acoustic communication in migrating humpback whales (Megaptera novaeangliae). Marine Mammal Science 24:613–629. Dunlop RA, Cato DH, Noad MJ (2014) Evidence of a Lombard response in migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 136:430– 437. Dunlop RA, Cato DH, Noad MJ, Stokes DM (2013) Source levels of social sounds in migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 134:706–714. Dunlop RA, Noad MJ (2016) The “risky” business of singing: Tactical use of song during joining by male humpback whales. Behavioral Ecology and Sociobiology 70:2149–2160.

159

Chapter 6: Discussion

Dunlop RA, Noad MJ, Cato DH, Stokes DM (2007) The social vocalization repertoire of east Australian migrating humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 122:2893–2905. Ey E, Fischer J (2009) The “acoustic adaptation hypothesis”—A review of the evidence from birds, anurans and mammals. Bioacoustics 19:21–48. Fischer J (2006) Categorical perception in animals. In: Brown K (ed) Encyclopedia of language and linguistics, 2nd edn. Elsevier Ltd, London, pp 248–251. Fischer J (2013) Information, inference and meaning in primate vocal behaviour. In: Stegmann U (ed) Animal Communication Theory: Information and Influence. Cambridge University Press, Cambridge, MA, pp 297–317. Fischer J, Farnworth MS, Sennhenn-Reulen H, Hammerschmidt K (2017a) Quantifying social complexity. Animal Behaviour 130:57–66. Fischer J, Kitchen DM, Seyfarth RM, Cheney DL (2004) Baboon loud calls advertise male quality: Acoustic features and their relation to rank, age, and exhaustion. Behavioral Ecology and Sociobiology 56:140–148. Fischer J, Wadewitz P, Hammerschmidt K (2017b) Structural variability and communicative complexity in acoustic communication. Animal Behaviour 134:229–237. Fitch WT, Neubauer J, Herzel H (2002) Calls out of chaos: The adaptive significance of nonlinear phenomena in mammalian vocal production. Animal Behaviour 63:407–418. Fournet ME, Szabo A, Mellinger DK (2015) Repertoire and classification of non-song calls in Southeast Alaskan humpback whales (Megaptera novaeangliae). The Journal of the Acoustical Society of America 137:1–10. Fournet MEH, Gabriele CM, Sharpe F, Straley JM, Szabo A (2018) Feeding calls produced by solitary humpback whales. Marine Mammal Science 34:851–865. Freeberg TM, Dunbar RIM, Ord TJ (2012) Social complexity as a proximate and ultimate factor in communicative complexity. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1785–1801. Garland EC, Castellote M, Berchok CL (2015) Beluga whale (Delphinapterus leucas) vocalizations and call classification from the eastern Beaufort Sea population. The Journal of the Acoustical Society of America 137:3054–3067. Gerhardt HC (1992) Multiple messages in acoustic signals. Seminars in Neuroscience 4:391–400. Green S, Marler P (1979) The Analysis of Animal Communication. In: Social Behavior and Communication. Plenum Press, New York (NY), pp 73–158. Hafner GW, Hamilton CL, Steiner WW, Thompson TJ, Winn HE (1979) Signature information in

160

Chapter 6: Discussion

the song of the humpback whale. Journal of the Acoustical Society of America 66:1–6. Hauser MD (1996) The evolution of communication. MIT Press, Cambridge, MA. Hayes SA, Josephson E, Maze-Foley K, Rosel PE (2020) US Atlantic and Gulf of Mexico Marine Mammal Stock Assessments - 2019 Hebets EA, Papaj DR (2005) Complex signal function: Developing a framework of testable hypotheses. Behavioral Ecology and Sociobiology 57:197–214. Indeck KL (2020) Acoustic communication of female-calf humpback whales during migration. PhD Thesis. The University of Queensland, Queensland. Indeck KL, Girola E, Torterotot M, Noad MJ, Dunlop RA (2020) Adult female-calf acoustic communication signals in migrating east Australian humpback whales. Bioacoustics Johnstone RA (1996) Multiple displays in animal communication: “backup signals” and “multiple messages.” Philosophical Transactions of the Royal Society B: Biological Sciences 351:329– 338. Jurasz CM, Jurasz VP (1979) Feeding modes of the humpback whale, Megaptera noavaengliae, in southeast Alaska. Scientific Report of the Whales Research Institute 31:69–83. Kappeler PM (2019) A framework for studying social complexity. Behavioral Ecology and Sociobiology 73:1–14. Kappeler PM, Clutton-Brock T, Shultz S, Lukas D (2019) Social complexity: patterns, processes, and evolution. Behavioral Ecology and Sociobiology 73:1–6. Kondo N, Watanabe S (2009) Contact calls: Information and social function. Japanese Psychological Research 51:197–208. Koren L, Geffen E (2009) Complex call in male rock hyrax (Procavia capensis): A multi- information distributing channel. Behavioral Ecology and Sociobiology 63:581–590. Kraus SD, Hatch JJ (2001) Mating strategies in the North Atlantic right whale. Journal of Cetacean Research and Management Special Is:237–244. Lingle S, Wyman MT, Kotrba R, Teichroeb LJ, Romanow CA (2012) What makes a cry a cry? A review of infant distress vocalizations. Current Zoology 58:698–726. Manser MB (2013) Semantic communication in vervet monkeys and other animals. Animal Behaviour 86:491–496. Marler P (1975) On the origin of speech from animal sounds. In: Kavanagh JF, Cutting J (eds) The Role of Speech in Language. MIT Press, Cambridge (MA), pp 11–37. Marler P (1976) Social organization, communication, and graded signals: The chimpanzee and the gorilla. In: Bateson PPG, Hinde RA (eds) Growing Points in Ethology. Cambridge University Press, Oxford (UK), pp 239–277.

