Improving Wildlife Detection Dog Team Selection and Training

Improving wildlife detection dog team selection and training

La Toya Jamieson

B.App.Sc. (Hons)

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2019

School of Agriculture and Food Sciences Abstract

Wildlife can only be properly managed when populations are accurately monitored. Commonly used monitoring methods, including camera-trapping and visual surveys, are often costly, labour intensive, with low detection rates. To address these issues wildlife detection dogs are increasingly being used in ecological research. These dogs non-invasively locate live individuals, their scats, carcasses, and denning/nesting sites. The success of this method is dependent most notably on the dog and handler, and their training. Whilst incorrectly selecting dogs and handlers is costly and a welfare concern, selection is often based on personal preference rather than scientific evidence. Working dog selection remains focused on breeding programs that are financially expensive with highly varied success. Certain breeds are therefore commonly excluded during selection. Selecting unsuitable individuals, or incorrectly managing these teams, will not only reduce team performance but may also tarnish wildlife detection dogs’ reputation. There is currently minimal research on the selection, training and management of wildlife detection dog teams, especially in Australia. Given wildlife detection dogs have unique working requirements research on other working dog fields is often not comparable. Thus, to investigate factors important to detection dog and handler selection and management, I trained 12 dogs from three breeds at detection work, experimentally assessing their training times and odour discrimination ability.

After reviewing the literature three breeds were selected. The breed with the greatest number of suitable behavioural and physical characteristics for wildlife detection (Border Collies); the breed with the least number of suitable characteristics (Greyhounds); and the breed used most commonly for detection work (Labrador Retrievers). These dogs were trained to detect Bengal Tiger (Panthera tigris tigris) scat as it was a novel odour which they would not encounter outside training. Training sessions were filmed to determine the time required to achieve specific training competencies, and behaviour coded to record smelling times and behaviours related to the dogs’ true and false indications. Once the dogs achieved all training competencies their odour discrimination ability was assessed during single-blind trials, with both a familiar and unfamiliar handler.

All Border Collies and Labrador Retrievers, and one Greyhound, completed training. Overall the Border Collies had the quickest training times and the highest accuracy scores. Individual variation was, however, significant within the breeds’ training times and accuracy. During training the dogs’ smelling times were significant factors influencing their indications, with specific behaviours (e.g. paw-lifting) being correlated more often with true, rather than false, positives. The only Greyhound to complete training had higher accuracy scores than half of the Labrador Retrievers during testing. There was therefore a weak correlation between the dogs’ training times and detection accuracy.

La Toya Jamieson 2

During testing the dogs had significantly higher accuracy scores when handled by their familiar handler. With the unfamiliar handler the dogs performed significantly more stress-related behaviours and were distracted for a higher proportion of time, which was negatively correlated to detection accuracy.

Important dog handler traits and skills were also determined through emailing questionnaires to Australian and New Zealand wildlife detection dog handlers. These questionnaires asked the handlers to complete personality assessments and rate handler skills based on importance for wildlife detection work. The handlers shared similar mean personality scores, however, these scores had large ranges. Handlers rated skills specific to their dog, such as understanding dog body language, as highly important for field success.

Individual variation was prominent in all major findings. Due to the large range in the dog handlers’ personality scores, personality may not be as important as their training or dog–handler relationship. The large variation within the breeds training times and accuracy further suggests that a dog’s breed may not be the best predictor of their trainability or detection aptitude. These dogs’ accuracy was further impacted by changing handlers. Future research is required to determine if professional dogs are impacted similarly, and the best ways to manage dog-handler transitions. Lastly my research demonstrated that dogs’ smelling times and their associated behaviours can assist handlers discriminate between dogs’ true and false indications.

My research challenges how working dogs are currently globally managed. Due to the level of individual variation among dogs suitable for working roles, dogs should not be excluded purely because of their breed. Individual team’s performances must also continue to be evaluated due to the highly site-specific nature of their effectiveness. Management strategies must also take into consideration how influential the dog-handler relationship is on team performance. Prior to my study no research had investigated how detrimental changing a dog’s handler is on their welfare and performance. It is therefore crucial to continue challenging and advancing best practises, not only for animal welfare but also for the success of the working dog industry. Continuing research on wildlife detection dogs, including best avenues to source dogs, is crucial for this emerging method and will ensure the greatest outcomes are achieved.

La Toya Jamieson 3

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.

La Toya Jamieson 4

Publications included in this thesis

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2018, ‘Who’s a good handler? Important skills and personality profiles of wildlife detection dog handlers’, Animals, vol. 8, pp. 222-36.

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2018, ‘You are not my handler! Impact of changing handlers on dogs’ behaviours and detection performance’, Animals, vol. 8, pp. 176-87.

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2017, ‘Identifying suitable detection dogs’, Applied Animal Behaviour Science, vol. 195, pp. 1–7.

Submitted manuscripts included in this thesis

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2019, ‘Benefits and limitations of wildlife detection dogs’, Wildlife Society Bulletin (Submitted April 2019 – under review).

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2019, ‘Can handlers be their dog’s lie detectors? Don’t believe those puppy eyes’, Applied Animal Behaviour Science (Re-submitted June 2019 – under review).

Other publications during candidature

No other publications.

Contributions by others to the thesis

No contributions by others.

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.

La Toya Jamieson 5

Research Involving Human or Animal Subjects Below are the ethics approval numbers for all chapters involving human or animal subjects. Ethics approval certificates are included in the Appendices. Chapter Ethics committee Approval number 4 University of Queensland’s Human Ethics committee 2016001089 5, 6, 7 University of Queensland’s Production and Companion SAFS/454/16 Animals committee

La Toya Jamieson 6

Acknowledgements

This thesis could not have been completed without the help of my PhD village. Thank you to my PhD supervisors, Associate Professor Peter Murray and Dr Greg Baxter, for your constant support and invaluable advice. Not many people would believe a Greyhound could be trained to detect tigers, but you always encouraged me to continue with my crazy plans (not that I ever actually gave you a choice). Thank you for also encouraging me to bring a never-ending menagerie of dogs to our meetings – they wouldn’t have been productive otherwise.

Thank you to Anne Russell for being my second dog handler during the odour discrimination testing. Whilst my dogs loved you, they did give you hell, so thank you for your patience and persistence. Thank you to all of my University of Queensland student volunteers who assisted me with training and testing the research dogs. Without your assistance this project wouldn’t have happened. Thank you to all of the dog owners who lent their dogs to me for this project and for trusting me to care for them for 3 months. Thank you to the University of Queensland’s Clinical Studies Centre staff for their constant help and support. Special thank you to: Samantha Kempster for providing me excellent advice and calming me down when faecal matter was hitting the fan; to Donna Marchiori for your invaluable help during all behaviour assessments and throughout training; and to Nikky Pitkin and Martina Smith for your help and support, and keeping me sane working seven days a week for nine months, looking after 12 dogs. A massive thank you to all the people who didn’t believe ex-racing Greyhounds were capable of (or worth) being trained for detection work – all you did was encourage me further.

Thank you to Amanda Hancock (Quebec Delta 1) for always supporting and encouraging my crazy ideas, and providing me with the inspiration and motivation when I needed it the most. If it wasn’t for you, I wouldn’t have even started a PhD. Thank you to Delma Greenway for providing her assistance with the statistics. Thank you to the wildlife detection dog handlers who completed the questionnaires and shared their stories with me - you are all true inspirations.

Thank you to my friends and family (of both blood and bond) for their support over what has been three very challenging, yet rewarding years. Thank you to my parents for your love and support, and for accepting that to love me you have to also love my research dogs. Thank you to my PhD Queens (Meg ‘Nerd’ Edwards and Tamielle ‘Sunny’ Brunt) for giving me Guinness, burgers, caramel slices and advice whenever I needed it, and for believing in me when I didn’t. Thank you to Ebony Reeves for always questioning everything (in the very best way), for supporting me when I really just needed a slap, and for loving my research just as much as me. Thank you to Ally Wood

La Toya Jamieson 7 for “basically being one of my limbs” and for flying to America with me just for moral support at my first international conference.

Thank you to the real heroes of my thesis – my research dogs. Thank you to “Jasper” for being a sweetheart and the perfect student – you were my little superstar. Thank you to “Rosie” for challenging me every day, but still scoring perfectly during testing. Thank you to “Lyla” for your cheekiness and for being almost as challenging as your daughter, “Rosie”. Thank you to my “Bailey” bear for being the most gorgeous, least motivated dog I’ve ever had the pleasure of training. Thank you to “Shaz” for having the energy, motivation and behavioural quirks of 10 dogs. Thank you to “Burky” for eventually trusting me that the tiger poo wasn’t going to bite you, and becoming my best Labrador. Thank you to my darling “Pozer” for always ‘protecting me’ from everyone when I was sick and teaching me to give Labradors treats like they are sharks. Thank you to “Zulu” for being such a sweetheart who always gave her best. Thank you to “Bacon” for being the most lovable, biggest drama queen Greyhound ever and making me constantly adapt my training approach. Thank you to “Olaf” for always surprising me and showing that Greyhounds can be highly motivated. Thank you to “Con” for being more interested in jumping into water fountains than any type of training or reward. Lastly thank you to my crazy, darling, superstar Greyhound “Marcie” who was not only the first Greyhound to complete training, but also challenged breed stereotypes, scoring higher than some of my Border Collies and Labradors. All 12 of you have a special place in my heart and taught me more than I could have ever hoped to teach you.

“Marcie” and the author (L), and “Pozer” and the author (R) – real scientists are covered in dogs, not lab coats.

La Toya Jamieson 8

Financial support

This research was supported by an Australian Government Research Training Program Scholarship. Funding totalling $1,000 was received from the Australian Geographic Society. No other financial support was provided to fund this research.

Keywords

Wildlife detection dogs, training, dog breeds, dog handler, behaviour, personality.

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 060801, Animal Behaviour, 50%

ANZSRC code: 070201, Animal Management, 40%

ANZSRC code: 050206, Environmental Monitoring, 10%

Fields of Research (FoR) Classification

FoR code: 0608, Zoology, 80%

FoR code: 0501, Ecological Applications, 20%

La Toya Jamieson 9

Table of Contents LIST OF TABLES ...... 13 LIST OF FIGURES ...... 15 ABBREVIATIONS ...... 15 CHAPTER 1: INTRODUCTION ...... 16

BACKGROUND INFORMATION ...... 16 RESEARCH AIMS ...... 17 STUDY DESIGN ...... 18 STRUCTURE OF THESIS...... 19 CHAPTER 2: BENEFITS AND LIMITATIONS OF WILDLIFE DETECTION DOGS ...... 21

ABSTRACT ...... 22 INTRODUCTION ...... 22 Who’s a good dog? ...... 23 Both ends of the leash ...... 24 DETECTION ACCURACY AND SUCCESS ...... 24 WORKING EFFICIENCY – COMPARISON OF SURVEY METHODOLOGIES ...... 26 LIMITATIONS ...... 32 Cost ...... 32 False indications and survey bias ...... 33 Environmental conditions ...... 34 CONCLUSIONS ...... 35 CHAPTER 3: IDENTIFYING SUITABLE DETECTION DOGS ...... 36

ABSTRACT ...... 37 INTRODUCTION ...... 37 DETECTION DOG TRAITS ...... 38 Ideal detection dog traits ...... 38 Undesirable detection dog traits ...... 40 SELECTING DETECTION DOGS ...... 42 COMMONLY USED DOG BREEDS FOR DETECTION WORK ...... 43 Variation between breeds...... 43 Breed performance and comparisons ...... 46 Influence of sex and neutering ...... 47 Is breed specific selection enough? ...... 48 THE IMPORTANCE OF THE DOG HANDLER ...... 49 CONCLUSIONS AND RECOMMENDATIONS ...... 49 CHAPTER 4: WHO’S A GOOD HANDLER? IMPORTANT SKILLS AND PERSONALITY PROFILES OF WILDLIFE DETECTION DOG HANDLERS ...... 51

ABSTRACT ...... 52 INTRODUCTION ...... 52 METHODS ...... 54 Questionnaire ...... 54 Questionnaire Distribution ...... 54

La Toya Jamieson 10

Data Analyses ...... 55 RESULTS ...... 55 General Handler Information ...... 55 Important Handler Characteristics and Knowledge...... 57 Personality Profiles ...... 58 DISCUSSION ...... 60 CONCLUSIONS ...... 63 ACKNOWLEDGMENTS ...... 63 CHAPTER 5: CAN SIGHTHOUNDS SCENT? DOG BREEDS’ (CANIS LUPUS FAMILIARIS) TRAINABILITY AT SCENT DETECTION ...... 64

ABSTRACT ...... 65 INTRODUCTION ...... 65 METHODS ...... 67 Selected breeds and dogs ...... 67 Kennelling ...... 68 Behaviour assessments ...... 68 Training ...... 68 Odour discrimination testing ...... 70 Data analyses ...... 71 RESULTS ...... 72 DISCUSSION ...... 75 ACKNOWLEDGEMENTS ...... 78 CHAPTER 6: YOU ARE NOT MY HANDLER! IMPACT OF CHANGING HANDLERS ON DOGS’ BEHAVIOURS AND DETECTION PERFORMANCE ...... 79

ABSTRACT ...... 80 INTRODUCTION ...... 80 METHODS ...... 82 Research Dogs ...... 82 Dog Handlers ...... 83 Testing Layout ...... 83 Testing Procedures ...... 84 Behaviour Coding ...... 85 Data Analyses ...... 85 RESULTS ...... 86 DISCUSSION ...... 88 CONCLUSIONS ...... 90 ACKNOWLEDGMENTS ...... 90 CHAPTER 7: CAN HANDLERS BE THEIR DOG’S LIE DETECTORS? DON’T BELIEVE THOSE PUPPY EYES ...... 91

ABSTRACT ...... 92 INTRODUCTION ...... 92 METHODS ...... 94 Research dogs ...... 94 Training ...... 95 Behaviour coding ...... 95

La Toya Jamieson 11

Data analyses ...... 96 RESULTS ...... 97 DISCUSSION ...... 101 ACKNOWLEDGEMENTS ...... 103 CHAPTER 8: GENERAL DISCUSSION AND CONCLUSIONS ...... 104

RESEARCH AIMS AND DISCUSSION ...... 104 Aim 1 – Wildlife detection dog benefits and suitable individuals ...... 105 Aim 2 – Important dog handler traits ...... 105 Aim 3 – Breed comparison ...... 107 Aim 4 – Impact of changing handlers...... 108 Aim 5 – Smelling times and behaviours ...... 108 CONCLUSIONS ...... 109 REFERENCES ...... 111 APPENDICES ...... 133

La Toya Jamieson 12

List of Tables Table 2.1. The reported accuracy (including search success) of wildlife detection dogs in the published literature………………………………………………………………………………….25

Table 2.2. Efficiency and accuracy of wildlife detection dogs compared against other environmental survey methods. X indicates the method(s) which dog surveys were not evaluated against. ……………………………………………………………………………………………...29

Table 3.1. Dog types and their common behavioural attributes……………………………………45

Table 4.1. Native and pest species, from Australia and New Zealand, listed as target species for the wildlife detection dogs. Total handlers currently handling these species-specific detection dogs are also listed.…………………………………………………………………………………………...56

Table 4.2. Scoring of skills and knowledge based on their importance and relevance for wildlife detection dog handlers (1: strongly disagree it is important, 5: strongly agree it is important; SD: standard deviation). …………………………………………………………………………...……57

Table 4.3. The dog handlers’ personality scores in the five main personality domains (in bold), and the traits (shown below each domain) within these domains. The standard deviation and range is also provided.……………………………………………………………………………………….59

Table 5.1. The dogs used in this research. …………………………………………………………67

Table 5.2. The list of competencies that the dogs’ training was shaped around…………………...70

Table 5.3. Behaviour assessment scores using the Match-Up II Shelter Dog Rehoming Program/Behavior Evaluation. ……………………………………………………………………..72

Table 5.4. The number of reinforcements (R), training minutes (T) and training sessions (S) required for the dogs to achieve the training competencies………………………………………...73

Table 5.5. The dogs’ total number of reinforcements (R), training minutes (T), and training sessions (S) required to achieve all training competencies…………………………………………………..74

Table 5.6. The dogs’ odour discrimination accuracy scores from testing.…………………………75

Table 6.1. Detection dogs used during this project. Due to their extended training time, these nine dogs were separated into three groups: March–June (Group 1), June–September (Group 2), and September–November 2017 (Group 3)……………………………………………………………..83