161

Chapter 6: Discussion

Mattila DK, Clapham PJ, Vasquez O, Bowman RS (1994) Occurrence, population composition, and habitat use of humpback whales in Samana Bay, Dominican Republic. Canadian Journal of Zoology 72:1898–1907. McDonald MA, Calambokidis J, Teranishi AM, Hildebrand JA (2001) The acoustic calls of blue whales off California with gender data. The Journal of the Acoustical Society of America 109:1728–1735. Mellinger DK, Nieukirk SL, Matsumoto H, Heimlich SL, Dziak RP, Haxel J, Fowler M, Meinig C, Miller H V. (2007) Seasonal occurrence of North Atlantic right whale (Eubalaena glacialis) vocalizations at two sites on the Scotian Shelf. Marine Mammal Science 23:856–867. Mesnick SL, Ralls K (2009) Mating Systems. In: Perrin WF, Wursig B, Thewissen JGM (eds) Encyclopedia of Marine Mammals. Academic Press, Amsterdam, pp 712–719. Mobley, Jr. JR, Herman LM (1985) Transience of social affiliations among humpback whales on the Hawaiian wintering grounds. Canadian Journal of Zoology 63:762–772. Morton ES (1977) On the occurrence and significance of motivation-structural rules in some bird and mammal sounds. The American Naturalist 111:855–869. Morton ES (1982) Grading, discreteness, redundancy, and motivation-structural rules. In: Kroodsma DE, Miller MH (eds) Acoustic communication in birds. Academic Press, New York (NY), pp 183–212. Munger LM, Mellinger DK, Wiggins SM, et al (2005) Performance of spectrogram cross- correlation in detecting right whale calls in long-term recordings from the Bering Sea. Canadian Acoustics 33:25–34. Noad, M. J., Cato, D. H. & Stokes MD (2004) Acoustic tracking of humpback whales: measuring interactions with the acoustic environment. In: Proceedings of Acoustics 2004. Gold Coast, Australia, pp 353–358. Notarbartolo-di-Sciara G, Zanardelli M, Jahoda M, Panigada S, Airoldi S (2003) The fin whale Balaenoptera physalus in the metiterranean sea. Mammal Review 33:105–150. Parks SE (2003a) Acoustic Communication in the North Atlantic Right Whale (Eubalaena glacialis). Massachusetts Institute of Technology. Parks SE (2003b) Response of North Atlantic right whales (Eubalaena Glacialis) to playback of calls recorded from surface active groups in both the North and South Atlantic. Marine Mammal Science 19:563–580. Parks SE, Brown MW, Conger LA, Hamilton PK, Knowlton AR, Kraus SD, Slay CK, Tyack PL (2007) Occurrence, composition, and potential functions of North Atlantic right whale (Eubalaena glacialis) surface active groups. Marine Mammal Science 23:868–887.

162

Chapter 6: Discussion

Parks SE, Cusano DA, Bocconcelli A, Friedlaender AS, Wiley DN (2016) Noise impacts on social sound production by foraging humpback whales. Proceedings of Meetings on Acoustics 27:1– 8. Parks SE, Cusano DA, Parijs SM Van, Nowacek DP (2019) North Atlantic right whale (Eubalaena glacialis) acoustic behavior on the calving grounds. JASA Express Letters 146:15–21. Parks SE, Cusano DA, Stimpert AK, Weinrich MT, Friedlaender AS, Wiley DN (2014) Evidence for acoustic communication among bottom foraging humpback whales. Scientific Reports 4:7508. Parks SE, Hamilton PK, Kraus SD, Tyack PL (2005) The gunshot sound produced by male North Atlantic right whales (Eubalaena glacialis) and its potential function in reproductive advertisment. Marine Mammal Science 21:458–475. Parks SE, Searby A, Célérier A, Johnson MP, Nowacek DP, Tyack PL (2011) Sound production behavior of individual North Atlantic right whales: implications for passive acoustic monitoring. Endangered Species Research 15:63–76. Parks SE, Tyack PL (2005) Sound production by North Atlantic right whales (Eubalaena glacialis) in surface active groups. The Journal of the Acoustical Society of America 117:3297–3306. Payne RS, McVay S (1971) Songs of humpback whales. Science 173:585–597. Peckre L, Kappeler PM, Fichtel C (2019) Clarifying and expanding the social complexity hypothesis for communicative complexity. Behavioral Ecology and Sociobiology 73:1–19. Pomilla C, Rosenbaum HC (2006) Estimates of relatedness in groups of humpback whales (Megaptera novaeangliae) on two wintering grounds of the Southern Hemisphere. Molecular Ecology 15:2541–2555. Ramp C, Hagen W, Palsbøll P, Bérubé M, Sears R (2010) Age-related multi-year associations in female humpback whales (Megaptera novaeangliae). Behavioral Ecology and Sociobiology 64:1563–1576. Reby D, McComb K (2003) Anatomical constraints generate honesty: Acoustic cues to age and weight in the roars of red deer stags. Animal Behaviour 65:519–530. Reby D, McComb K, Cargnelutti B, Darwin C, Fitch WT, Clutton-Brock T (2005) Red deer stags use formants as assessment cues during intrasexual agonistic interactions. Proceedings of the Royal Society B: Biological Sciences 272:941–947. Redondo T, Exposito F (1990) Structural variations in the begging calls of nestling magpies Pica pica and their role in the development of adult voice. Ethology 84:307–318. Rekdahl ML, Dunlop RA, Noad MJ, Goldizen AW (2013) Temporal stability and change in the social call repertoire of migrating humpback whales. The Journal of the Acoustical Society of