La Toya Jamieson 13

Table 6.2. Sensitivity, specificity, PPV and NPV scores with Handler 1 and Handler 2. Mean scores and standard deviations (SD) are also provided. Dogs 4, 5, and 8 did not work for Handler 2 as demonstrated by their scores of zero for sensitivity and PPV………………………………………86

Table 6.3. The dogs’ behaviour assessment scores for Friendliness (score/23), Excitability (score/23), Playfulness (score/17), Fearfulness (score/24), Aggressiveness (score/24), and Trainability (score/15). The dogs who performed well for both handlers are in bold……………...88

Table 7.1. The research dogs and their relevant information………………………………………95

Table 7.2. Ethogram used for behaviour coding……………………………………………………96

Table 7.3. Behaviours (B) performed during or directly after smelling the samples. P is the percentage of these behaviours that were completed for each indication type (e.g. True positive or true negative)………………………………………………………………………………………100

La Toya Jamieson 14

List of Figures Figure 6.1. Strongly correlated negative relationship between the proportion of time (%) spent ‘distracted’ and the dogs’ sensitivity scores with Handler 1 and 2……………….………………..87

Figure 6.2. Strongly correlated relationship between the proportion of time (%) spent ‘scenting’ and the dogs’ sensitivity scores with Handler 1 and 2……………………………………………...88

Figure 7.1. Frequency of smelling times for true and false positives, and true and false negatives.97

Figure 7.2. Mean smelling times for true and false positives, and true and false negatives……….98

Figure 7.3. The dogs’ mean times for completing true and false positive indications. The bars that are a mixture of both shades (blue and pink) demonstrate both true positive and false positive indication time frequencies………………………………………………………………………….99

Abbreviations Abbreviations Meaning PPV Positive Predictive Value NPV Negative Predictive Value GLM Generalised linear models DOC Department of Conservation IPIP-NEO-120 International Personality Item Pool-NEO SD Standard deviation BORIS Behavioral Observation Research Interactive Software

La Toya Jamieson 15

Chapter 1: Introduction

Background information To develop wildlife management strategies, population estimates must be made (Fryxell et al. 2014; DeMatteo et al. 2018). Locating and collecting the number of biological samples required to determine this information is challenging (Orkin et al. 2016), particularly for cryptic and endangered wildlife. A range of monitoring techniques are available to collect this information. Human visual surveys for wildlife traces; camera- and cage-trapping; hair snares; auditory species surveys; and aerial and ground based surveys are often labour intensive, time and resource consuming, with potentially high error rates (e.g. false positives and false negatives) and low detection rates (Krausman 2002; Krebs 2006; Long et al. 2007a; De Bondi et al. 2010; Watkins et al. 2010; Duggan et al. 2011; Nadeau & Conway 2012; Cristescu et al. 2015; Meek et al. 2015). This reduces a survey’s accuracy and efficiency, and therefore the validity of the results. If inaccurate information is used to develop management programs this can have catastrophic effects on wildlife management efforts, especially endangered species conservation. An alternative survey method would therefore be beneficial, especially for endangered and cryptic species (Homan et al. 2001; Dahlgren et al. 2012).

Dogs have historically been used for detection tasks by humans, with their exceptional odour discrimination ability leading to their use in diverse working fields (Hurt & Smith 2009; Johnen et al. 2015). We rely on dogs to ensure our airports are safe; our sporting events are drug-free; vulnerable members of society are cared for; and our borders are safe from pests, carried in food or mail. Dogs are continually being trained for new and novel working fields. Wildlife detection dogs are an emerging method for wildlife population monitoring, having high accuracy and detection rates (Arandjelovic et al. 2015; Cristescu et al. 2015). Detection dogs locate individual wildlife (terrestrial and marine), their biological samples, denning/nesting sites and carcasses (Rolland et al. 2006; Hurt & Smith 2009). Wildlife detection dogs are frequently reported to significantly outperform other survey methods, such as having 5-20 times higher detection rates than camera- trapping surveys (Vynne et al. 2011). Detection dogs are also efficient, with some dogs reportedly surveying 26 ha/hour (Reindl-Thompson et al. 2006). This allows for large areas to be rapidly surveyed which minimizes disruption to wildlife (Duggan et al. 2011). Detection dogs’ noses are also more sensitive than current DNA technology and artificial detection systems (Beckman 2005; Shelby et al. 2006; Bomers et al. 2012; Horvath et al. 2013). Detection dogs are therefore highly important, not only for wildlife detection but also many other human-relevant fields (e.g. narcotics,

La Toya Jamieson 16 cadaver, explosives and biomedical detection), which warrants research to improve their management and therefore success.

Whilst the benefits of wildlife detection dogs are extensively listed in the literature, survey success is highly dependent on several factors. These include: the dog and handler selected (Rebmann et al. 2000; Svartberg & Forkman 2002; Sinn et al. 2010); the dog-handler relationship (Hoummady et al. 2016); the team’s training (Hurt et al. 2016) and field surveying strategy (Glen & Veltman 2018). A variety of working dog industries, including guide dogs and narcotics detection dogs, have reported that less than 50% of their purposely bred dogs became professional working dogs (Wilsson & Sundgren 1997; Slabbert & Odendaal 1999; Maejima et al. 2007; Batt et al. 2008a, b; Sinn et al. 2010). Not only is this a major concern for animal welfare, considering the already overpopulated domestic dog population, but this is not financially efficient (Cobb et al. 2015). These low success rates indicate a problem not only in the dogs being selected, but also with the handlers they are paired with and their management.

Professional working dog teams may further be impacted if the dog-handler relationship isn’t seen as a priority, or if the handlers are not properly educated about dog behaviour and humane training methods (Lefebvre et al. 2007). These factors are especially concerning given certain military organisations class newly paired detection dog teams as ‘operative’ after working two weeks together (Haverbeke et al. 2010). Research is therefore crucial to determine the best individuals for wildlife detection work, both dog and handler, and the most appropriate training and management schemes. As wildlife detection dogs are an emerging field there are currently limited studies which have explored these issues. Determining qualities and skills that are crucial for dog and handler performance will improve selection, regardless of where the dogs are sourced (e.g. breeders or rescue organisations). This information is important for this field and will ensure this method is feasible and effective, with high standard welfare practises.

Research aims My thesis presents research which explores dog and handler selection, including the influence of a dog’s breed; the impact changing a dog’s handler has on their detection performance and welfare; and the impact dogs’ smelling behaviours has on their detection accuracy.

La Toya Jamieson 17

My research aims are:

Aim 1 Collate the published literature to determine the benefits and limitations of wildlife detection dogs, and determine ideal characteristics, both physical and behavioural, of detection dogs (Chapters 2 and 3).

Aim 2 Experimentally determine important wildlife detection dog handler characteristics, personality traits and knowledge (Chapter 4).

Aim 3 Experimentally determine the variation in dog breeds odour discrimination training success and detection accuracy (Chapter 5).

Aim 4 Experimentally compare dogs’ detection performances and stress-related behaviours when handled by a familiar and unfamiliar handler (Chapter 6).

Aim 5 Experimentally determine behaviours and measure smelling times detection dogs perform prior to completing true and false positives, and true and false negatives (Chapter 7).

Study design Previous research has touched on some aspects of the research aims, but only peripherally. That research is limited in several ways with crucial aspects either not being included or thoroughly explored. I have therefore addressed these issues when planning my research design.

The first issue is the lack of time dedicated to individuals during dog breed comparison research. Certain studies have eliminated dogs from projects, possibly prematurely, due to the dogs not progressing at the desired or anticipated pace. Potentially important data has therefore not been collected, which may have impacted their final conclusions. Whilst this may be due to financial or time constraints, this does not provide a total picture of the dogs’ or breeds’ capabilities. Therefore, for all my studies, individual dogs were allowed an extensive training period (based on what is currently stated in the literature), with the results reflecting their whole learning progression and not simply their initial drive or learning aptitude.

The second issue is the lack of evidence regarding where important dog handler skills and knowledge are sourced. Authors rarely reveal where this information is derived, or their own experience if it is their personal opinion. Dog handlers, particularly wildlife detection dog handlers, rarely have their voices heard in the literature, especially regarding skills crucial for their work. To

La Toya Jamieson 18 collect this information, I contacted active and current wildlife detection dog handlers directly and not only provided a list of skills which the literature stated were important, but also allowed them to provide their own list and opinions.

Lastly, whilst there are studies which have explored the level of attachment between a dog and their handler, few studies have practical outcomes or focus. Previous studies have instead focused on the behavioural changes of the dogs in different scenarios with different people. I instead explored the behavioural and performance impact of changing a dog’s handler during detection tasks, which is a transition often required for working dogs. This research is therefore very applied, rather than purely exploratory.

Structure of thesis My thesis is comprised of an introduction chapter, two review chapters, four data-based chapters, and a discussion and conclusion chapter.

Chapter 1 Provides background information specific to my research focus (wildlife detection dogs and their management), the research aims, study design and thesis structure.

Chapter 2 Explores the benefits and limitations of the wildlife detection dog methodology, and compares it against alternative survey methods. This review chapter is currently under review by Wildlife Society Bulletin.

Chapter 3 Determines the characteristics of suitable wildlife detection dogs and compares the success of different dog breeds at detection work. The impact of the dog handler is also discussed. This review chapter has been published (Jamieson, LJ, Baxter, GS & Murray, PJ 2017, ‘Identifying suitable detection dogs’, Applied Animal Behaviour Science, vol. 195, pp. 1–7).

Chapter 4 Determines knowledge and skills which detection dog handlers believe are important for dog handlers to possess. Handler personality assessments were also completed to determine if they shared similar personality scores. This data-based chapter has been published (Jamieson, LJ, Baxter, GS & Murray, PJ 2018, ‘Who’s a good handler? Important skills and personality profiles of wildlife detection dog handlers’, Animals, vol. 8, pp. 222-36).

Chapter 5 Compares three dog breeds’ detection training times and evaluates their detection accuracy. Behaviour assessments were also completed to

La Toya Jamieson 19

determine if specific individuals were trained quicker or had higher detection accuracy scores.

Chapter 6 Compares dogs’ detection performances and stress-related behaviours when handled by a familiar and unfamiliar handler. This data-based chapter has been published (Jamieson, LJ, Baxter, GS & Murray, PJ 2018, ‘You are not my handler! Impact of changing handlers on dogs’ behaviours and detection performance’, Animals, vol. 8, pp. 176-87).

Chapter 7 Measures dogs’ smelling and indication times during various detection tasks and records their behaviours. This data-based chapter is currently under review by Applied Animal Behaviour Science who requested it to be re- submitted with modifications.

Chapter 8 Discusses how the project aims were achieved and the overall results. The importance of the research findings are highlighted, and future research ideas are proposed.

La Toya Jamieson 20

Chapter 2: Benefits and limitations of wildlife detection dogs

Contributor Statement of contribution

Conceived/designed review (100%) La Toya Jamieson Collated literature (100%) Wrote the article (90%)

Greg Baxter Wrote/edited the article (5%) Peter Murray Wrote/edited the article (5%)

Quoll detection dog “Sparky”

La Toya Jamieson 21

Abstract Wildlife management requires accurate monitoring of animal numbers and population trends. This is often challenging, especially when target species are cryptic and/or in low-density populations. Whilst there are a range of monitoring techniques available, some methods can be labour intensive with high error rates and low detection rates. Wildlife detection dogs trained to safely locate wildlife and their signs are more frequently being used for population monitoring. As with any method, wildlife detection dogs have both benefits and limitations which must be recognised. Through collating the literature, this paper examines the benefits and limitations of wildlife detection dogs and compares this against other methods. Overall, detection dogs have high detection rates and are more efficient than other survey methods, including camera- and cage- trapping and human visual surveys. The use of detection dogs can be limited due to high survey costs or extreme environmental conditions. These limitations are manageable and their impact can be minimised. The literature has highlighted the efficiency and accuracy of detection dogs, and their application versatility. The benefits of this method typically outweigh its limitations, and is proven to be highly effective, especially for endangered or cryptic species. The possible future uses of detection dogs for conservation has only begun to be explored and undoubtedly will only continue to evolve and improve.

Introduction To develop a suitable wildlife management strategy population estimates and an understanding of habitat use must be determined (Fryxell et al. 2014; DeMatteo et al. 2018). Locating the number of biological samples required to determine this information is a limitation of species monitoring and wildlife management (Orkin et al. 2016). Wildlife monitoring aims to maximise the survey area, whilst minimising cost, time and resources required (Reed et al. 2011). A range of monitoring techniques are available to collect ecological information, with all having benefits and limitations (Hurt & Smith 2009; DeMatteo et al. 2018). Human visual surveys for scat, carcasses and evidence of burrows/dens; camera- and cage-trapping; hair snares; auditory species surveys; and aerial surveys are often labour intensive, time and resource consuming, with potentially high error rates (e.g. false positives and false negatives) and low detection rates (Krausman 2002; Krebs 2006; Long et al. 2007a; De Bondi et al. 2010; Watkins et al. 2010; Duggan et al. 2011; Nadeau & Conway 2012; Cristescu et al. 2015; Meek et al. 2015). These deficiencies reduce a survey’s accuracy (Arandjelovic et al. 2015) and therefore the validity of the results. An alternative survey method, especially for endangered species, would therefore be beneficial (Homan et al. 2001; Dahlgren et al. 2012).

La Toya Jamieson 22

Novel methods for monitoring wildlife are beginning to be explored, such as microchip-automated technology (Edwards et al. 2018). Wildlife detection dogs are dogs used for locating scat, live animals, plants, denning or nesting sites to determine species presence and distribution (Hurt & Smith 2009). Detection dogs have been successfully used globally for ecological surveys, including in Canada (Wasser et al. 2004), the United States of America (Cablk & Heaton 2006; Kapfer et al. 2012), Australia (Cristescu et al. 2015; Leigh & Dominick 2015), New Zealand (Browne et al. 2015), Argentina (DeMatteo et al. 2009), Brazil (Vynne et al. 2011), and China (Orkin et al. 2016). Detection dogs are particularly beneficial for cryptic, low density or endangered species surveys (Wasser et al. 2004; DeMatteo et al. 2009; Orkin et al. 2016), and have many benefits including: higher detection rates than alternative monitoring methods (Arandjelovic et al. 2015); the ability to survey significantly more area, in less time, than alternative methods (Homan et al. 2001; Smith et al. 2001; Nussear et al. 2008; Cristescu et al. 2015); and the ability to work in challenging environments where alternative methods are not suitable (Oliveira et al. 2012). As a result, wildlife detection dogs have been reported to be more efficient and accurate than alternative methods for many species (Harrison et al. 2002; Cablk & Heaton 2006; Long et al. 2007a).

A wildlife detection dog team’s performance, however, is dependent on many factors including: the dog selected (Svartberg & Forkman 2002; Sinn et al. 2010; Jamieson et al. 2017), their handler (Gutzwiller 1990; Rebmann et al. 2000; Hurt & Smith 2009; Jamieson et al. 2018a,b), and the weather conditions they are worked in (Wright & Thomson 2005; Greatbatch et al. 2015). As a result, wildlife detection dog success is likely specific to the individual working team, their survey location and environment, and time/season.

Who’s a good dog? Whilst wildlife detection dogs are similar in many aspects to other detection dogs; their unique working requirements demands that they must also pose no threat to wildlife. To maximise the likelihood of wildlife detection dog success the correct dog must firstly be selected and properly trained. Traits which are important wildlife detection dog characteristics are: a medium sized, agile and highly athletic individual that is highly play/food motivated, yet poorly motivated to chase other animals; and is highly intelligent and independent, yet obedient working off-leash (Jamieson et al. 2017). Individuals meeting these criteria can be found in a variety of breeds (Jamieson et al. 2017). Certain breeds such as Labrador Retrievers, German Shepherds and English Springer Spaniels, however, are typically favoured for detection work (Rooney & Bradshaw 2004; Maejima et al. 2007; Jezierski et al. 2014). The preference of dog breeds for working purposes is often a

La Toya Jamieson 23 personal choice, however, rather than scientific justification (Hall et al. 2015; Minhinnick et al. 2016).