163

Chapter 6: Discussion

America 133:1785–1795. Rekdahl ML, Tisch C, Cerchio S, Rosenbaum H (2017) Common nonsong social calls of humpback whales (Megaptera novaeangliae) recorded off northern Angola, southern Africa. Marine Mammal Science 33:365–375. Rendall D, Owren MJ, Rodman PS (1998) The role of vocal tract filtering in identity cueing in rhesus monkey (Macaca mulatta) vocalizations. The Journal of the Acoustical Society of America 103:602–614. Riede T, Owren MJ, Arcadi AC (2004) Nonlinear acoustics in pant hoots of common chimpanzees (Pan troglodytes): Frequency jumps, subharmonics, biphonation, and deterministic chaos. American Journal of Primatology 64:277–291. Root-Gutteridge H, Cusano DA, Shiu Y, Nowacek DP, Van Parijs SM, Parks SE (2018) A lifetime of changing calls: North Atlantic right whales, Eubalaena glacialis, refine call production as they age. Animal Behaviour 137:21–34. Schehka S, Zimmermann E (2009) Acoustic features to arousal and identity in disturbance calls of tree shrews (Tupaia belangeri). Behavioural Brain Research 203:223–231. Seger KD (2016) Ambient acoustic environments and cetacean signals, baseline studies from humpback whale and gray whale breeding grounds. University of California, San Diego. Seyfarth RM, Cheney DL (2003) Signalers and receivers in animal communication. Annual Reviews in Psychology 54:145–173. Sharpe FA (2001) Social foraging of the southeast Alaskan humpback whale, Megaptera novaeangliae. PhD Thesis. Simon Fraser University, Ann Arbor, MI. Širović A, Rice A, Chou E, Hildebrand JA, Wiggins SM, Roch MA (2015) Seven years of blue and fin whale call abundance in the Southern California Bight. Endangered Species Research 28:61–76. Slocombe KE, Townsend SW, Zuberbühler K (2009) Wild chimpanzees (Pan troglodytes schweinfurthii) distinguish between different scream types: Evidence from a playback study. Animal Cognition 12:441–449. Slocombe KE, Zuberbühler K (2007) Chimpanzees modify recruitment screams as a function of audience composition. Proceedings of the National Academy of Sciences of the United States of America 104:17228–17233. Snowdon CT (2009) Plasticity of Communication in Nonhuman Primates. In: Naguib M, Janik VM, Zuberbühler K, Clayton NS (eds) Advances in the Study of Behavior. Academic Press, Amsterdam, pp 239–276. Stafford KM, Moore SE, Laidre KL, Heide-Jørgensen MP (2008) Bowhead whale springtime song

164

Chapter 6: Discussion

off West Greenland. The Journal of the Acoustical Society of America 124:3315–3323. Stimpert AK, Wiley DN, Au WWL, Johnson MP, Arsenault R (2007) ‘ Megapclicks ’: acoustic click trains and buzzes produced during night-time foraging of humpback whales (Megaptera novaeangliae). Biology Letters 1–4. Taylor AM, Reby D (2010) The contribution of source-filter theory to mammal vocal communication research. Journal of Zoology 280:221–236. Valsecchi E, Hale P, Corkeron P, Amos W (2002) Social structure in migrating humpback whales (Megaptera novaeangliae). Molecular Ecology 11:507–518. van Hooff J, Preuschoft S (2003) Laughter and smiling: the intertwining of nature and culture. In: de Waal F, Tyack P (eds) Animal social complexity: intelligence, culture and individualized societies. Harvard University Press, Cambridge, MA, pp 260–287. Wadewitz P, Hammerschmidt K, Battaglia D, Witt A, Wolf F, Fischer J (2015) Characterizing Vocal Repertoires — Hard vs . Soft Classification Approaches. PLoS ONE 10:1–16. Watkins WA, Tyack P, Moore KE, Bird JE (1987) The 20-Hz signals of finback whales (Balaenoptera physalus). Journal of the Acoustical Society of America 82:1901–1912. Wein A, Schwing R, Huber L (2019) Kea Nestor notabilis mothers produce nest-specific calls with low amplitude and high entropy. Ibis Published:1–12. Weissman YA, Demartsev V, Ilany A, Barocas A, Bar-Ziv E, Shnitzer I, Geffen E, Koren L (2019) Acoustic stability in hyrax snorts: Vocal tightrope-walkers or wrathful verbal assailants? Behavioral Ecology 30:223–230. Whitehead H (1983) Structure and stability of humpback whale groups off Newfoundland. Can J Zool 61:1391–1397. Wiley D, Ware C, Bocconcelli A, Cholewiak D, Friedlaender A, Thompson M, Weinrich M (2011) Underwater components of humpback whale bubble-net feeding behaviour. Behaviour 148:575–602. Wursig B, Dorsey EM, Fraker MA, Payne RS, Richardson WJ (1985) Behavior of bowhead whales summering in the Beaufort Sea: A description. Fishery Bulletin 83:357–377. Würsig B, Guerrero J, Silber GK (1993) Social and sexual behavior of bowhead whales in fall in the Western Arctic: A re‐examination of seasonal trends. Marine Mammal Science 9:103–115. Zahavi A (1982) The pattern of vocal signals and the information they convey. Behaviour 80:1–8.

165

Appendices

Appendix 1: Humpback whale social call production

Table A1.1 The 26 tags used in analyses. Total tag duration is the time spent in a stable group composition, excluding the first 10 minutes after tag deployment and. FC: female-calf pair, FCE: female-calf-escort group, FCME: female-calf-multiple escort group, SNR: signal-to-noise ratio. Group Number of Tag Calls > -3 Year Tag ID Composition Escorts Duration dB SNR 2010 mn10_267a FCME 2 1.66 18 2010 mn10_272a FCME 2 4.13 1109 2010 mn10_278a FCE 1 2.84 86 2010 mn10_280a FC 0 1.95 24 2010 mn10_290a FC 0 1.91 8 2010 mn10_291a FC 0 1.05 2 2010 mn10_296a FCE 1 4.12 132 2010 mn10_297a FC 0 1.92 63 2011 mn11_259a FC 0 2.53 12 2011 mn11_259b FC 0 1.80 8 2011 mn11_267a FC 0 0.35 1 2011 mn11_273a FC 0 3.52 26 2011 mn11_273b FCE 0-1 2.53 106 2011 mn11_280a FCE 0-1 3.70 141 2011 mn11_285a FCE 0-1 3.20 727 2011 mn11_289b FC 0 2.71 8 2011 mn11_289c FCE 1 1.80 51 2011 mn11_298c FC 0 4.14 20 2011 mn11_301a FC 0 4.09 22 2014 mn14_269a FC 0 0.30 5 2014 mn14_280a FC 0 1.50 11 2014 mn14_289a FCE 1 2.87 152 2014 mn14_290a FCE 1 0.73 47 2014 mn14_293b FC 0 1.43 16

166

Appendices

2014 mn14_296a FC 0 2.93 4 2017 mn17_283a FCME 0-3 2.23 230

Group Cluster Dur FMIN FMAX FP 1 0.25±0.17 389.9±321.8 2448.8±1216.4 607.8±426.6 FC 2 0.34±0.22 141.5±124.2 2274.2±873.0 154.4±72.5 1 0.23±0.21 529.2±472.1 4702.8±4552.5 855.2±674.1 FCE 2 0.35±0.32 78.8±82.4 3095.8±2463.9 159.4±97.0 1 0.19±0.30 700.0±696.6 4337.3±3359.4 1018.0±771.9 FCME 2 0.28±0.49 201.7±104.9 2661.9±2260.2 317.1±143.3 3 0.41±0.61 60.9±50.2 2243.6±1533.9 127.2±51.4

Group Cluster FC FQ1 FQ3 FIQ 1 677.8±411.0 565.1±387.4 828.2±476.1 263.2±238.9 FC 2 188.0±65.6 140.2±58.9 255.7±95.6 115.5±80.0 1 939.4±660.7 774.9±578.2 1185.3±798.7 410.3±445.8 FCE 2 193.5±94.5 144.9±87.6 257.9±107.6 113.1±56.0 1 1144.1±790.8 949.6±742.9 1423.6±847.8 474.0±394.7 FCME 2 358.8±125.0 297.0±119.5 447.5±144.9 150.6±104.0 3 156.4±44.1 116.8±40.6 210.5±70.5 93.7±60.8

Group Cluster FC05 FC95 FIC FTREND 1 450.3±345.4 1129.5±628.1 679.1±522.0 1.47±1.45 FC 2 100.0±58.5 445.2±311.1 345.2±317.6 1.47±1.17 1 620.7±509.7 1728.3±1133.3 1107.6±964.8 1.28±1.49 FCE 2 99.1±83.1 453.4±257.2 354.3±244.3 2.06±5.64 1 794.1±717.5 2063.8±896.6 1269.7±662.0 1.20±0.87 FCME 2 237.1±111.6 708.8±280.4 471.8±284.5 1.19±0.95 3 80.1±43.8 344.9±165.8 264.9±169.6 1.88±4.39

167

Appendices

Appendix 2: Quantifying the number of call types

Table A2.1 Random Forest (RF) confusion matrix, with the call types from aural-visual (AV) classification in the first columns and the RF distribution of each call type in the remaining columns. The last column is the classification error for each call type, providing a quantitative assessment of how well the RF categorisation agreed with AV analysis. Bop Crow Eeaw Grumble Grunt Meow Moan Paired Croaks Bop 491 0 0 0 0 0 0 0 Crow 0 6 3 3 0 0 2 0 Eeaw 0 0 82 2 0 1 3 0 Grumble 1 1 3 61 0 0 7 4 Grunt 3 0 1 1 49 0 0 1 Meow 9 0 2 0 0 148 0 0 Moan 2 0 5 3 0 0 25 1 Paired Croaks 0 0 2 4 0 0 1 33 Knock 47 0 0 0 0 0 0 0 Snort 38 0 1 8 1 2 3 0 Spiccato 0 2 1 2 0 0 3 2 Squawk 23 0 4 2 0 1 0 0 Squeak 21 0 4 0 0 1 0 0 Thwop 0 0 2 2 0 0 0 2 Wop 1 0 1 5 0 1 0 2 Wup 47 0 4 0 1 1 4 0 Knock Snort Spiccato Squawk Squeak Thwop Wop Wup Error Bop 24 26 0 4 8 0 0 12 0.13 Crow 0 0 2 0 2 0 2 0 0.70 Eeaw 0 3 3 3 2 0 0 4 0.20 Grumble 0 15 1 3 2 2 3 1 0.41 Grunt 0 12 0 0 0 0 0 3 0.30 Meow 0 3 0 0 2 0 0 1 0.10 Moan 0 9 5 0 1 0 1 3 0.55 PCroaks 0 1 0 0 0 1 1 0 0.23 Knock 70 25 0 0 3 0 0 4 0.53 Snort 12 301 0 13 8 1 0 28 0.28