Both ends of the leash The detection dog is only half the team. The dog handler, and the dog-handler relationship, play a crucial role in the team’s success (Jamieson et al. 2018b). The dog handler will ideally possess certain traits and knowledge to maximise team compatibility and working outcomes. Some of these traits include knowledge of dog behaviour and body language, and training principles; trust in their dogs’ behaviours, especially working indications; strong work ethic; knowledgeable in target species ecology; and ability to read and understand influential environmental conditions, such as wind direction (Rebmann et al. 2000; Jamieson et al. 2018a). A successful and compatible wildlife detection dog team is therefore created from the careful selection of both dog and handler, and the continued training and evaluation of their working performance (Hurt & Smith 2009; Jamieson et al. 2017; 2018a).

Whilst the success of wildlife detection dogs is influenced by dog and handler selection, collating information on their uses, success and limitations may reveal ways to improve this methodology or areas which require further research. Whilst Beebe et al. (2016) has previously reviewed certain wildlife detection dog studies, their review more specifically examined how the detection dog itself affects survey success. Our review paper, however, aims to: (1) collate the published literature on the use and success of wildlife detection dogs and compare this against alternative survey methods; (2) determine factors that impact wildlife detection dog team success (not limited to the detection dog and handler selected); and (3) highlight the limitations of this methodology and strategies to minimise these issues.

Detection accuracy and success Wildlife detection dogs are required to discriminate between target and non-target wildlife samples (Oldenburg et al. 2016). This skill is exceptionally important for the accuracy and efficacy of a field survey. Investigating an area where a dog has falsely indicated (false positive) is fatiguing for both the dog and handler (Greatbatch et al. 2015), and a waste of time and resources. It is therefore desirable that false positives occur as infrequently as possible (Greatbatch et al. 2015). Quantifying the dog-handler success rate is therefore important. It is also crucial to governing agencies and potential employers (Helton 2009) and allows confidence in the dogs’ indications. Whilst their success varies, overall the literature reports the dogs’ high detection accuracy across multiple countries and for a variety of species (Table 2.1). The variation in survey locations and target species also highlights the versatility of this methodology.

La Toya Jamieson 24

Table 2.1. The reported accuracy (including search success) of wildlife detection dogs in the published literature.

Samples Target Location Accuracy Reference located Brown tree snake (Boiga Live Savidge et al. Northern Guam 35% irregularis) animals (2011) Allen’s lappet-browed bat (Idionycteris phyllotis), Coconino National Chambers et al. Arizona myotis (Myotis Forest, northern 29-77% Bat roosts (2015) occultus) and Long-legged Arizonna, USA myotis (M. volans) 40% of radio- Purua, New Live Robertson & Kiwi (Apteryx spp.) collared Zealand animals Fraser (2009) individuals Maned wolf (Chrysocyon brachyurus), puma (Puma concolor), jaguar (Panthera 60-85% Emas National Vynne et al. onca), giant anteater (depending on Scats Park, Brazil (2011) (Myrmecophaga tridactyla), year) and giant armadillo (Priodontes maximus) Spotted Knapweed Gallatin County, Goodwin et al. 81.1% Live plants (Centaurea stoebe) Montana, USA (2010) Black-footed ferret (Mustela South Dakota, Live Reindl-Thompson 82% nigripes) USA animals et al. (2006) 83% detection Illinois, Indiana, Franklin’s ground squirrel probability Live Duggan et al. Iowa, Missouri and (Poliocitellus franklinii) with two dog animals (2011) Wisconsin, USA teams Norway rats (Rattus Maungatautari, 83.5% search Live norvegicus) and house mice North Island, New Gsell et al. (2010) success animals (Mus musculus) Zealand 87% (98% Scats Lazovsky State Amur tigers (Panthera tigris during (matching Kerley & Salkina Nature altaica) repeated-trial to (2007) Zapovednik, Russia tests) individuals) Bobcats (Lynx rufus) New Mexico, USA 88% Scats Harrison (2006) 91% success - animals on Desert tortoises (Gopherus Las Vegas, surface; and Live Cablk & Heaton agassizii) Nevada, USA 90% success - animals (2006) animals in burrows 91% mean success (100% Live Eastern indigo snake Georgia and Stevenson et al. skin sheds and animals and (Drymarchon couperi) Florida, USA (2010) 81% live skin sheds snakes)

La Toya Jamieson 25

Western black crested gibbon (Nomascus concolor), Stump- 92% (81% if tailed macaque (Macaca unidentifiable Orkin et al. Yunnan, China Scats arctoides) and Indochinese samples are (2016) gray langur (Trachypithecus included) crepusculus) Koala (Phascolarctos Queensland, Cristescu et al. 97% Scats cinereus) Australia (2015) Black bears (Ursus Centennial americanus), grizzly bears Mountains, Idaho- (Ursus arctos horribilis), 98.6% Scats Beckman (2005) Montana border, cougars (Puma concolor) and USA gray wolves (Canis lupus) 100% accurate in discrimination Spotted-tailed quoll New South Wales, Leigh & trials. 83-87% Scats (Dasyurus maculatus) Australia Dominick (2015) detection rate depending on habitat. Live Red Palm Weevil animals in Nakash et al. Israel 100% (Rhynchophorus ferrugineus) infested (2000) trees Carrizo Plain San Joaquin kit fox (Vulpes National Smith et al. 100%* Scats macrotis mutica) Monument, (2001) California, USA Carrizo Plain National Monument and Smith et al. San Joaquin kit fox 100%* Scats LoKern Natural (2003) Area, California, USA *All scat collected was target scat

Even when accuracy is reported as 100% it is still possible that errors were made. Detection dogs may have falsely indicated to a sample, however, these were correctly interpreted by their handler as a false positive and therefore were not recorded. Detection dogs are therefore not perfect machines and it is unreasonable to expect them to consistently work with 100% accuracy for prolonged periods under adverse environmental conditions (Greatbatch et al. 2015). Measures can, however, be implemented to maximise their likelihood of success.

Working efficiency – Comparison of survey methodologies For wildlife detection dogs to be a viable method for wildlife monitoring they not only need to be highly accurate and efficient, but they must be at least as successful as other methods (Cablk &

La Toya Jamieson 26

Heaton 2006). Typically when compared against alternative survey methods detection dogs are the most accurate and/or efficient method (Table 2.2). For example, detection dogs had a mean detection probability of 20.1%, whilst vocal surveys had a detection probability of 7.3% for Northern Spotted (Strix occidentalis caurina) and Barred Owls (S. varia; Wasser et al. 2012). Scat detection dogs also had 5-20 times higher detection rate success for Maned wolves (Chrysocyon brachyurus), pumas (Puma concolor), giant anteaters (Myrmecophaga tridactyla), and giant armadillos (Priodontes maximus; Vynne et al. 2011), than the previously reported camera-trapping study (Silveira et al. 2003). Of these collected scat samples, 71% were not located near roads, and therefore would unlikely be collected by human observers (Vynne et al. 2011). Previous studies have also reported that dogs effectively locate samples missed during human visual surveys (Hagell 2010). Rat detection dogs have previously been reported to locate 33% of targets within 10 m from the walking transect (Glen et al. 2018), whilst carnivore scat detection dogs have located greater than 75% of scats within the same distance (Reed et al. 2011). Detection dogs have therefore been estimated to have an effective sweep width of 16.8 m from a stationary target (e.g. dead rodent; Glen et al. 2018). These detection distances, however, are highly impacted by the environmental conditions (Reed et al. 2011).

Another major benefit of wildlife detection dogs is their ability to rapidly survey an area (Duggan et al. 2011). For example, black-footed ferret detection dogs surveyed on average 26 ha/hour (Reindl- Thompson et al. 2006). Detection dogs also cover more area than their handler. Carnivore detection dogs reportedly covered 2.1 times greater distance than humans (Beckman 2005). This example illustrates another advantage of detection dogs; they allow ready calculation of the area surveyed which other methods, such as camera- and cage-traps, do not. Rodent detection dogs have also reportedly located single rodents when other methods were unsuccessful (e.g. trapping and bait stations), thereby ensuring an area was rodent free (Russell et al. 2008). Detection dogs are also capable of surveying areas where other methods would not be suitable, due to the risk of equipment theft or non-target interference (Glen et al. 2016) or habitat complexity (e.g. log piles and rocky slopes). Dog-handler teams also typically only require one site visit, which is an important consideration when long driving distances and time constraints are involved (Long et al. 2007a). Whilst dogs can disturb wildlife (Langston et al. 2007), detection dog survey speed means wildlife species aren’t disrupted for long time periods, and no animals, such as potential predators, need to be attracted to the area (Duggan et al. 2011). The use of detection dogs also eliminates the need for wildlife to be trapped and handled, further reducing stress (Duggan et al. 2011).

La Toya Jamieson 27

Research suggests dogs’ noses are more sensitive than current DNA technology and man-made detection systems (Beckman 2005; Shelby et al. 2006; Bomers et al. 2012; Horvath et al. 2013). Wildlife detection dogs can locate samples that cannot be identified by DNA technology due to sample degradation (Beckman 2005; Orkin et al. 2016). More research therefore needs to be completed on what level of degradation or dilution a detection dog can locate a sample (Orkin et al. 2016). In the likely event that detection dogs are accurately locating samples which machines cannot identify, dogs should be the method of choice until these machines are developed to ensure they can collect valuable genetic information (Orkin et al. 2016). When scat samples are genetically analysed information including the individual’s’ species, sex, diet and parasite burden can be determined (Kohn & Wayne 1997; Mills et al. 2000; Elbroch & Evans 2012).

La Toya Jamieson 28

Table 2.2. Efficiency and accuracy of wildlife detection dogs compared against other environmental survey methods. -- indicates the method(s) which dog surveys were not evaluated against.

Detection Human Camera- Cage- Hair Target Reference dog surveys trapping trapping snares 4 x more effective at San Joaquin Smith et al. locating scat ------kit fox (2001) than human observers 8.4 – 61.9 – 67.7% Swift foxes 11.5% Harrison et al. search ------(Vulpes velox) trapping (2002) success success 10 x positive Bobcat -- 10 x less -- 10 x less Harrison (2006) detections Locate more Smith et al. Bobcat scat than ------(2006) humans North Atlantic right whales Located 97 Located 30 Rolland et al. ------(Eubalaena scats scats (2006) glacialis) 87% 33% 8% Long et al. Black bear detection -- detection -- detection (2007a) probability probability probability Fisher 84% 28% Long et al. (Martes detection -- detection -- -- (2007a) pennanti) probability probability 27% 13% Long et al. Bobcat detection -- detection -- -- (2007a) probability probability Searched five times the Desert Nussear et al. distance of ------Tortoises (2008) humans in 70% of time Spotted Knapweed 81.1% 58.9% Goodwin et al. ------(Centaurea accuracy accuracy (2010) stoebe) Spider monkey 59% 82% ------Hagell (2010) (Ateles accuracy* accuracy geoffroyi) Franklin’s Detection Probability Duggan et al. ground probability of -- -- of -- (2011) squirrel 44-59% detection

La Toya Jamieson 29

(increased to 61% 83% when (increased two dog- to 84% handler with two teams were trap days) used) Common quail 96% mean 12.3% mean Paula et al. carcasses ------accuracy accuracy (2011) (Coturnix coturnix) Detection Deer success of No samples Oliveira et al. (Mazama ------0.2 located (2012) spp.) samples/km Carcasses of multiple bat 75% species detection including 20% success, in < Mathews et al. Plecotus detection ------25% of time (2013) auratus, success taken by Pipistrellus humans pipistrellus, P. pygmaeus 7-8 weeks 2 days required to required to have 90% have 90% Clare et al. Bobcat -- detection -- -- detection (2015) probability probability (4 (4 km2 km2 area) area) 153% more accurate and 19 x more Cristescu et al. Koala ------efficient than (2015) human surveys 0.54 0.45 Cat (Felis detection detection Glen et al. ------catus) probability probability (2016) per search per search Western black 92% 74% crested detection detection gibbon, accuracy accuracy Orkin et al. Stump-tailed (81% if (45% if ------(2016) macaque and unidentifiable unidentifiable Indochinese samples are samples are gray langur included) included)

La Toya Jamieson 30

Detected Detected Eurasian lynx Hollerbach et target in -- target in -- -- (Lynx lynx) al. (2018) 21/44 sites 7/44 sites *Several trials completed during a rainstorm and the dog was ill. If compromised trials are removed accuracy increases to 64-92%

Survey design and objectives, and the study area, including vegetation, climate and topography, will impact detection dog effectiveness and determine which method is most appropriate for a study (Long et al. 2007a,b; Duggan et al. 2011). How effective a detection dog team is will also depend on the search strategy employed, with search methods being used in other detection fields often not being appropriate for wildlife detection (Glen & Veltman 2018).

The search pattern employed during detection dog surveys influences the search effort and probability of detection (Glen & Veltman 2018), thereby impacting overall survey effectiveness. Search patterns employed may follow a pre-determined path, such as grid lines (Gsell et al. 2010), or may be flexible depending on the terrain and environmental conditions (Glen & Veltman 2018). Searches can also be focused on specific areas where the target is likely to be (Brook et al. 2012; O’Kelly et al. 2012; Fuller et al. 2016). When comparing linear and diamond-shaped transects, and ‘opportunistic’ (ad hoc) search patterns for large carnivore scat, ‘opportunistic’ searches collected the most scats (Wultsch & Kelly 2012). Detection distances and probability of detection are also influenced by the target sample, such as the age and size of the scat samples (Wasser et al. 2004; Reed et al. 2011; Dahlgren et al. 2012), and vary between habitats (Smith et al. 2005). With skilled dog teams and favourable wind conditions, however, detection distances can be up to 60 m (Cablk et al. 2008; Gsell et al. 2010).

To ensure survey consistency the same dog teams should be used throughout projects (Dahlgren et al. 2012), with the dog team’s experience possibly being included as a covariate during analyses (Glen & Veltman 2018). The reported accuracy and efficiency of detection dogs is also specific to that survey or testing scenario, as different environmental conditions may have created different results. Detection dogs therefore may not always be the most suitable method depending on the target, research aims and survey conditions (including terrain and weather). Further research is needed to compare wildlife detection dogs against alternative survey methods. This will improve our understanding of projects which detection dogs are best suited for. Even if detection dogs are correctly selected for wildlife surveys, the success of this method is still limited by several factors which must be recognised and managed.

La Toya Jamieson 31

Limitations Whilst wildlife detection dogs are highly effective, no method is perfect, and their capabilities must not be over-stated (Cablk & Heaton 2006; Long et al. 2007b). Using detection dogs to the exclusion of trapping animals means that mark-recapture data, morphometric measures or blood/tissue samples cannot be collected (Duggan et al. 2011). Detection dogs can, however, be used in conjunction with trapping. Methodology selection depends on the data required; survey budget; climate; and the effectiveness of the currently used survey method. If an effective survey method is available and suitable, the costs and logistics associated with wildlife detection dog surveys may outweigh their benefits (Long et al. 2007b). This section will explore this methodology’s limitations and determine if and how they can be minimised.

Cost Due to the extensive training requirements of the dog-handler team, and the continued cost of dog maintenance and management, wildlife detection dog surveys are often financially expensive. This cost can be prohibitive for some organisations (Beckman 2005). For example, a two-month survey for Cross-river gorillas (Gorilla gorilla diehli) in Cameroon using detection dogs cost $98,000 USD (Arandjelovic et al. 2015). One-month detection dog surveys were also reported to cost 33% more than four-month camera trapping surveys, however, cameras required a significantly longer time to achieve high detection probabilities (90%, Clare et al. 2015). In some cases, financial costs have been reduced considerably. Research groups have independently trained their own dogs (Cristescu et al. 2015) or partnered with local dog trainers, such as the police (Orkin et al. 2016). In another example international primatologists partnered with the Chinese Ministry of Public Security to achieve surveys costs of under $3,000 USD for a project which ran over several months at multiple survey sites (Orkin et al. 2016).

It is important to highlight that dog-handler teams require extensive training and detection dogs are life-long commitments (Orkin et al. 2016). Therefore considerable thought and planning must be made prior to acquiring a dog for detection work, whether they are purchased trained or not. Care must also be taken when purchasing detection dogs to ensure they are compatible with their intended handler, as poor handler compatibility and unfamiliarity may reduce dogs’ detection performance and welfare (Hurt & Smith 2009; Jamieson et al. 2018b).