168

Appendices

Spiccato 0 1 21 1 1 2 4 0 0.48 Squawk 0 13 0 42 12 0 0 4 0.58 Squeak 1 8 1 7 135 0 0 6 0.27 Thwop 0 0 3 0 0 4 10 0 0.83 Wop 0 2 2 0 0 5 25 0 0.43 Wup 1 37 0 2 4 0 0 193 0.34

Table A2.2 Results of the fuzzy cluster analysis, indicating how many of each call type from the aural-visual analysis were assigned to each cluster. A call was assigned to the cluster in which it had the largest typicality coefficient. Cluster Call Type 1 2 3 4 5 6 7 8 9 10 Bop 22 83 26 2 47 12 20 152 89 112 Crow 0 5 8 1 0 0 5 1 0 0 Eeaw 1 41 8 2 9 6 24 1 0 11 Grumble 11 23 29 15 3 1 10 5 4 3 Grunt 18 13 27 1 2 1 3 1 1 3 Moan 5 12 13 0 1 0 3 18 0 3 Knock 41 16 20 31 1 2 2 29 1 6 Snort 134 49 45 96 17 6 16 27 7 19 Squawk 2 13 4 0 17 16 20 1 11 17 Squeak 0 5 6 1 13 65 23 4 59 8 Wup 14 127 34 3 18 4 14 34 10 36

169

Appendices

Table A2.3 Summary of the acoustic parameters (mean ± SD) by call type. The description was based on the aural-visual sound types most prevalent in the cluster and/or the defining features of the calls. N (% Total Aggregate Minimum Maximum Minimum Call Type Description Duration Calls) Entropy Time Time Frequency Cluster 1 Low entropy snort/knock 248 (10.4%) 4.14 ± 0.69 0.18 ± 0.13 0.07 ± 0.09 0.10 ± 0.06 85.1 ± 46.4 Cluster 2 Wup/low frequency eeaw 387 (16.2%) 4.98 ± 0.73 0.26 ± 0.20 0.10 ± 0.15 0.16 ± 0.14 227.0 ± 96.0 Cluster 3 Long snort/grumble 220 (9.3%) 4.63 ± 0.74 0.32 ± 0.31 0.15 ± 0.27 0.18 ± 0.18 146.7 ± 66.7 Cluster 4 Discrete snort 152 (6.4%) 4.31 ± 0.79 0.22 ± 0.29 0.08 ± 0.13 0.12 ± 0.15 1.54 ± 2.08 Cluster 5 Broadband bop 128 (5.4%) 5.57 ± 0.60 0.18 ± 0.13 0.07 ± 0.10 0.11 ± 0.10 451.7 ± 184.3 Cluster 6 High frequency squeak 113 (4.8%) 6.03 ± 1.01 0.20 ± 0.14 0.08 ± 0.11 0.11 ± 0.08 1418.3 ± 810.2 Cluster 7 Squeak/high frequency eeaw 140 (6.0%) 6.10 ± 0.93 0.35 ± 0.27 0.10 ± 0.16 0.24 ± 0.21 665.4 ± 305.3 Cluster 8 Low entropy bop 273 (11.5%) 3.74 ± 0.69 0.14 ± 0.08 0.05 ±0.06 0.09 ± 0.07 273.5 ± 109.6 Cluster 9 High frequency bop 182 (7.7%) 4.24 ± 0.74 0.14 ± 0.07 0.05 ± 0.06 0.08 ± 0.05 703.5 ± 366.8 Cluster 10 High entropy bop 218 (9.2%) 5.08 ± 0.66 0.15 ± 0.08 0.06 ± 0.07 0.09 ± 0.06 288.7 ± 102.7 Meow Meow 165 (6.9%) 5.53 ± 1.02 0.24 ± 0.10 0.09 ± 0.09 0.14 ± 0.09 365.5 ± 335.1 Paired croaks Paired croaks 43 (1.8%) 3.64 ± 0.90 0.40 ± 0.05 0.14 ± 0.12 0.24 ± 0.09 61.7 ± 55.3 Spiccato Spiccato 40 (1.7%) 4.47 ± 0.83 1.78 ± 1.61 0.73 ± 1.01 0.95 ± 0.78 148.8 ± 126.8 Thwop Thwop 23 (0.09%) 4.30 ± 0.71 0.68 ± 0.24 0.17 ± 0.14 0.47 ± 0.27 46.0 ± 52.8 Wop Wop 44 (1.8%) 4.66 ± 0.63 0.79 ± 0.64 0.25 ± 0.33 0.48 ± 0.31 65.8 ± 53.4 Peak Peak Frequency Peak Centre Inter-quartile Call type Maximum Frequency Frequency 1st Frequency 3rd Trend Frequency Frequency Bandwidth Quartile Time Quartile Time Cluster 1 2173.0 ± 1269.4 138.8 ± 39.9 174.4 ± 42.0 106.0 ± 62.8 163.4 ± 68.7 129.2 ± 68.7 1.48 ± 1.01 Cluster 2 4326.4 ± 4027.2 345.7 ± 110.8 403.5 ± 102.5 181.3 ± 144.0 320.5 ± 120.1 319.9 ± 120.1 1.17 ± 0.92 Cluster 3 3195.6 ± 1923.9 215.2 ± 64.0 259.6 ± 58.4 141.7 ± 109.4 213.8 ± 101.0 201.8 ± 101.0 1.24 ± 1.09 Cluster 4 3419.7 ± 3281.1 121.0 ± 59.4 135.4 ± 55.5 113.0 ± 90.1 162.8 ± 164.8 130.9 ± 164.8 2.27 ± 5.24 Cluster 5 4318.7 ± 2844.3 823.9 ± 286.8 906.1 ± 229.2 497.8 ± 341.8 706.3 ± 303.6 724.3 ± 303.6 1.21 ± 0.83

170

Appendices

Cluster 6 6229.0 ± 4206.7 2336.7 ± 735.4 2334.4 ± 660.1 537.1 ± 430.4 2144.0 ± 743.3 2387.1 ± 743.3 0.97 ± 0.45 Cluster 7 5772.3 ± 5534.1 1158.9 ± 521.8 1333.5 ± 467.8 868.5 ± 531.5 1163.8 ± 590.6 1104.8 ± 590.6 1.64 ± 2.89 Cluster 8 1209.4 ± 694.0 417.0 ± 135.0 428.3 ± 120.5 106.7 ± 61.5 374.2 ± 130.3 331.7 ± 130.3 1.25 ± 0.55 Cluster 9 2785.3 ± 1988.0 990.5 ± 351.5 1008.6 ± 340.8 155.4 ± 77.2 891.8 ± 277.8 937.0 ± 277.8 1.01 ± 0.35 Cluster 10 4142.0 ± 2982.7 496.4 ± 142.5 563.1 ± 121.4 294.0 ± 205.5 488.7 ± 243.8 473.7 ± 243.8 1.27 ± 0.99 Meow 5048.7 ± 2760.8 583.7 ± 506.1 624.6 ± 523.6 286.6 ± 377.1 624.8 ± 535.5 575.9 ± 535.5 1.22 ± 0.65 Paired croaks 1759.4 ± 995.4 114.6 ± 50.4 139.1 ± 45.0 74.9 ± 30.3 105.9 ± 43.8 113.0 ± 43.8 2.63 ± 8.06 Spiccato 3552.5 ± 3142.2 223.2 ± 176.1 283.2 ± 186.3 150.9 ± 118.0 182.0 ± 174.9 214.1 ± 174.9 1.59 ± 4.16 Thwop 3571.7 ± 1796.3 132.2 ± 90.4 157.4 ± 77.6 101.8 ± 53.7 118.6 ± 71.0 141.7 ± 71.0 2.97 ± 9.20 Wop 3106.9 ± 2504.2 154.9 ± 150.9 187.8 ± 120.8 125.7 ± 112.6 142.2 ± 147.5 161.2 ± 147.5 1.76 ± 4.90

171

Appendix 3: A link between social and communicative complexity

Figure A3.1 Spectrograms of two representative calls from each cluster: (a) cluster one, ‘low entropy snort/knock’, (b) cluster two, ‘wup/low frequency eeaw’, (c) cluster three, ‘grumble/long snort’, (d) cluster four, ‘discrete snort’, (e) cluster five, ‘broadband bop’, (f) cluster six, ‘high

172 Appendices frequency squeak’, (g) cluster seven, ‘squeak/high frequency eeaw’, (h) cluster eight, ‘low entropy bop’, (i) cluster nine, ‘high frequency bop’, and (j) cluster ten, ‘high entropy bop’.

Table A3.1 Summary of the 24 focal follows used in analysis. The number of calls excludes calls with a signal-to-noise ratio < 5 dB and without accompanying behavioural data. No. of No. of Time Hours of No. of Year Tag ID Escorts Bins Data Calls 2010 mn10_267a 2 10 1.7 11 2010 mn10_272a 2 22 3.7 684 2010 mn10_278a 1 12 2 37 2010 mn10_280a 0 11 1.8 21 2010 mn10_291a 0 6 1 2 2010 mn10_296a 1 22 3.7 42 2010 mn10_297a 0 11 1.8 29 2011 mn11_259a 0 15 2.5 11 2011 mn11_259b 0 10 1.7 7 2011 mn11_267a 0 2 0.3 1 2011 mn11_273a 0 21 3.5 11 2011 mn11_273b 0-1 11 1.8 56 2011 mn11_280a 0-1 18 3 31 2011 mn11_285a 0-1 12 2 220 2011 mn11_289b 0 14 2.3 1 2011 mn11_289c 1 17 2.8 29 2011 mn11_298c 0 17 2.8 3 2011 mn11_301a 0 23 3.8 8 2014 mn14_280a 0 8 1.3 2 2014 mn14_289a 1 8 1.3 45 2014 mn14_290a 1 16 2.7 26 2014 mn14_293b 0 8 1.3 1 2014 mn14_296a 0 17 2.8 1 2017 mn17_283a 1-4 10 1.7 592