The slightly higher costs of using wildlife detection dogs are typically offset by their increased effectiveness and higher detection rates (Long et al. 2007a,b). For example, Duggan et al. (2011) reported that live-trapping in 62 sites cost $9,907 USD, whilst the detection dog surveys in the same sites cost $11,156 USD. The detection dog surveys, however, were completed in half the time and

La Toya Jamieson 32 only required one site visit, whilst live-trapping required multiple (Duggan et al. 2011). Detection dog surveys for bat roosts were also reported to have similar costs to radiotelemetry, however, for the dog surveys the bats didn’t require trapping or radio-tagging (Chambers et al. 2015). Similarly, detection dogs were also reported to have comparable costs to camera-trapping (3-week deployment of 80 cameras) when surveying nine search cells for feral cats (Camera-trapping = $5,100, detection dogs = $4,820; Glen et al. 2016). Camera-trapping costs, however, did not include the purchase of the camera traps, which totalled NZ$56,000 (Glen et al. 2016). This demonstrates that whilst the cost of wildlife detection dog surveys can be high, their benefits typically outweigh this expense. These price comparisons need to be completed for each individual project, as these costs will vary significantly depending on the research requirements, time of year and location.

False indications and survey bias Wildlife detection dogs can achieve a very high level of accuracy and detection success (Table 1). Despite the best training and handling however, false positives and negatives can occur. For example, during a survey for Eurasian lynx, the detection dogs falsely indicated to a large amount of non-target scat (42.6% of collected samples), predominantly fox scat (Hollerbach et al. 2018). DNA analysis showed some Eurasian lynx scats also contained fox DNA (Hollerbach et al. 2018) suggesting these scats were likely contaminated with fox urine (Wikenros et al. 2017). This is supported by other studies where DNA in urine altered the target scat’s genetic profile (DeMatteo et al. 2018). This could not have been determined visually by the dog handlers and therefore the dogs were inadvertently rewarded for locating fox DNA, potentially resulting in the large collection of fox scat (Hollerbach et al. 2018).

Other studies, however, have stated that surveys should focus on minimising targets missed, even at the expense of high false positive rates (Goodwin et al. 2010; Duggan et al. 2011). This may not always be feasible especially with the cost of genetic analyses and survey time (Hollerbach et al. 2018). As with any methodology, it is unlikely that detection dogs can locate every target within an area, especially when under time constraints (Smith et al. 2003). Wildlife detection dogs, however, have been reported to increase detection probability and success when compared to alternative methods (Smith et al. 2003). The potential of false positives and negatives should therefore not be a deterrent for using wildlife detection dogs, especially as detection dogs typically outperform the currently available alternative methods.

Scenting is taxing on detection dogs, especially when target odours are minimal, as is common for endangered species surveys (Papet & Minhinnick 2016). Survey efficacy can therefore be reduced due to loss of motivation and fatigue (Glen & Veltman 2018). To minimise the likelihood of dogs

La Toya Jamieson 33 completing false positives and increase search motivation, target samples can be provided in the field to allow the dogs to be rewarded (Wasser et al. 2004; Papet & Minhinnick 2016). This will also ensure the dogs are being rewarded for locating the correct target. It is imperative that these planted target samples are then removed from the field to not bias the survey results.

As with other survey methods, detection dog surveys aren’t free from bias. More research is needed on potential detection dog survey bias, such as the dogs age preference of target samples (i.e. fresher samples; Cablk & Heaton 2006). The dog handler can also introduce bias, such as through their selection of search strategies and survey sites (Glen & Veltman 2018), or level of experience (Wasser et al. 2004). Bias can also be introduced when compensating for challenging terrain or environmental conditions, such as through surveying easier areas (Arandjelovic et al. 2015), which will be discussed in the following section. Dogs may also behave and perform differently depending on their handler, which must be acknowledged if dogs have multiple handlers (Jamieson et al. 2018b). These potential biases must be considered when developing survey plans and analysing collected data.

Environmental conditions Environmental conditions including temperature, humidity, rainfall, wind speed and direction all influence scent movements in the air and therefore detection dog search performance (Wright & Thomson 2005; Reed et al. 2011; Greatbatch et al. 2015). Missed target samples may therefore be attributed to the absence of a constant scent and not a result of dog or handler error (Wright & Thomson 2005). As temperatures increase, evaporation and bacterial activity also increases, thereby intensifying odours of biological material (Wasser et al. 2004). This intensity is only short-lived, with higher temperatures eventually killing the bacteria and therefore scent production (Reed et al. 2011). During low temperatures bacterial activity is slowed and odours are less intense. Air temperature impacts dogs’ detection abilities differently, with certain dogs’ performances increasing as the temperature increases whilst others decrease (Reed et al. 2001).

Environmental conditions also impact detection dog health. A dog’s sniffing ability is related to its’ panting rate (Goldblatt et al. 2009). A dog that is panting heavily will have a reduced sniffing ability and may struggle to locate their target (Gazit & Terkel 2003). Whilst detection dogs have located targets in 29.9˚C heat (Nussear et al. 2008), limitations due to temperature have been suspected at 23˚C (Smith et al. 2003). Previous studies have also reported a reduction in dog’s performances if they are not allowed a sufficient acclimation period (recommended 2-week period, Hagell 2010). High temperatures should therefore be avoided where possible to maintain survey efficiency and animal welfare. The complexity of the survey terrain will also influence the dog’s

La Toya Jamieson 34 panting rate, thereby influencing their working ability. The dog’s handler, however, is arguably more impacted by terrain complexity than the dog itself (Arandjelovic et al. 2015). Detection dogs are less constrained by terrain complexity than methods which rely solely on humans accessing these areas (Arandjelovic et al. 2015).

Whilst the adverse effects of environmental conditions might not be avoidable during field work, measures can be put into place to minimise the level of impact. This includes working dogs during cooler parts of the day (e.g. early morning and late afternoon) and avoiding extreme temperatures (Smith et al. 2003). Wildlife detection dogs, when properly trained and acclimatised, however, can effectively work under a range of environmental conditions (Cablk & Heaton 2006; Reindl- Thompson et al. 2006; Long et al. 2007a,b; Reed et al. 2011) and habitat types (Leigh & Dominick 2015). It is therefore crucial that prior to beginning surveys all stakeholders understand the dogs’ limitations, and that their success is not over-stated, in order to avoid possible disappointment and develop appropriate survey plans to meet project aims.

Conclusions This review has highlighted the benefits of wildlife detection dogs, including their high detection rates and survey efficiency. Detection dogs were also demonstrated as the most accurate and efficient method for a variety of species during specific surveys. Whilst several limitations were explored, the majority could be managed including survey cost and environmental conditions. These limitations should therefore not necessarily prevent the use of this method. However, it does highlight the need to evaluate the success and effectiveness of dog-handler teams in each specific circumstance where they may operate. The possible future uses of detection dogs for conservation are almost unlimited and will certainly benefit wildlife monitoring and management. As this methodology evolves it is likely that these benefits, and our understanding of them, will increase. Wildlife managers and researchers should therefore continue exploring these more novel survey methods and not rely solely on traditional methods.

References

Complete reference list is provided at the end of the thesis due to reference repetition throughout the chapters.

La Toya Jamieson 35

Chapter 3: Identifying suitable detection dogs

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2017, ‘Identifying suitable detection dogs’, Applied Animal Behaviour Science, vol. 195, pp. 1–7.

Contributor Statement of contribution

Conceived/designed review (100%) La Toya Jamieson Collated literature (100%) Wrote the article (90%)

Greg Baxter Wrote/edited the article (5%) Peter Murray Wrote/edited the article (5%)

Research dogs “Jasper” (L) and “Rosie” (R)

La Toya Jamieson 36

Abstract Domestic dogs (Canis lupus familiaris) are versatile resources for humans due to a number of their physical and behavioural characteristics. Because of dogs’ olfactory acuity they have been used to detect cryptic or concealed items such as narcotics, explosives and wildlife. However, there is a wide variation in performance. This variation is often not correlated with their breed and has not been rigorously tested. Little research has compared dog breeds for their suitability as detection dogs, and even fewer studies have concluded which characteristics should be selected. This is important considering the number of dogs produced for detection work. This paper has collated the scientific literature to present important behavioural and physical traits, and traits which should be avoided, in detection dogs. The important traits include: highly play motivated; high level of cooperativeness with their handler; boldness; obedience yet independence when off-leash; and high athleticism. Although wildlife detection dogs are this paper’s focus, these proposed traits are relevant in any detection field.

Introduction Behavioural and performance differences between breeds of dogs (Canis lupus familiaris) can be a controversial topic (Fadel et al. 2016). Behavioural differences between breeds are often inappropriately generalised, however, they are a distinct group of genetic units (Ostrander & Wayne 2005; Clarke et al. 2013). Whilst it has been emphasised that each breed has specific behavioural characteristics, studies highlight the variation amongst individuals within breeds (Mehrkam & Wynne 2014). This variability is likely a result of the change in dog breeding priorities, from breeding for abilities to breeding for appearance (Mirkó et al. 2012). Domestic dogs have traditionally been utilised by humans due to their ability to perform specific working roles, including guarding, hunting, herding and detection (Rooney & Bradshaw 2004; Serpell & Duffy 2014). Dogs are a highly versatile detection tool and have been utilised in over 30 different tasks (Lorenzo et al. 2003; Hall et al. 2014). As detection dogs are a modern phenomenon no dog has been bred solely for this purpose (Rooney & Bradshaw 2004).

Both physical and behavioural traits are important when selecting working dogs (Coppinger & Coppinger 2001; McGarrity et al. 2016). The variation between dogs’ working performances can be attributed to behavioural differences, emphasising the importance of selecting a dog that is physically capable and behaviourally suited to detection work (Slabbert & Odendaal 1999; Svartberg & Forkman 2002; Rooney et al. 2007; Sinn et al. 2010). When selecting a detection dog certain behavioural and physical characteristics are typically desired. This has resulted in certain breeds being favoured for detection work. The current lack of breed comparative studies and the

La Toya Jamieson 37 variation within each breed provides a challenge when selecting a suitable detection dog (Rooney & Bradshaw 2004; Jezierski et al. 2014). If only individuals from one or two breeds are evaluated for their working potential, as is common in Military working dog programs (Moore et al. 2001; Sinn et al. 2010), this may restrict the detection program’s success. Breeders are not always aware of important characteristics of successful detection dogs, due to poor communication with the dog- handlers (Rocznik et al. 2015). Once a working program is established and traditions are formed, there is often minimal feedback received from the dog-handlers in relation to choosing breeding dogs (Rocznik et al. 2015). To maximise the efficiency of identifying suitable detection dog candidates, important physical and behavioural traits for specific detection tasks must be determined.

Wildlife detection dogs are a unique category of detection dogs trained to locate wildlife scats, carcasses or live animals (Hurt & Smith 2009). Whilst Beebe et al. (2016) have proposed certain important wildlife detection dog traits, commonly used dog breeds and their suitability for detection work has not been discussed. Irrespective of the breed selected, a dog’s temperament should always be gauged prior to choosing it for detection work (Graham & Gosling 2009; McGarrity et al. 2016).

Characteristics of the most suitable individual for detection work, with a focus on wildlife detection, will be explored here. Depending on dog breeders’ genetic selection criteria, breeds which were typically chosen for traditional traits and functions (e.g. tracking) may no longer possess such qualities (Adamkiewicz et al. 2013). Continually selecting the same dog breeds, without inspecting other breeds, may reduce the effectiveness of detection dog programs. This review will discuss the physical and behavioural characteristics of a suitable detection dog; commonly used dog breeds for detection, and the variation within these breeds.

Detection dog traits Ideal detection dog traits Most detection dogs to date have been herding, hunting or sporting breeds (Brownell & Marsolais 2002). A detection dog should be athletic and trainable, to ensure the dog is physically capable of completing the work, whilst also having desirable motivations (Brownell & Marsolais 2002). Differences between dog breed physical characteristics undoubtedly influence their skills and capabilities (Coppinger & Coppinger 2001). Behavioural traits have also been investigated to improve animal welfare, and dog training and management (Svartberg 2002; Clarke et al. 2013; McGarrity et al. 2016). There are multiple physical and behavioural traits which are important for detection dogs to possess. Choosing dogs with the following traits, gleaned from the literature, should increase both the dog’s suitability for detection work and their working performance.

La Toya Jamieson 38

Speed is important in any working dog field, ensuring working efficiency (Helton 2010). Detection dogs should work quickly, whilst not missing the intended targets nor exhausting themselves prematurely (Jezierski et al. 2014). In difficult terrain detection dogs should be agile, with exceptional stamina, allowing them to traverse the terrain (Rebmann et al. 2000; Hurt & Smith 2009). Medium-built dogs with suitably long legs are preferable, which is also beneficial if the dog becomes injured and needs to be carried (Hurt & Smith 2009). Medium-built dogs, with shorter coats, can also be advantageous for heat tolerance (Chesney 1997; Hurt & Smith 2009). Large dog breeds retain too much body heat and small dog breeds retain too little (Coppinger & Coppinger 2001). Heat-tolerant dogs are able to work more efficiently with fewer breaks, without the risk of overheating, which is not only detrimental to the dog’s working performance but can be fatal (Hurt & Smith 2009). The choice of a dog’s build and size is therefore a reflection of the dog’s working environment (Rebmann et al. 2000).

In the following sections the term ‘drive’ and its importance in relation to detection dogs will be discussed. A dog’s drive is their impulse or motivation to perform a behaviour or action (Brownell & Marsolais 2002). This concept is not current in behavioural science; however, it is widely and currently used in working dog science (Beebe et al. 2016; Minhinnick et al. 2016). Due to its importance in working dog science, the term ‘drive’ will be used. It should be highlighted that a dog’s ‘drive’ or motivation can be influenced by external factors (e.g. environmental), and therefore can be modified over time.

Detection dog handlers typically select working dogs who have strong motivational drives (Beebe et al. 2016). Motivators which are important during detection dog selection are play-, prey-, and hunt-drives (Maejima et al. 2007; Hurt & Smith 2009; Reed et al. 2011; Beebe et al. 2016; Minhinnick et al. 2016). A dog’s play-drive is the desire to be entertained, which ensures the dog values a toy or play reward in exchange for performing a particular behaviour (Cablk & Heaton 2006; Hurt & Smith 2009; Duggan et al. 2011). A detection dog will ideally be highly play motivated, to the point of obsession (Rebmann et al. 2000; Hurt & Smith 2009; Beebe et al. 2016; Minhinnick et al. 2016). This will ensure the dog is willing to perform hundreds of repetitions to receive their toy, which is crucial for training and work (Hurt & Smith 2009).

A dog’s desire to search is referred to as their hunt-drive and is important for sustaining motivation (Cablk & Heaton 2006; Hurt & Smith 2009). This motivation to search is crucial for dogs during surveys where the work is fatiguing and target odours are minimal (Cablk & Heaton 2006; Hurt & Smith 2009; McGarrity et al. 2016). A dog’s prey-drive is their desire to chase and kill (Hurt & Smith 2009; Minhinnick et al. 2016). Whilst this may not be problematic for a drug detection dog

La Toya Jamieson 39 which only works indoors, a wildlife detection dog with an uncontrolled high prey-drive can be catastrophic (Cablk & Heaton 2006; Hurt & Smith 2009; Beebe et al. 2016; Minhinnick et al. 2016). Ideally a detection dog would not be motivated to chase or kill another animal, which would minimise the risk to both wildlife and the dog (Cablk & Heaton 2006; Hurt & Smith 2009; Beebe et al. 2016; Minhinnick et al. 2016). The most suitable definitions and interpretations of these behavioural drives are, however, being challenged in the scientific literature (Minhinnick et al. 2016).

A detection dog must be able to work cooperatively with humans (also known as pack-drive), and follow both visual and auditory cues (Hurt & Smith 2009; Beebe et al. 2016). This ensures the dog is working efficiently and obediently in the field. Detection dogs should demonstrate minimal aggression to both humans and dogs, allowing for a peaceful home or kennel environment (Rooney & Bradshaw 2004). Whilst a dog should be willing to work with their handler, a detection dog should have a certain degree of independence when working (Rebmann et al. 2000; Rooney & Bradshaw 2004; Hurt & Smith 2009; Adamkiewicz et al. 2013). This enables them to make their own choices in the field when required. Trained dogs typically look to their handlers for guidance less than untrained dogs, which indicates independence and their problem-solving ability (Prato- Previde et al. 2008; Marshall-Pescini et al. 2009). Caution should be made when selecting an independent individual, with dogs possessing too much independence commonly becoming disobedient (Rebmannetal. 2000). Obedience off-leash, yet a certain degree of independence, is critical especially for the safety of both explosive and wildlife detection dogs (Rebmann et al. 2000; Hurt & Smith 2009; Adamkiewicz et al. 2013).

A dog’s ability to adapt to, and cope with, stress-producing stimuli within their environment are important working dog traits (Brownell & Marsolais 2002; Hurt & Smith 2009). This coping mechanism is crucial for detection dogs who are frequently exposed to a variety of visual, aural, olfactory and tactile environmental stimuli (Brownell & Marsolais 2002; Hurt & Smith 2009). This is related to not only its breed, but also their training, socialisation, early life experiences and environmental exposure (Brownell & Marsolais 2002; Hurt & Smith 2009).

Undesirable detection dog traits Whilst fear and anxious responses are both crucial for survival (Ohl et al. 2008), they are not ideal for detection dogs (van Rooy et al. 2014). Fearful dogs are undesirable due to the amount of stimuli in their working environment (Graham & Gosling 2009; Adamkiewicz et al. 2013). Typically dogs with poor concentration are also more anxious, providing further reason to reject individuals that

La Toya Jamieson 40 are easily distracted (Murphy 1998). High performing detection dogs typically score higher for boldness in comparison to lower performing dogs (Svartberg 2002).

An individual dog’s olfactory ability is dictated by a variety of factors, including their breed, anatomy and age (Rauth-Widmann 2006; Hurt & Smith 2009). The dog’s nasal cavity contains millions of sensory neurons within the olfactory epithelium (Craven et al. 2010). The dog’s body size influences their olfactory epithelium, with a Fox Terriers’ being ∼84 cm2 and a German Shepherds being ∼150 cm2 (Rauth-Widmann 2006). This large epithelium, put simply, means a larger area for sensory neurons (Rauth-Widmann 2006). More sensory neurons can increase the dog’s olfactory accuracy; however, this alone does not indicate the dog’s working ability (Rauth- Widmann 2006).

Brachycephalic breeds typically have poor olfactory abilities, due to the minimal space allowing for the olfactory epithelium to expand within their noses (Rauth-Widmann 2006; Bartels et al. 2015). Brachycephalic breeds also have notably less olfactory cells, thereby reducing their olfactory sensitivity (Rauth-Widmann 2006). Breeds with elongated noses (mesocephalic breeds) have more olfactory receptor cells, allowing them to identify and retain more scent (Craven et al. 2007; Abney 2009). Brachycephalic breeds also typically have breathing issues, resulting in less oxygen provision to the brain causing the dogs to tire easily (Rauth-Widmann 2006; Packer et al. 2012; Bartels et al. 2015). Brachycephalic breeds should therefore be avoided when selecting a detection dog.

Even a physically and behaviourally ideal detection dog is not suitable for field work if it is unhealthy. Breeds which have common health issues, such as German Shepherds and Bloodhounds with hip and elbow dysplasia, may not be the most suitable candidates (Palika 2007). Through gene analysis, however, breeding programs are now concentrating on reducing these common health issues within breeds (Janutta et al. 2006; Fels & Distl 2014). A dog’s breed is also a factor which influences their longevity (Fleming et al. 2011). This is demonstrated in the variation in longevity between breeds; with Great Danes having an average life expectancy of 6.5 years, in comparison to Jack Russell Terriers of 14 − 16 years (Palika 2007; Adams et al. 2010). A dog’s body weight and sexual entirety is also correlated to longevity, with smaller dog breeds and neutered individuals typically living longer (Moore et al. 2001; Galis et al. 2007; Greer et al. 2007; Adams et al. 2010). Neutering will also decrease distractibility, roaming and aggressive tendencies (Hart & Eckstein 1997; Maejima et al. 2007).

La Toya Jamieson 41

Selecting detection dogs There is commonly performance variation in any working dog environment (Brownell & Marsolais 2002). Certain dog and handler teams consistently perform to a high standard, both in accuracy and efficiency (Brownell & Marsolais 2002). The success or failure of these teams can often be traced to the dog’s selection (Brownell & Marsolais 2002).

There are multiple screening tests which can evaluate and indicate a dog’s future working performance (Brownell & Marsolais 2002). However, there is a lack of uniformity regarding screening tests amongst the working dog community (Brownell & Marsolais 2002; Early et al. 2014). Certain screening tests may be given too much weight for their reliability and potential to indicate a dog’s working potential (Brownell & Marsolais 2002). Screening tests should be completed prior to selecting dogs for work, however, these tests should not be the sole indicator of a dog’s potential. Any behavioural or trait evaluation must be objective, reliable, meaningful and repeatable (Wilsson & Sundgren 1997). It would be of great benefit to the canine research and behaviour community for testing techniques to be standardised, allowing for results to be pooled and studies compared (Early et al. 2014; van Rooy et al. 2014).

The Dog Mentality Assessment is commonly used as a behavioural test for dogs, with the results being comparable to the dog owner’s questionnaire responses (Serpell & Hsu 2005; Svartberg 2005). This assessment measures the dog’s sociability, playfulness, fearlessness and boldness, through the use of multiple behavioural assessments (Svartberg 2005). These results are, however, greatly influenced by external factors, such as the scoring judge (Ruefenacht et al. 2002; Saetre et al. 2006).

Phenotyping, the recording and analysis of phenotypes, is used in behavioural studies through the use of owner questionnaires, battery testing and observation studies (van Rooy et al. 2014). A commonly used testing assessment is the behavioural test battery (Jones & Gosling 2005), where dogs’ reactions are gauged when they’re exposed to specific situations and stimuli (Svartberg & Forkman 2002; Svartberg 2002, 2005; Serpell & Hsu 2005). Similarly ad hoc observational tests occur in uncontrolled environments, where the stimulus is naturally occurring (Goddard & Beilharz 1984; Mirkó et al. 2012). These tests can be used to determine commonly demonstrated traits in a naturalistic environment, allowing for conclusions about a dog’s behavioural patterns and temperament to be made (Goddard & Beilharz 1984; Murphy 1995, 1998; Jones & Gosling 2005). For phenotyping to be useful it must be sensitive, reliable and objective (van Rooy et al. 2014). Whilst observational studies can provide a wealth of data, they are commonly less used, due to their high time and financial costs (van Rooy et al. 2014).

La Toya Jamieson 42

A major difficulty with behavioural studies has been measuring and defining certain behaviours, along with how they are expressed (van Rooy et al. 2014). Improved methods for measuring dog behaviour are required (McGarrity et al. 2016). As multiple ways to assess and measure behaviour have been utilised, behavioural studies cannot simply be compared (van Rooy et al. 2014). International standards in testing protocols for dog behavioural evaluations would therefore be beneficial (van Rooyetal. 2014). With the cost of breeding and rearing dogs in mind, improving behavioural measurements is crucial for improving the efficiency of working dog breeding programs (McGarrity et al. 2016).

Commonly used dog breeds for detection work There are over 400 breeds of dogs displaying a high level of diversity in behaviour and morphology (Bradshaw et al. 1996; Svartberg 2006; Serpell & Duffy 2014). Of these dog breeds, few are chosen to become working detection dogs. Breeds have specific traits that are linked to genetic mutations or to artificial selection by humans (Serpell & Duffy 2014). For detection dogs, genetic differences, often expressed morphologically, may determine working aptitude (Maejima et al. 2007).

For drug detection, Labrador Retrievers, German Shepherds, Terriers (e.g. Jack Russell) and English Springer Spaniels are commonly selected breeds (Jezierski et al. 2014). In the United Kingdom the most common breeds for drug and explosives detection are English Springer Spaniels, Labrador Retrievers, Cross breeds and Border Collies (Rooney & Bradshaw 2004). Similarly the Labrador Retriever is the most common narcotics detection dog in Japan (Maejima et al. 2007). Within these narcotic detection programs 30% of dogs that enter narcotics detection training become working detection dogs (Maejima et al. 2007). Whilst this percentage may be a result of strict selection criteria, it still presents a large proportion of dogs that require rehoming. This percentage may also indicate a problem with the chosen dog breed, the methods utilised to select the potential detection dogs, or the training (McGarrity et al. 2016).

Variation between breeds Whilst anatomical differences are typically compared between dog breeds, the greatest variation is behavioural (Table 3.1)(Coppinger & Schneider 1995). All canine behaviours have a genetic component, which can be breed-specific, such as livestock guarding in Maremmas (van Rooy et al. 2014). Behaviour is influenced by learning, epigenetics and their surrounding environment (van Rooy et al. 2014). Dog breed variation can typically be explained by their original selection for working use (Helton 2010; Adamkiewicz et al. 2013). For example sight-hounds, such as Greyhounds, were chosen for their speed, and terriers for their ability to hunt underground (Helton 2010). These specialist dog breeds were created through continual artificial selection, however, as

La Toya Jamieson 43 previously mentioned, there is currently no specialist breed for detection work (Rooney & Bradshaw 2004).

Dog breeds are perceived as differing in trainability and intelligence, with this perception further complicating identifying a suitable working individual (Rooney & Bradshaw 2004; Serpell & Hsu 2005; Helton 2009; Ley et al. 2009). Intelligence can be defined as an individual’s ability to learn, perceive and process specified information and apply it in a specific situation (Zhong et al. 2015). Trainability can be defined as a dog’s ability to learn skills or tasks, and can be measured through evaluating a dog’s performance and speed at learning a task (Helton 2010; Turcsán et al. 2011). Coren’s ranking is a ranking of 133 dog breeds for their working intelligence, which should have probably been termed ‘trainability’ (Helton 2010), based on the opinion of professional obedience judges (208 North American experts in total) (Coren 1994). According to Coren’s ranking, the most intelligent dogs are Border Collies, Poodles, German Shepherds, Golden Retrievers, Doberman Pinschers, Shetland Sheepdogs, Labrador Retrievers, Papillons, Rottweilers and Australian Cattle Dogs (Helton 2010). The least intelligent breeds are Basset Hounds, Mastiffs, Beagles, Pekingese, Bloodhounds, Borzoi, Chow Chow, Bulldog, Basenji and Afghan Hounds (Helton 2010). As previously stated, this ranking of a breed’s intelligence may be more related to the breed’s trainability (Helton 2010). Trainability is not the equivalent of an unguided problem-solving ability, thereby demonstrating that trainability isn’t necessarily related to intelligence (Frank & Frank 1985). This perception of variation amongst dogs’ trainability and intelligence is remarkable, considering the minimal evidence of differences in cognitive abilities between breeds (Gagnon & Doré 1992; Pongrácz et al. 2005).

A dog breed’s cooperativeness has influenced how they have been used by humans (Gácsi et al. 2009). ‘Cooperative worker’ breeds, such as gun and herding dogs, typically work with continual human cues and visual contact (Serpell & Hsu 2005; Gácsi et al. 2009). ‘Independent worker’ breeds, such as scent-hounds and livestock guarding dogs, typically work with minimal human cues (Gácsi et al. 2009). As a result, ‘cooperative worker’ breeds are able to respond to human cues more successfully and cooperatively than ‘independent worker’ breeds (Gácsi et al. 2009). This ability to cooperate with humans is invaluable for a detection dog (Gácsi et al. 2009; Hurt & Smith 2009).

Due to the importance of a dog’s personality when acting as a detector, knowledge regarding their personality is important when selecting and training individuals (Svartberg 2002; Svartberg & Forkman 2002; McGarrity et al. 2016). Table 3.1 demonstrates common qualities of five dog types.

La Toya Jamieson 44

Table 3.1. Dog types and their common behavioural attributes.

Dog Type Common attributes References Longevity References (years) Gun dogs  Low aggression Vas et al. 11 – 13 (Labrador Palika (e.g. levels (2005) Retrievers) (2007); Labrador  Cooperative with 12 – 15 (English Mehus-Roe Retrievers, humans Springer (2009) English  Highly trainable Helton Spaniels) Springer (2010); Spaniels) Turcsán et al. (2011)  Strong chase Palika (2007); instincts – Mehus-Roe potential threat to (2009) small animals  High stamina Herding  Highly trainable Coren (1994); 13 – 15 (Border Palika dogs (e.g. and intelligent Serpell & Hsu Collie) (2007); Border (2005); 12 – 14 Mehus-Roe Collies, Helton (Australian Cattle (2009) Australian (2010); Dog) Cattle Dogs, Turcsán et al. 9 – 11 (German German (2011) Shepherds) Shepherds)  Strong play-drive Svartberg & Forkman (2002)  Independent Serpell & Hsu problem solvers (2005);  Cooperative with Jakovcevic et humans al. (2010)  High stamina Svartberg  Bold (2002); Palika (2007); Mehus-Roe (2009) Terriers (e.g.  Poor trainability Helton (2010) 14 – 16 (Jack Palika Jack Russell  High energy Palika (2007); Russell Terriers) (2007); Terriers) levels Mehus-Roe Mehus-Roe  Strong play- and (2009) (2009) prey-drive  Prone to behavioural problems  Aggressive Duffy et al. tendencies (2008) Sight-  Susceptible to Alpak et al. 12 – 14 Palika hounds (e.g. injury (due to (2004); Kemp (Greyhounds) (2007); Greyhounds) low bone mass) et al. (2005);

La Toya Jamieson 45

 Poor Helton Mehus-Roe maneuverability (2007a) (2009)  Poor trainability Helton (2010); Turcsán et al. (2011)  Single-minded Palika (2007);  Disobedient off- Mehus-Roe leash (2009) Scent-  Sociable yet Svartberg and 9 – 11 Palika hounds (e.g. stubborn Forkman (Bloodhounds) (2007); Bloodhounds,  Moderate energy (2002) 14 – 15 (Beagles) Mehus-Roe Beagles) levels (2009)  Relatively poor Coren (1994); trainability Ley et al. (2009)  Single-minded Palika (2007);  Disobedient off- Mehus-Roe leash (2009)

It must be acknowledged that dog breeds continue to change over time, with the possibility that they no longer possess the physical or behavioural traits they were originally bred for (van Rooy et al. 2014). Breed specific characteristics are unlikely to be lost, however, unless there is active artificial or natural selection (van Rooy et al. 2014).

Breed performance and comparisons It is reasonable that, in order for them to become useful tools, detection dogs should be required to have their accuracy validated. Variation amongst performance can be breed related, with certain studies comparing specific breeds’ performances (Rooney & Bradshaw 2004; Maejima et al. 2007; Jezierski et al. 2014). This section will explore these comparative studies.

During police drug detection dog testing in Poland, German Shepherds had the highest accuracy and efficiency in comparison to Labrador Retrievers, English Cocker Spaniels and Terriers (Fox, Welsh and Jack Russell) (Jezierski et al. 2014). Terriers had the longest detection times and the highest proportions of false positives (Jezierski et al. 2014). Whilst Terriers demonstrated a relatively poor accuracy rate, their small size can be advantageous, allowing them to investigate confined areas (Jezierski et al. 2014).

When evaluating the difference between German Shepherds and Labrador Retrievers at the Swedish Dog Training Centre, a significant difference was reported between the breeds (Wilsson & Sundgren 1997). German Shepherds scored significantly higher for ‘defence drive’ and ‘sharpness’

La Toya Jamieson 46

(Wilsson & Sundgren 1997). Labrador Retrievers scored higher for ‘nerve stability’, ‘hardness’, ‘courage’ and reacted less to gun fire (Wilsson & Sundgren 1997). Labrador Retrievers were also more cooperative and affable than German Shepherds (Wilsson & Sundgren 1997). During a similar study Rooney and Bradshaw (2004) created a list of traits and asked dogs-handlers to evaluate their dogs against these traits. For ‘tendency to be distracted’ and ‘stamina’ Labrador Retrievers were significantly further from ideal than English Springer Spaniels (Rooney & Bradshaw 2004). For ‘food motivation’, Border Collies were significantly closer to ideal than Labrador Retrievers, and were scored closest to ideal for ‘tendency to be distracted’ (Rooney & Bradshaw 2004). Based solely on this evaluation, Border Collies and English Springer Spaniels would be the most suitable breed for drug/explosives detection work (Rooney & Bradshaw 2004).