173

Appendices

Table A3.2 Comparison of the coefficient of variation (CV) within each call type for the four acoustic parameters across each social state. A significant result (p < 0.05) from the modified signed-likelihood ratio test is indicated by a * next to the parameter, with the highest and lowest CVs indicated in bold. ‘Meows’ were only detected once in the ‘low social complexity’ state, so no CV is available. D: discrete, G: graded. Social State Call Type Parameter Test statistic FC FCE Unstable Competitive MSLRT=4.50 Duration 48.86 62.21 44.99 71.87 p=0.2125 Peak MSLRT=5.59 Low 33.59 30.71 15.46 27.54 frequency p=0.1334 entropy Inter- snort/knock MSLRT=13.15 quartile 110.93 44.95 49.89 57.65 (G) p=0.0043* frequency Aggregate MSLRT=39.94 4.46 14.69 11.53 19.34 Entropy p<0.0001* MSLRT=10.04 Duration 45.34 85.62 76.73 81.21 p=0.0183* Peak MSLRT=3.18 33.02 31.51 24.69 28.44 Wup/low frequency p=0.3648 frequency Inter- MSLRT=2.40 eeaw (G) quartile 61.46 77.72 77.15 83.26 p=0.4942 frequency Aggregate MSLRT=12.31 8.62 13.55 15.52 14.54 Entropy p=0.0056* MSLRT=6.60 Duration 53.71 121.43 89.40 92.01 p=0.0858 Peak MSLRT=5.91 36.65 29.25 21.16 31.00 Grumble/ frequency p=0.1163 long snort Inter- MSLRT=1.94 (G) quartile 63.28 75.79 60.19 78.04 p=0.5856 frequency Aggregate MSLRT=27.09 4.87 15.54 16.79 16.85 Entropy p<0.0001* MSLRT=12.84 Duration NA 65.79 NA 153.07 p=0.0003* Peak MSLRT=21.46 NA 21.84 NA 44.80 frequency p<0.0001* Discrete Inter- snort (D) MSLRT=26.80 quartile NA 38.95 NA 106.32 p<0.0001* frequency Aggregate MSLRT=2.95 NA 15.26 NA 19.16 Entropy p=0.0857

174

Appendices

MSLRT=1.80 Duration 61.72 68.32 44.99 68.60 p=0.6147 Peak MSLRT=2.60 53.32 30.10 34.77 34.10 frequency p=0.4580 Broadband Inter- bop (G) MSLRT=2.75 quartile 69.74 38.09 54.75 51.69 p=0.4312 frequency Aggregate MSLRT=4.37 8.99 9.92 15.70 9.42 Entropy p=0.2238 MSLRT=2.72 Duration 33.36 42.67 68.93 49.07 p=0.4362 Peak MSLRT=1.53 19.36 27.25 28.99 32.81 High frequency p=0.6766 frequency Inter- MSLRT=2.09 squeak (G) quartile 26.90 74.46 92.82 81.70 p=0.5537 frequency Aggregate MSLRT=4.34 8.74 19.76 12.20 14.66 Entropy p=0.2270 MSLRT=3.33 Duration 46.78 85.03 56.60 95.70 p=0.3439 Peak MSLRT=9.00 69.72 53.81 39.86 27.35 Squeak/high frequency p=0.0292* frequency Inter- MSLRT=0.60 eeaw (G) quartile 62.62 66.16 69.07 56.90 p=0.8968 frequency Aggregate MSLRT=7.23 2.82 12.56 16.68 14.53 Entropy p=0.0650 MSLRT=1.09 Duration 50.94 62.66 54.00 57.60 p=0.7790 Peak MSLRT=5.58 30.34 25.97 31.02 33.94 Low frequency p=0.1337 entropy bop Inter- MSLRT=4.66 (G) quartile 43.28 65.36 66.03 52.21 p=0.1987 frequency Aggregate MSLRT=6.99 11.66 17.02 19.57 18.31 Entropy p=0.0722 MSLRT=2.34 Duration 32.37 49.70 59.19 46.01 p=0.5055 High Peak MSLRT=5.99 16.15 34.73 33.10 33.30 frequency frequency p=0.1123 bop (G) Inter- MSLRT=1.08 quartile 48.18 53.31 56.93 44.86 p=0.7824 frequency

175

Appendices

Aggregate MSLRT=0.37 16.74 18.40 16.22 18.10 Entropy p=0.9472 MSLRT=1.30 Duration 42.12 47.43 54.19 52.80 p=0.7295 Peak MSLRT=6.57 20.70 24.25 37.50 31.15 High frequency p=0.0870 entropy bop Inter- MSLRT=1.07 (G) quartile 54.92 71.43 73.85 71.74 p=0.7833 frequency Aggregate MSLRT=3.76 9.91 12.33 17.43 12.77 Entropy p=0.2885 MSLRT=43.31 Duration NA 31.66 53.67 39.36 p<0.0001* Peak MSLRT=31.07 NA 83.58 66.26 85.80 frequency p<0.0001* Meow (D) Inter- MSLRT=16.24 quartile NA 111.41 143.88 128.23 p=0.0003* frequency Aggregate MSLRT=41.17 NA 16.48 18.55 17.51 Entropy p<0.0001* MSLRT=0.14 Duration NA 17.63 NA 11.51 p=0.7083 Peak MSLRT=0.60 NA 31.74 NA 45.40 frequency p=0.4374 Paired Inter- Croaks (D) MSLRT=1.22 quartile NA 18.59 NA 34.93 p=0.269 frequency Aggregate MSLRT=5.37 NA 3.27 NA 20.82 Entropy p=0.0205* Duration NA NA NA NA 90.68 Peak NA NA NA NA 78.86 frequency Spiccato Inter- (D) quartile NA NA NA NA 78.19 frequency Aggregate NA NA NA NA 18.51 Entropy MSLRT=1.17 Duration NA 42.59 NA 25.81 p=0.2800 Thwop (D) Peak MSLRT=1.16 NA 67.86 NA 40.11 frequency p=0.2806