Detection dogs are bred for specific personality traits (McGarrity et al. 2016). Evaluating the outcome of these dog breeding programs can be challenging as even within these specifically selected populations individual variation is prominent (Jones & Gosling 2005; Graham & Gosling 2009; Fratkin et al. 2013). For example when examining a group of 1,310 German Shepherds and 797 Labrador Retrievers at the Swedish Dog Training Centre, 17 German Shepherds and 87 Labrador Retrievers were successfully trained as detection dogs (Wilsson & Sundgren 1997). Of the original 1,310 German Shepherds, 788 (60.1%) were rejected as working dogs (e.g. police and detection work), and 147 (11.2%) were euthanised for behavioural reasons (Wilsson & Sundgren 1997). Of the original 797 Labrador Retrievers, 530 (66.5%) were rejected as working dogs, and 42 (5.2%) were euthanised due to behavioural reasons (Wilsson & Sundgren 1997). The apparent minimal success of these breeding programs raises the ethical implications of breeding such a large quantity of dogs, further contributing to the domestic dog overpopulation.

When evaluating breed differences it is important to remember a dog’s early experiences may be just as influential on their behaviour as their innate tendencies (Rooney & Bradshaw 2004). The extent to which the innate characteristics of a breed contribute to adult dog behaviour is largely unresolved and should be further investigated (Serpell & Jagoe 1995; Willis 1995; Appleby et al. 2002).

Influence of sex and neutering Whilst there are commonly seen variations within breeds, there is also variation between the sexes. Females are typically easier to control, due to their smaller sizes, and have less aggressive tendencies (Rebmann et al. 2000). This is important if the dog is to live in a home or kennel environment where contact with other dogs is inevitable (Rebmann et al. 2000; Rooney & Bradshaw 2004). Whilst there are many considerations when choosing which sex of dog to use, it is

La Toya Jamieson 47 important to note that individuals may not possess traits typical of their sex. Regardless of sex, it could be argued if breeding is not intended dogs should be de-sexed (Moore et al. 2001). Neutering has also been reported to increase a dog’s trainability in certain breeds (Serpell & Hsu 2005). The benefits of neutering, however, are widely debateable, especially in regard to health benefits (Beauvais et al. 2012).

During a study (Maejima et al. 2007) male, neutered Labrador Retrievers had higher scores for ‘desire to work’, when compared to females and non-neutered males. Neutered dogs had significantly lower distractibility scores than non-neutered dogs (Maejima et al. 2007). Neutering therefore reduced the dog’s distractibility (Maejima et al. 2007). Previous studies reported male Labrador Retrievers scored higher than females for defence drive and hardiness, whilst females scored higher for ability to cooperate and lower level of aggression towards other dogs (Wilsson & Sundgren 1997). Male German Shepherds and Labrador Retrievers also scored significantly higher than females for defence- and prey-drive, and courage (Wilsson & Sundgren 1997). Male German Shepherds also scored significantly higher for cooperation, whilst the opposite was found for Labradors with females scoring significantly higher (Wilsson & Sundgren 1997). Whilst their traits and abilities are varied, entire male dogs are still more commonly used for detection work than females and neutered male dogs (Wilsson & Sundgren 1997; Maejima et al. 2007). This may be due to personal preference or bias during the dog selection process.

Is breed specific selection enough? Historically it was widely agreed that animals of a certain age, species and sex would behave similarly (Feaver et al. 1986). There is now substantial evidence suggesting a large variation amongst individuals’ behaviours within species and breeds (Manteca & Deag 1993; Wielebnowski 1999; Buffington 2002; Serpell & Hsu 2005; King et al. 2012).

Temperamental factors are the primary determinant of a dog’s working ability (Maejima et al. 2007; King et al. 2012; McGarrity et al. 2016) making certain individuals more suited to tasks than others (Serpell & Hsu 2001; Svartberg 2002). Different breed lines (e.g. working lines vs show lines) are also likely to result in different behavioural attributes (Houpt & Willis 2001; Serpell & Hsu 2005). While the link between dog health and physical characteristics has been established (van Rooy et al. 2014), a similar link with behaviour has been discovered for a relatively narrow range of physical characteristics (Gácsi et al. 2009). For example, brachycephalic dogs are more successful in following human pointing signals than dolichocephalic dogs (Gácsi et al. 2009). A dog, however, should not be selected for detection work based solely on their breed (Rooney & Bradshaw 2004; McGarrity et al. 2016).

La Toya Jamieson 48

A dog’s environment and lifetime experiences are highly influential and cannot be overlooked when investigating a dog’s working suitability (van Rooy et al. 2014; McGarrity et al. 2016). All behaviours must be viewed with environmental context in mind, which may shine light on the behavioural response (van Rooy et al. 2014). Learning plays an important role in behavioural development, with dogs repeating previously successful behaviours (van Rooy et al. 2014). Early experiences therefore shape a dog’s development and future behavioural tendencies (van Rooy et al. 2014).

Whilst selecting the best dog breed may increase the likelihood of an individual’s success at detection work, it will not guarantee it. Individual variation within breeds is typically immeasurable. What is clear is that each individual should be thoroughly tested and evaluated prior to selection, no matter the breed.

The importance of the dog handler A dog and its handler together are a partnership, which strongly influences their work. A strong bond is typically formed, which is of great importance when working in the field (Rebmann et al. 2000; Abney 2009). Personality conflicts will severely compromise the dog and handler team’s working ability and success (Smith et al. 2003). A dog handler must be able to trust their dog’s indications (a trained behaviour which demonstrates a target sample has been located) and similarly the dog must have trust in their handler’s commands. A dog which has been injured or overworked through following their handler’s commands will unlikely work well for them. A dog handler’s experience will also influence the success of the team (Gutzwiller 1990).

As with dogs, not everyone is a suitable dog handler (Rebmann et al. 2000). Whilst there are many characteristics that make up a suitable detection dog handler, some of the most important include: high level of fitness; knowledge of dog training and handling principles; trust in their dog’s behaviours; and ability to understand dog behaviour and body language (Rebmann et al. 2000). A successful dog and handler team is therefore created from the proper selection of both a handler and their dog, and assessment of their compatibility; adequate training of both parties; and continual performance evaluation (Hurt & Smith 2009).

Conclusions and recommendations From the literature reviewed, a suitable detection dog is an individual of medium size, with a high level of agility; highly play motivated; and high level of intelligence and obedience, yet independence when working off-leash. Whilst certain breeds commonly possess the proposed ideal traits (i.e. Herding and gun dogs), this does not mean that every individual will be suited to this

La Toya Jamieson 49 work. Nor does it mean that no other breeds will have suitable individuals. In the future the focus should therefore not be the mass breeding of specific breeds deemed best suited, but instead locating specific individuals that have the characteristics required to be successful detection dogs.

References

Complete reference list is provided at the end of the thesis due to reference repetition throughout the chapters.

La Toya Jamieson 50

Chapter 4: Who’s a good handler? Important skills and personality profiles of wildlife detection dog handlers

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2018, ‘Who’s a good handler? Important skills and personality profiles of wildlife detection dog handlers’, Animals, vol. 8, pp. 222-36.

Contributor Statement of contribution Conceived/designed study (90%) Designed questionnaire (90%) La Toya Jamieson Collected data (100%) Data analyses (100%) Wrote the article (90%) Designed study (5%) Greg Baxter Designed questionnaire (5%) Wrote/edited the article (5%) Designed study (5%) Peter Murray Designed questionnaire (5%) Wrote/edited the article (5%)

Quoll detection dog “Kuna” (RIP) and handler, Amanda Hancock

La Toya Jamieson 51

Abstract Wildlife detection dog teams are employed internationally for environmental surveys, and their success often depends on the dog handler. Minimal research is available on the skills that dog handlers believe are important, and no research has been published on the personality profiles of wildlife detection dog handlers. This may reveal the skills that people should acquire to be successful at, or suitable for, this work. An online questionnaire was distributed to Australian and New Zealand wildlife detection dog handlers. This questionnaire provided a list of skills to be rated based on importance, and a personality assessment measured their five main personality domains (Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness). A total of 35 questionnaires were collected, which represented over half of the estimated Australian wildlife detection dog handler population. The handlers had on average 7.2 years of dog handling experience, and 54% were female. More than half (57%) of the handlers stated that they were very emotionally attached to their dogs; however, 9% stated they were either not attached or mildly attached to their working dogs. The skill that was rated highest for importance was ‘ability to read dog body language’, and the lowest was ‘skilled in report writing’. On average, the handlers scored high in the Agreeableness domain, low in the Neuroticism domain, and average in the Extraversion, Conscientiousness, and Openness domains. However, all of the personality scores had large ranges. Therefore, a dog handler’s personality may not be as influential on their success as their training or their dog–handler bond. Further research would be beneficial regarding the direct impact that the dog–handler bond and the handler’s knowledge have on working team outcomes.

Introduction The dog handler is potentially the most important factor affecting working dog performance (Cobb et al. 2015; McGarrity et al. 2016; Jamieson et al. 2018a, b). The handler is responsible for more than simply following their dog in the field. They must be able to monitor and interpret their dog’s often subtle behaviours, and determine the most suitable areas and directions in which to work (Bird 1997). Underestimating the dog handler’s role can jeopardise not only the working dog’s performance, but also their welfare (Cobb et al. 2015). Whilst there is a large literature on ideal detection dog traits (reviewed by Jamieson et al. 2017), there is currently minimal information on the optimal profile of a dog handler, with handler and owner personality profiles only beginning to be researched (Schoeberl et al. 2012; Curb et al. 2013; Cobb et al. 2015; Payne et al. 2015). Understanding what makes certain people more suited to dog handling, such as their knowledge, skills, and personality types, requires further research (Payne et al. 2015).

La Toya Jamieson 52

Wildlife detection dogs are dogs that are used for an environmental purpose, such as locating scat, plants, or live animals to determine species presence or distribution (Hurt & Smith 2009). Wildlife detection dogs have been used globally for environmental surveys (Kapfer et al. 2012; Browne et al. 2015; Cristescu et al. 2015), and are particularly useful for species that occur at low densities (DeMatteo et al. 2009; Orkin et al. 2016). Training a detection dog handler to operational standard requires significant time and dedication (Orkin et al. 2016). Therefore, it is crucial that the best candidates are selected prior to dedicating significant resources. Whilst the literature lists certain important traits for dog handlers to possess (e.g. Rebmann et al. 2000, Hurt et al. 2016, Minhinnick et al. 2016), there is no study that determines what wildlife detection dog handlers believe are important knowledge and skills. This is valuable information for new dog handlers or established handlers wanting to further their skills. Wildlife detection dog handlers will likely require unique skills in comparison with other detection fields due to the nature of wildlife detection work. Wildlife detection work is often required in highly remote areas, which poses multiple threats to the dog– handler team. The handler is also responsible for ensuring that their dogs pose no threat to wildlife, and may also be required to have a similar level of knowledge to wildlife ecologists to ensure survey success, which is specific to this field. Currently, there is little information published regarding the factors that influence dog handler selection (Minhinnick et al. 2016). Therefore, this field would benefit from an analysis of the skills and knowledge that are crucial for a detection dog team’s success (Minhinnick et al. 2016).

When examining a person’s personality, their five main personality domains are typically measured - Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness (Johnson 2014). Previous research has determined that dog handlers have unique qualities, as shown by their five main personality domains, when compared to the personality profiles of the general population (Kaleta et al. 2011). In some cases, handler personality traits have also been correlated with their dog-handling practises (Payne et al. 2015). For example, male Polish police dog handlers were reported to have very high Conscientiousness scores, slightly above average Agreeableness and Extraversion scores, slightly below average Openness scores, and very low Neuroticism scores (Kaleta et al. 2011). However, no research has determined if wildlife detection dog handlers share similar traits. As previously highlighted, wildlife detection dog handlers have unique working requirements, so determining this information may assist with dog handler selection. As a handler’s personality impacts their dog training and handling practises (Payne et al. 2015), it may also allow for dog-handler training programs to be better constructed for individuals.

La Toya Jamieson 53

This research had three main aims: (1) collect general information from Australian and New Zealand wildlife detection dog handlers to improve our understanding of their working roles; (2) determine the characteristics and knowledge that dog handlers believe are important to be successful in this specialised field; and (3) use a well-tested psychological assessment to determine if wildlife detection dog handlers share similar personality profiles. This information may assist with future wildlife detection dog handler selection and highlight skills that need to be attained to be successful in this niche field. This is especially needed in Australia, where wildlife detection dogs are still a relatively new survey method. The research hypotheses were: (1) handlers will rate the skills specific to dog handling and training highly (> 4.5); and (2) handlers will have high Conscientiousness scores and low Neuroticism scores.

Methods Questionnaire We constructed a questionnaire to collect information regarding the important skills and personality profiles of wildlife detection dog handlers. The questionnaire was in three sections (see Appendices). The first listed the traits and characteristics that the published literature has stated as being important dog handler qualities. These qualities were rated from one to five (with one being not important, and five being very important). The second section was a personality assessment, which was comprised of 120 questions that the participants used to describe themselves. The personality assessment was the International Personality Item Pool-NEO (IPIP-NEO-120; Johnson 2014). The IPIP-NEO-120 is a free, web-based personality assessment that has been demonstrated to be highly reliable with the results showing high similarity to commercial personality assessments (Johnson 2014; Maples et al. 2014). The IPIP-NEO-120 assessed the handlers’ five main personality domains: Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness (Johnson et al. 2014). The final section collected personal information such as the handlers’ age, sex, and years working as a detection dog handler.

Questionnaire Distribution The questionnaire was distributed via email to dog handlers actively working with wildlife detection dogs in Australia and New Zealand. At the time of the survey, there was no formal wildlife detection dog community in Australia; therefore, it was difficult to determine the potential sample size. Based on working groups, individual dog handlers’ internet presence, and inclusion in the published literature, we estimated that there were approximately 50 individuals working within this field in Australia. New Zealand’s detection dog community is governed by the Department of Conversation (DOC). Australian wildlife detection dog teams require no formal selection or

La Toya Jamieson 54 assessment. In contrast, New Zealand detection dog teams are selected by the New Zealand Department of Conservation (DOC), and teams must complete both an interim and full certificate assessment (Thomas, S pers. comm.). DOC estimated that there were approximately 70 dog handlers working in this field (Thomas, S pers. comm.). This department distributed our questionnaire to the New Zealand dog handlers. Professional dog trainers in Australia, who were also responsible for training certain handlers, were also contacted via email to distribute the questionnaire to their students (handlers). A reminder email was sent to all of the parties two weeks after initial contact in an attempt to increase sample size. All of the contacted parties were encouraged to share this questionnaire with other dog handlers who were known to them. All of the research that was completed had the University of Queensland’s Human Ethics approval (approval number: 2016001089).

Data Analyses Personality profiles were calculated online using the scoring process developed by Johnson (2014). The personality profiles that were created revealed where each dog handler ranked on the scale within each personality domain. General linear models were constructed to determine the influence of the dog handler’s age, gender, country of origin, employment (professional or volunteer), experience level, or target species (native or pest) on their personality scoring or rating of skills. Pearson correlations were then measured between the dog handler’s age and years of dog handling experience, and their personality scoring. Minitab 18 was used for all of the statistical analyses. Significance was set at p < 0.05.

Results General Handler Information A total of 35 completed questionnaires were collected from Australian and New Zealand dog handlers. Thirty-two individual dog handlers and eight working dog organisations were sent the questionnaire. As some working dog organisations distributed these questionnaires independently, it was not possible to calculate the exact response rate. Of the returned questionnaires, 31 were from Australian handlers and four were from New Zealand. The dog handlers had a mean age of 43.9 ± 9.1 years, with 54% being female. The majority of the handlers were professionals (80%), with a mean of 7.2 years of dog handling experience. Based on their detection dogs’ target species, 57% handled native species detection dogs, 37% handled pest species detection dogs, and 6% handled both native and pest species detection dogs (Table 4.1).

La Toya Jamieson 55

Table 4.1. Native and pest species, from Australia and New Zealand, listed as target species for the wildlife detection dogs. Total handlers currently handling these species-specific detection dogs are also listed.

Country Native Total Pest Total Electric ants Bell’s turtle 1 (Wasmannia 1 (Myuchelys bellii) auropunctata) Black-tailed antechinus Feral cat 1 7 (Antechinus arktos) (Felis catus) Fireweed Eastern bristlebird 1 (Chamerion 1

(Dasyornis brachypterus) angustifolium) Fire ants Emu 2 (Solenopsis 1 (Dromaius novaehollandiae)

Australia geminate) Koala Fox 10 6 (Phascolarctos cinereus) (Vulpes vulpes) Pygmy Blue-tongue lizard Hawkweed 1 1 (Tiliqua adelaidensis) (Hieracium spp.) Northern and Tiger quolls Introduced rodents (Dasyurus hallucatus and D. 8 2 (e.g., Rattus rattus) maculatus) Blue duck Feral cat (Hymenolaimus 2 1 (Felis catus) malacorhynchos) Brown teal Introduced rodents 1 1 (Anas chlorotis) (e.g., Rattus rattus) Kakapo 2 (Strigops habroptila) Kea 1

New Zealand New (Nestor notabilis) Kiwi 1 (Apteryx sp.) South Island Takahe 1 (Porphyrio hochstetteri)

The detection dogs used by the respondents belonged to a variety of breeds, with the majority coming from herding, gun dog, or terrier breeds (including cross-breeds). The dogs that are most commonly and currently used in Australia are Border Collies (29%), Labrador Retrievers (17%), and English Springer Spaniels (17%). Of the dog breeds listed, 26% of handlers recorded currently working with cross-breeds. Whilst the handlers varied in how emotionally attached they were to their detection dogs, all of the handlers agreed that their behaviours and stress levels impacted their dogs’ behaviours. Of the 35 handlers, 57% stated they were very emotionally attached to their dogs, while 34% were moderately emotionally attached, and 9% were mildly or not emotionally attached.

La Toya Jamieson 56

Important Handler Characteristics and Knowledge The dog handlers were asked to rate a series of important dog handler skills and knowledge, which were listed in the literature. Qualities rated as most important were ‘ability to read dog body language’, ‘ability to trust in a dog’s indications’, ‘strong working ethic’, and ‘knowledgeable on dog behaviour’ (Table 4.2). The qualities that were rated least important were ‘skilled in report writing’, ‘strong leader’, and ‘theoretical background in ecology’.

Table 4.2. Scoring of skills and knowledge based on their importance and relevance for wildlife detection dog handlers (1: strongly disagree it is important, 5: strongly agree it is important; SD: standard deviation).

Rank Skill Mean SD Ability to read dog body language 4.7 0.5 1 Ability to trust in a dog’s indications 4.7 0.5 2 Strong work ethic 4.6 0.6 3 Knowledgeable on dog behaviour 4.5 0.6 Skilled in dog handling 4.4 0.8 4 Ability to read wind direction 4.4 0.7 Navigational skills 4.3 0.8 5 Sound knowledge of target species 4.3 0.8 6 High level of physical fitness/stamina 4.0 0.9 7 Practical ecological experience 3.9 1.0 8 Team player 3.7 1.1 Knowledgeable of canine olfactory physiology 3.6 0.9 9 Experienced in dog training 3.6 1.0 Theoretical background in ecology 3.5 1.0 10 Strong leader 3.5 1.0 11 Skilled in report writing 3.0 1.0

General linear models determined no significant differences between how the handlers rated these traits based on the handlers’ age, sex, country, employment (professional or volunteer), experience level (years dog handling), or target species (native or pest species). The only exception was trait three (skilled in dog handling), which was significantly impacted by the handlers’ employment,

La Toya Jamieson 57 with volunteers rating it as significantly more important (mean = 4.9) than the professional handlers (mean = 4.3; p = 0.011). Of the handlers who listed additional skills or qualities that they believed were important, 45% stated patience, and 35% stated experience or knowledge of their working environment.

Personality Profiles The handlers’ personality domains that were the most different to the average scores were the Agreeableness and Neuroticism domains. This scoring system is based on the classifications of ‘low’, ‘average’, or ‘high’ that are provided when the personality scores were calculated online, which was developed by Johnson (2014). This online program classified a person’s results based on whether their score was in the lowest 30%, middle 40%, or highest 30% based on age and sex. For the purpose of this research, due to the differences in participants’ ages and sex, scores between 35– 65 were classified as ‘average’. Scores above 65 were classified as ‘high’, and scores below 35 were classified as ‘low’. Whilst handlers’ mean personality scores differed from the average, there was a large range between the individual handlers’ scores (Table 4.3).

La Toya Jamieson 58

Table 4.3. The dog handlers’ personality scores in the five main personality domains (in bold), and the traits (shown below each domain) within these domains. The standard deviation and range is also provided.

Domain Mean SD Range Extraversion 57.6 25.7 13–99 Friendliness 58.1 27.3 7–98 Gregariousness 44.9 29.2 2–99 Assertiveness 49.4 25.8 4–95 Activity Level 71.9 18.6 36–99 Excitement seeking 51.0 25.1 12–95 Cheerfulness 56.5 22.6 5–88 Agreeableness 67.5 24.0 13–99 Trust 62.9 21.7 8–95 Morality 63.2 23.3 4–89 Altruism 55.7 25.4 13–95 Co-operation 61.3 23.9 7–89 Modesty 68.4 26.5 16–99 Sympathy 55.9 26.2 2–99 Conscientiousness 57.7 25.3 2–96 Self-efficacy 48.6 24.4 1–97 Orderliness 46.9 22.5 16–90 Dutifulness 53.3 23.1 5–95 Achievement-striving 62.3 24.8 2–91 Self-discipline 59.4 27.6 6–99 Cautiousness 60.9 24.5 18–97 Neuroticism 34.9 27.7 1–82 Anxiety 36.6 29.0 1–97 Anger 31.5 27.4 1–76 Depression 34.7 25.0 3–77 Self-consciousness 44.9 23.3 1–90

La Toya Jamieson 59

Immoderation 43.8 24.4 1–85 Vulnerability 43.1 28.7 1–99 Openness 52.1 27.51 2–95 Imagination 42.1 27.4 1–83 Artistic interests 49.7 27.0 1–89 Emotionality 41.7 30.5 1–97 Adventurousness 66.4 24.7 15–99 Intellect 47.5 28.9 1–91 Liberalism 62.0 17.5 25–89

General linear models determined that the handlers’ personality domains were not affected by age, except for the Conscientiousness domain (p = 0.04). Age was negatively correlated to Conscientiousness (Pearson correlation = −0.348, p = 0.04). Similarly, the dog handlers’ personality domains were not impacted by which target species they detected (native, pest, or both), except for the Openness domain (p = 0.018). The dog handlers’ gender, country of residence (i.e., Australia or New Zealand), employment (professional or volunteer handler), or attachment to their dog did not impact their personality scores. The only exception was Neuroticism, which was impacted by sex (p = 0.049), with males having higher mean scores than females (44 and 27.3, respectively). However, there was no significant difference between the different sexes’ personality domains. Further, the dog handler’s personality domains weren’t impacted by their level of emotional attachment to their dogs.

Discussion The returned questionnaires provided information on current wildlife detection dog handlers in Australia and New Zealand, skills they believed important, and their personality profiles. There are extensive lists of skills, knowledge, and traits that the literature states are important for detection dog handlers to possess (e.g. Rebmann et al. 2000; Hurt et al. 2016; Minhinnick et al. 2016). These lists aren’t always created by dog handlers. Therefore, important skills may be overlooked.

Whilst all the skills listed in the distributed questionnaire were sourced from the literature, the dog handlers clearly rated certain skills to be more important than others. The skills that were listed the highest were focused on the detection dog itself. The only traits that were directly related to the detection dogs and were rated low were knowledge of canine olfaction and experience in dog training. This was unexpected, not only because knowledge on canine olfaction has been listed as

La Toya Jamieson 60 important dog handler knowledge (Goldblatt et al. 2009; Minhinnick et al. 2016), but also due to its relationship with environmental conditions. Dog handlers who have little experience in dog training or knowledge on dog learning theory are anticipated to struggle handling dogs in novel situations (Minhinnick et al. 2016). Dog handlers with dog training knowledge have been reported to be more self-confident, more aware of their dog’s working abilities, and use significantly less aversive handling methods (Haverbeke et al. 2010). Ensuring handlers have theoretical knowledge of dog training principles allows them to implement training practises that are efficient and humane. Therefore, future research would be beneficial to determine the performance variation between dog handlers who are also dog trainers, or have trained their current dogs, and handlers who are not experienced in dog training.

Whilst the dog handlers rated a sound knowledge of the target species relatively highly, both practical and theoretical ecological knowledge was not rated highly. This was unexpected, due to the impact that it may have on field survey success. Poor ecological knowledge may reduce the effectiveness of surveys and minimise information recorded on non-target species and habitat structure (Hurt et al. 2016). If wildlife detection dog handlers are working in close partnership with ecologists or land managers, this knowledge may not be as important. However, this knowledge is likely to aid dog handlers and potentially improve their future work (Hurt et al. 2016).

There was a large variation in the handlers’ responses to how emotionally attached they were to their current detection dog/s. However, this level of attachment did not impact their personality scores. Based on the published literature, which highlights the importance and impact of the dog– handler relationship (Zubedat et al. 2014; Payne et al. 2015; Hoummady et al. 2016), it was anticipated that all of the handlers would be very emotionally attached to their dogs. This was not demonstrated in our findings. Previous research has determined that dog–handler teams with higher quality relationships also had higher performance levels and better dog–handler communication (Hoummady et al. 2016; Diverio et al. 2017). It could be argued that the higher the dog–handler relationship, the higher the dog’s dependence on their handler, thereby reducing the dog’s independence and performance (Udell 2015). However, a larger proportion of the literature to date has supported the hypothesis that the dog–handler relationship is positively correlated to the dogs’ problem-solving abilities and working performances (Lefebvre et al. 2007; Horváth et al. 2008; Horn et al. 2013a; Hoummady et al. 2016; Diverio et al. 2017). During this project, the dog handler’s working performance was not evaluated. However, this would be beneficial in future research, with the dog teams’ performances being compared against their attachment levels. Further research is also needed on other factors that may impact dog–handler attachment and working

La Toya Jamieson 61 performance, such as detection dogs’ living arrangements and conditions, daily time spent with the dog, and where the dog was sourced (e.g. adopted, purchased, or purposefully bred).

The dog handler’s personality influences the dog–handler interface and therefore the team’s performance (Zubedat et al. 2014). As a result, previous research has investigated how these differences in a handler’s personality impact a working dog’s performance (Payne et al. 2015). Within our study, there were differences between the handlers’ mean personality scores and the ‘average’ scores; however, there was a large range. This was most strongly demonstrated in the Agreeableness and Neuroticism domains. The handlers scored highly for Agreeableness. Agreeableness has previously been positively correlated with team co-operation and avoidance of conflicts, and negatively associated with dog owner-directed aggression (Herron et al. 2009; Payne et al. 2015). Handlers scoring high for Agreeableness also use less verbal corrections, potentially creating a more positive dog–handler working relationship (Payne et al. 2015).

The dog handlers collectively scored low for the Neuroticism domain. Neuroticism in humans is related to anxious tendencies (Kis et al. 2012). Dog handlers scoring high for Neuroticism are reported to have dogs who perform less efficiently at working tasks and take longer to respond to commands (Kotrschal et al. 2009; Kis et al. 2012; Payne et al. 2015). Having handlers with low Neuroticism scores may also improve a dog’s stress levels, with dogs having lower cortisol levels when handled by calm handlers (Slotta-Bachmayr & Schwarzenberger 2007). Collectively, the dog handlers scored in the higher percentile of the ‘average’ score for the Conscientiousness and Extraversion domains. It has been postulated that high Conscientiousness scores will be positively associated with the dog–handler relationship and their working performance (Payne et al. 2015). However, this association requires further research, and was not demonstrated in our results where Conscientiousness scores were not correlated with dog–handler emotional attachment. High scores for Extraversion are related to more confident individuals (Schaefer et al. 2004). Whilst dog handlers should be confident in their team’s abilities, it is important that these scores don’t become too high, or else over-confidence could negatively impact their performance. Further research on the relationship between a handler’s Extraversion scores and their working performance would be beneficial. The dog handlers had an average score for the Openness domain. This was surprising, as Openness has previously been related to adaptability within a working environment (Payne et al. 2015). However, there is currently minimal research on the relationship between a dog handler’s Openness score and their work.

La Toya Jamieson 62

Conclusions Wildlife detection dogs and their handlers must have a compatible partnership. To achieve this, we need to begin to focus on the handlers—their selection, skills, and training—as much as we do on the dogs. Improving how we select and train detection dogs alone is not enough, nor is placing a high performing dog with an unsuitable or unknowledgeable handler going to result in good outcomes. Our research has highlighted the skills and knowledge that dog handlers believe to be important for working success. Whilst these skills were focused on wildlife detection work, certain skills are likely transferable to other detection fields. As these results weren’t always consistent with the published literature, it would be beneficial to determine how different skill sets and knowledge may impact a detection dog team’s success and performance. Further research is also required to determine the correlation between mean personality scores and dog handlers’ performance. On average, the handlers scored high in the Agreeableness domain and low in the Neuroticism domain. Assessing handlers’ personality scores may assist with highlighting individuals who have personalities suitable for this work, but this should not be the sole screening process. Whilst the mean scores of the handlers were mainly consistent with the literature, there was a very large range. Therefore, personality may not be the overall determinant of a person’s suitability for wildlife detection dog work. Rather, a potential handler’s suitability is related to how someone applies themselves to their training and work, and their dog–handler bond. Future research should begin to focus on improving how detection dog handlers are selected and trained, and ensuring that handlers are provided with the best available information to allow them to shape their own dog’s training and working plans. Optimising this dog–handler pairing will likely improve working success and efficiency (Kydd & McGreevy 2017). Therefore, improving dog handler selection and training may improve field success and working dog welfare, which should be priorities for this highly applicable methodology.

Acknowledgments Thank you to all the dog handlers who participated in this project, and to all the organisations who helped distribute the questionnaire. The first author would also like to acknowledge she received an Australian Government Research Training Scholarship to complete this research.

References

Complete reference list is provided at the end of the thesis due to reference repetition throughout the chapters.

La Toya Jamieson 63

Chapter 5: Can Sighthounds scent? Dog breeds’ (Canis lupus familiaris) trainability at scent detection

Contributor Statement of contribution Conceived/designed study (90%)

Dog care and training (100%) La Toya Jamieson Collected data (100%) Data analyses (100%) Wrote the article (90%) Designed study (5%) Greg Baxter Assisted with accuracy testing Wrote/edited the article (5%) Designed study (5%) Peter Murray Wrote/edited the article (5%)

La Toya Jamieson 64

Abstract Assumptions are often made that there are variations between dog breed’s aptitudes for scent detection. This has resulted in specific breeds being favoured for detection tasks, even when recognising the large individual variation within breeds. Our preliminary research has compared Border Collies, Labrador Retrievers and Greyhounds for their trainability and accuracy at scent detection. These dogs were trained to locate Bengal Tiger (Panthera tigris tigris) scat. All training sessions were filmed and behaviour coded allowing for the total training time required to reach specific training milestones to be calculated. Once training was complete, the dogs’ odour discrimination ability was assessed, using 144 target, non-target and control samples. Behaviour assessments were completed both prior to and post training. The four Border Collies and four Labrador Retrievers completed training, and were therefore assessed, whilst only one Greyhound out of four did. On average, Border Collies had the quickest mean training time and the highest accuracy scores. The one Greyhound, however, had higher accuracy scores than half of the Labrador Retrievers. There was no correlation found between the dogs’ training times and their detection accuracy. Individual variation was significant in the breeds’ training times (F value = 78.11, p < 0.001). A dog’s breed may therefore not be the best indicator of their trainability, nor is their training time a sufficient indicator of their future detection accuracy.

Introduction The dog’s exceptional scenting and odour discrimination ability has led to them being versatile for a variety of detection tasks (Hurt & Smith 2009). A recently developed use of working dogs are wildlife detection dogs. These dogs are trained to locate wildlife through identifying their traces, such as scat or den/nesting sites, or even live animals (de Oliveira et al. 2012; Wasser et al. 2012; Browne et al. 2015). There is, however, a large variation in their reported success and accuracy (e.g. 60-85% detection success (Vynne et al. 2011); 92% detection success (Orkin et al. 2016)). The initial step to having a successful wildlife detection dog is selecting the correct individual for training.

A dog’s breed is often a point of focus during working dog selection (Minhinnick et al. 2016). Working dogs are typically bred for specific personality traits (McGarrity et al. 2016), with certain breeds being favoured (e.g. Labrador Retrievers and German Shepherds; Graham & Gosling 2009). It is important to highlight that whilst dog breeds have no personality, personality traits obtained from individuals within a specific breed are commonly used to characterise an entire breed or breed grouping (Miklósi 2015). The selection of specific dog breeds is therefore often more a result of personal preference, than scientific reasoning (Hall et al. 2015; Minhinnick et al. 2016). For wildlife

La Toya Jamieson 65 detection work, behavioural and physical traits that are viewed as ideal are: medium sized agile individual with an elongated nose that is highly play motivated, obedient and intelligent (Jamieson et al. 2017). There are few quantitative studies that demonstrate the variation between dog breeds for their detection ability (Serpell & Hsu 2005). Similarly, breed profiling is typically based on anecdotal or historical notions, rather than scientific-based evidence (Turcsán et al. 2011). Working dog programs, however, continue to select dogs from a small pool of breeds (Jamieson et al. 2017).