176

Appendices

Inter- MSLRT=0.03 quartile NA 17.89 NA 19.38 p=0.8678 frequency Aggregate MSLRT=0.71 NA 18.69 NA 13.47 Entropy p=0.401 MSLRT=5.59 Duration 35.03 44.07 37.31 104.08 p=0.1335 Peak MSLRT=3.93 19.42 54.02 36.86 64.26 frequency p=0.2697 Wop (D) Inter- MSLRT=4.66 quartile 26.82 31.84 23.87 55.65 p=0.1988 frequency Aggregate MSLRT=2.53 0.00 18.51 8.96 12.10 Entropy p=0.2827

Appendix 4: Discrete and graded calls during intraspecific conflict

Table A4.1 Focal follow information for the four years of data collection on the Great Barrier Reef. A: adult, E: escort, FC: female-calf pair, J: juvenile, Intensity Level 1: low intensity, Intensity Level 2: moderate intensity, Intensity Level 3: high intensity. Time Time of Recording Intensity Date Group Calls On Follow Time Level 21-Aug-16 FC9E 17:16 00:34:00 00:15:08 298 2 22-Aug-16 6A 15:36 01:49:00 00:19:05 14 1-2 27-Aug-16 FC2E 08:52 00:24:00 00:05:45 8 1 27-Aug-16 5A 09:49 00:56:00 00:11:04 108 1 29-Aug-16 5A 06:50 01:07:00 01:01:50 65 1-2 17-Jul-17 7A 08:34 01:39:00 00:22:20 366 1-2 18-Jul-17 4A-4A1J 10:51 00:50:00 00:19:49 165 1-2 18-Jul-17 5A2J 15:19 00:58:00 00:14:44 22 1-2 19-Jul-17 3A 10:08 01:01:00 00:22:35 0 1 19-Jul-17 5A 15:20 00:21:00 00:13:06 116 1 23-Jul-17 3-5A 15:37 01:01:00 00:26:15 88 3 31-Jul-17 5A 11:48 00:54:00 00:31:04 495 2 1-Aug-17 4A 08:45 01:29:00 00:13:44 0 1 3-Sep-17 FC2E 13:40 00:20:00 00:15:15 1 1 3-Sep-17 6A 16:15 01:04:00 00:28:19 82 1 6-Sep-17 8A 10:00 02:50:00 00:57:54 252 2-3

177

Appendices

11-Sep-17 FC3E 12:47 00:55:00 00:37:33 62 2-3 17-Jul-18 8A 12:55 01:03:00 00:24:12 26 1 18-Jul-18 FC4E 10:38 00:46:00 00:22:20 52 1-2 19-Jul-18 4A 09:39 01:41:00 01:01:15 593 2 20-Jul-18 5-9A 12:35 01:36:00 00:56:03 60 1-2 26-Aug-18 4A 17:10 00:15:00 00:12:29 24 2 28-Aug-18 5A 14:42 01:43:00 00:37:32 220 2 30-Aug-18 3A 11:42 01:33:00 00:27:11 1 1-2 31-Aug-18 6A1J 11:04 01:13:00 00:35:35 327 1 2-Sep-18 11A 16:52 00:36:00 00:32:14 28 2 3-Sep-18 6A 08:44 01:00:00 00:29:15 57 2 6-Sep-18 FC3E 15:41 00:38:00 00:17:23 45 2 7-Sep-18 12A4J 10:00 01:33:00 00:56:19 610 3 18-Aug-19 3-5A 13:51 01:06:00 00:26:53 63 2 26-Aug-19 3-4A1J 11:00 02:07:00 00:35:04 12 2-3 26-Aug-19 3A1J 09:40 01:59:00 00:58:14 77 1-3 27-Aug-19 FC2E 09:50 00:25:00 00:04:14 0 1 27-Aug-19 5A 12:29 01:39:00 00:43:13 133 2 27-Aug-19 7A 15:41 01:39:00 01:01:51 100 2-3 27-Aug-19 5A 14:46 01:51:00 00:55:46 102 2 28-Aug-19 4A 09:19 01:48:00 00:05:14 2 2 2-Sep-19 5A-5A1J 14:10 02:05:00 00:35:32 80 2 2-Sep-19 6A 16:32 01:01:00 00:23:01 485 1-2 3-Sep-19 FC3E 09:15 01:15:00 00:22:06 222 2 3-Sep-19 6A 12:58 01:26:00 00:16:23 21 1 4-Sep-19 4A 10:35 00:56:00 00:04:24 7 1 5-Sep-19 3A1J 10:57 00:50:00 00:18:50 4 1-2

178

Appendix 5: Animal ethics approval

179 Appendices

180

Appendices

181

Appendices

182

Appendices

183

Appendices

184

Appendices

185

Appendices

186