Even within specifically selected and bred working dog populations, individual variation is significant (Jones & Gosling 2005; Graham & Gosling 2009; Fratkin et al. 2013). Dog breeding and training programs are financially and resource expensive. Reportedly only 30-50% of dogs bred for working purposes, such as detection work (Maejima et al. 2007), police and military work (Slabbert & Odendaal 1999; Sinn et al. 2010), guide dogs (Batt et al. 2008a, b), and herding dogs (Arnott et al. 2014a, b) complete training or work in their originally intended field (Wilsson & Sundgren 1997). These programs may not only be financially unsustainable but may also be impacting animal welfare (Cobb et al. 2015). A working dog’s training success is dependent on multiple factors, including their genetic makeup (i.e. breeding lineage), rearing, housing, and training and handling techniques used by the trainers (Cobb et al. 2015).

A dog’s trainability is crucial for working success and the dog-handler relationship (Serpell & Hsu 2005; Minhinnick et al. 2016). Determining the variation between and within breed trainability has significant implications for working dogs’ welfare and their attributed value (Serpell & Hsu 2005). Previous studies have explored breed variation for trainability and working suitability (Scott & Fuller 1965; Serpell & Hsu 2005; Adamkiewicz et al. 2013; Hall et al. 2015). No study, however, has explored dog breeds’ training times to specific field-relevant competencies and determined its relationship with the dogs’ overall odour discrimination ability. Our study therefore attempted to mimic training requirements which would be necessary for field success to provide a better representation of the dogs’ trainability.

For this preliminary study we purposely selected Border Collies, Greyhounds and Labrador Retrievers based on our review of breeds that commonly possess suitable and unsuitable characteristics for detection work (Jamieson et al. 2017). Our aims therefore were: (1) determine the variation in training progress between and within breeds, (2) determine if training time is related to odour discrimination ability, and (3) determine if the most successful dogs (i.e. dogs trained fastest or with the highest accuracy scores) shared similar characteristics (i.e. age or behaviour assessment scores). We hypothesised that Border Collies and Labrador Retrievers would be trained faster and achieve a higher level of detection accuracy than Greyhounds.

La Toya Jamieson 66

Methods Selected breeds and dogs Based on our review (Jamieson et al. 2017) three dog breeds were selected for their commonly attributed physical and behavioural characteristics. The breeds selected were: (1) the breed identified as having the most ideal physical and behavioural characteristics for wildlife detection work – Border Collie; (2) the breed listed as having the least ideal physical and behavioural characteristics for wildlife detection work – Greyhound; and (3) the breed listed most frequently being used for detection work – Labrador Retriever. The selected breeds therefore belong to different dog breed types, respectively Herding dogs, Sighthounds and Gun dogs.

Twelve dogs were trained in total, with four from each breed (Table 5.1). Due to the extended training requirements the dogs were separated into three groups, with each group having at least one member of each breed. This was to minimise the environmental variation between the groups. Group 1 was trained between March – June, Group 2 between June – September and Group 3 between September – November 2017.

Table 5.1. The dogs used in this research.

ID Group Breed Sex Neuter status Age (years) BC1 1 Border Collie Male De-sexed 6 BC2 1 Border Collie Female Entire 2 BC3 2 Border Collie Female Entire 4 BC4 3 Border Collie Male Entire 4 L1 1 Labrador Retriever Female Entire 2.5 L2 2 Labrador Retriever Male Entire 2 L3 2 Labrador Retriever Female Entire 2 L4 3 Labrador Retriever Female Entire 5 GHD1 1 Greyhound Male Entire 3 GHD2 2 Greyhound Male Entire 3 GHD3 3 Greyhound Male De-sexed 2.5 GHD4 3 Greyhound Female Entire 3.5

All selected dogs were either pedigree or purebreds. All Border Collies and Labrador Retrievers were sourced from breed specific breeders. The Greyhounds were sourced from a rescue organisation who had them surrendered from the Greyhound racing industry. All dogs had to be at least 12 months old to ensure they were mentally mature enough for extensive training (Hurt & Smith 2009), with no known health problems. None of the dogs had previous detection training, and

La Toya Jamieson 67 had minimal, if any, obedience training. The use and temporary housing of these dogs was approved by The University of Queensland’s Animal Ethics committee (approval permit: SAFS/454/16).

Kennelling The dogs were housed for up to three months at The University of Queensland’s School of Veterinary Science’s Clinical Studies Centre. The dogs were housed individually in kennels and had free access to both an outside area and an air-conditioned indoor area. The dogs had constant access to water and clean bedding. To ensure their care was standardised, the same feeding and cleaning schedule was completed daily by the same person (their trainer). The dogs were all fed the same food twice daily via ‘scatter feeding’ (scattering dry biscuits around their kennel) or with ice blocks (biscuits frozen within water) to increase feeding time. On days the dogs weren’t trained (two days per week) they went for a walk with their trainer or to a 30 m x 10 m fenced exercise yard for free- play.

Behaviour assessments The dogs were assessed using the Match-Up II Shelter Dog Rehoming Program/Behavior Evaluation. This assessment is a standardised dog behaviour evaluation, specifically shelter dogs, and its’ results and reliability have been explored and validated previously (Marder et al. 2013). This evaluation consists of 11 sub-tests that are used to score a dog’s Friendliness, Excitability, Playfulness, Fearfulness, Aggressiveness and Trainability (Marder et al. 2013). For this study three sub-tests were not completed as they weren’t necessary for detection training (sub-test 8 part 2: Pig’s Ear, sub-test 9 Toddler Doll and sub-test 11 Dog-To-Dog Interaction). Suitable adjustments were therefore made to the score totals.

The first behaviour assessment occurred 5 – 7 days after the dogs entered the kennels. The second behaviour assessment occurred one week before accuracy testing. Assessments were done in an indoor room with which the dogs weren’t familiar. The dog’s handler for their assessment was their trainer. The behaviour recorder was the Clinical Studies Centre animal behaviourist. All assessments were filmed using a GoPro Hero4 silver camera.

Training There is currently no standard training protocol for wildlife detection work (Oldenburg et al. 2016). We therefore followed the training principles of general detection work, similar to that outlined and validated in Fischer-Tenhagen et al. (2017). Several dog trainers and behaviourists were also consulted prior to and during this training. Training detection dogs can be separated into three steps: adapting to their training environment; imprinting on their target odour; and odour discrimination

La Toya Jamieson 68 training (Göth et al. 2003; Fischer-Tenhagen et al. 2011). These steps were followed during this project. The dogs’ training began after a one week settling in period at the kennels. This period was spent getting the dogs accustomed to the kennels and bonding with their trainer. The dogs were trained five days per week by the same trainer, with sessions typically running for 5-10 minutes. These shorter sessions are favourable for increasing the speed of skill acquisition (Demant et al. 2011). Operant conditioning, with a significant focus on positive reinforcement (e.g. Demant et al. 2011), was used for training. The dogs were rewarded with food and verbal praise, which is recommended to ensure the reward is varied (Papet & Minhinnick 2016). To ensure the dogs were motivated during training high value treats were used. The goal of the dogs’ training was having them independently locating and indicating to their target odour (Bengal Tiger scat) amongst an assortment of non-target odours. This target odour was selected as we could guarantee it was novel and be sure the dogs would not encounter it again when returned to their owners.

Dog training was shaped around a competency list (Table 5.2). These competencies were important milestones in training a dog to accurately locate a target odour. As detection dog teams should be blind to the sample order as early as possible during training (Johnen et al. 2015), volunteers were used weekly during training to randomise the sample order. All training sessions were filmed using a GoPro Hero4 silver camera to record the dogs’ behaviours and calculate how much time and reinforcements were required to achieve a competency. A competency was classed as being achieved if the dogs accurately and confidently completed the competency five consecutive times.

La Toya Jamieson 69

Table 5.2. The list of competencies that the dogs’ training was shaped around.

Competency Description 1.0 Clicker trained The dog was associating the clicker with a reward 2.0 Learnt indication behavior The dog had learnt their indication behavior (e.g. sit) 3.0 Indicating to target being The dogs were actively walking up to the correct target held by handler sample jar, held by the handler, and indicating 4.0 Locating target on ground in The dogs were locating and indicating to the target training container odour on the ground within a training container The dogs walked along a line of odours and 5.0 Locating target in bricks demonstrated interest (e.g. stopping at sample, looking (with guidance from handler) at handler after smelling sample) to the target 6.0 Locating target and The dogs walked along a line of odours and indicated indicating with guidance form their handler 7.0 Locating target and The dogs walked along a line of odours and indicated indicating independently without guidance The dogs performed competency 7.0 but with their 7.1 With handler blind to order handler blind to the order The dogs searched a line of odours, absent of target, 8.0 Completes true negatives and did not falsely indicate The dogs performed competency 8.0 but with their 8.1 With handler blind to order handler blind to the order The dogs walked along a line of odours, containing at 9.0 Discriminating between least one non-target species sample, and indicated to the target and 1 non-target target The dogs performed competency 9.0 but with their 9.1 With handler blind to order handler blind to the order The dogs walked along a line of odours, containing at 10.0 Discriminating between least two non-target species sample, and indicated to target and 2 non-targets the target odour The dogs performed competency 10.0 but with their 10.1 With handler blind to order handler blind to the order 11.0 Working accurately with a The dogs located their target when there was only 1 target:non-target ratio of 1:8 target out of 8 samples The dogs performed competency 11.0 but with their 11.1 With handler blind to order handler blind to the order

Odour discrimination testing The dogs were only tested if they achieved all training competencies within 11 weeks. Previous studies have trained dogs for detection tasks within 10 weeks (Rooney et al. 2007). This limit was therefore set as it was presumed, based on their training program, if the dogs were going to complete training it would be achieved by 11 weeks. Accuracy testing followed the layout and

La Toya Jamieson 70 procedures listed in Jamieson et al. (2018b). Testing occurred outdoors over two days, with both morning and afternoon sessions, in a paddock previously grazed by cattle. The dog-handler team walked along the line of samples contained in plastic 5 mL vials placed within the holes of a standard house brick. Within each line (18 bricks total) was target (Bengal Tiger scat), non-target (bovine, Bos taurus, and Brush-tailed Phascogale, Phascogale tapoatafa scat), control and null samples. The control samples consisted of a brick with an empty plastic vial, and null samples were an empty brick. Thus each brick had one of five sample types. All samples and bricks were handled with species-specific gloves.

The sample order within these lines were randomised and unknown to the handler. A person who was following the team at approximately a 5 m distance recorded the dogs’ indications and stated whether these indications were correct. When correct the dogs were rewarded. No reward was given for a false indication. The dogs were free to complete multiple indications along the line of samples. By testing conclusion, the dogs had worked along eight lines, totalling 144 samples, of which 16 were targets.

Data analyses From the dogs’ indications during testing, their accuracy was evaluated by calculating the dogs’ sensitivity, specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV) scores. A dog’s sensitivity is their ability to identify a target sample and specificity is their ability to identify, but not indicate to, a non-target sample (Frederick & Bowden 2009). A dog’s PPV refers to the correct proportion of their target-present indications and NPV refers to the correct proportion of their target-absent indications (Frederick & Bowden 2009).

Minitab 18 was used for all statistical analyses. Due to the Greyhounds low success in training they were excluded from the statistical analyses. The Greyhounds were, however, included in the behaviour assessment data analyses as these data were complete. Pearson correlations were completed to determine the correlation between the dogs’ behaviour assessment scores, their training times and accuracy scores. Generalized linear models (GLM) were constructed to determine the effect of breed and sex on the behaviour assessment scores. Paired t-tests were completed to determine if there were significant differences between the dogs’ first and second behaviour assessments. To determine the effect of breed and sex on training time and detection accuracy, GLM were constructed. Statistical significance was set at p < 0.05. Due to the small sample size minimal analysis was completed on the training and testing data.

La Toya Jamieson 71

Results There were differences between the dogs’ behaviour scores, both within and between breeds (Table 5.3). There was significant variation among the dogs’ first assessed Trainability scores (F value = 5.63, p = 0.049), but not among their second test (F value = 2.75, p = 0.173). There were also differences between the dogs’ behaviour scores in the first and second assessments. This was most clearly demonstrated in their Fearfulness scores, which were significantly lower during the second test (p < 0.0001). The Labrador Retrievers, on average, had the highest scores for Friendliness (mean = 19.5 ± 2.6), Excitability (mean = 7.25 ± 4.9), Playfulness (mean = 5.75 ± 3.2) and Fearfulness (mean = 4 ± 1.8). Border Collies scored highest for Trainability (mean = 12 ± 2.4) and Aggression (mean = 0.5 ± 1). Only one dog displayed ‘aggressive’ behaviours during the behaviour assessments. This was during a sub-test where the dog’s feet needed to be handled. This dog had a previous foot injury and was therefore very sensitive of his feet. The Greyhounds scored average across the board, except for Trainability (mean = 4.5 ± 1.2) where they scored the lowest.

Table 5.3. Behaviour assessment scores using the Match-Up II Shelter Dog Rehoming Program/Behavior Evaluation.

Friendliness Excitability Playfulness Fearfulness Aggressiveness Trainability ID (score/23) (score/23) (score/17) (score/24) (score/24) (score/15) Test 1 2 1 2 1 2 1 2 1 2 1 2 BC1 14 14 1 6 1 0 6 6 0 2 11 12 BC2 10 9 3 1 1 1 8 2 0 0 13 15 BC3 6 5 0 1 1 1 9 3 1 0 6 9 BC4 16 20 7 9 3 5 4 1 0 0 9 12 Mean 11.5 12 2.75 4.25 1.5 1.75 6.75 3 0.25 0.5 9.75 12 L1 22 23 7 14 7 8 7 2 0 0 15 15 L2 16 20 4 7 7 7 12 5 0 0 9 9 L3 17 17 2 2 0 1 13 6 0 0 6 9 L4 19 18 8 6 5 7 7 3 0 0 3 8 Mean 18.5 19.5 5.25 7.25 4.75 5.75 9.75 4 0 0 8.25 10.25 GHD1 12 15 1 3 0 3 8 3 0 0 2 6 GHD2 12 13 3 2 4 3 6 5 0 0 2 3 GHD3 14 18 2 4 3 2 5 1 0 0 3 4 GHD4 16 19 2 1 4 8 6 2 0 0 3 5 Mean 13.5 16.25 2 2.5 2.75 4 6.25 2.75 0 0 2.5 4.5 Overall 14.5 15.9 3.3 4.7 3 3.8 7.6 3.25 0.08 0.2 6.8 8.9 mean SD 4.21 5.07 2.64 3.96 2.48 3.01 2.67 1.81 0.28 0.57 4.51 4.01 Note. The dark shading represents scores that are above, or equal to, the average (overall mean). Variation in competency training time was demonstrated both between and within the breeds (Table 5.4). One of the largest between breed differences was training success. Whilst all Border Collies and Labrador Retrievers completed training, only one Greyhound achieved all competencies.

La Toya Jamieson 72

Because of this, the greyhounds were removed from statistical analyses. The majority of the competencies were achieved fastest by either the Border Collies or the Labrador Retrievers, with the exception of competency 11.1 (Table 5.2) that was achieved fastest by a Greyhound (GHD4). The Border Collies achieved nine competencies in the shortest mean training time, and the Labrador Retrievers achieved five competencies in the shortest mean training time. At the beginning stages of training one dog from each breed responded fearfully when smelling the tiger sample. Two of these three dogs went on to complete training, but this initial fear did increase their training times.

Table 5.4. The number of reinforcements (R), training minutes (T) and training sessions (S) required for the dogs to achieve the training competencies.

Labrador Competency Border Collie Greyhound Overall mean Retriever R T S R T S R T S R T S C1 72.3 10.7 2.5 78.8 14.1 2.5 119 18.6 3.25 90 14.4 2.8 C2 5 1 1 23 5.4 2 239 34.4 13 88.7 13.6 5.3 C3a 73.3 14.2 3 118 28.7 4.8 226 41.5 9 139 28.1 5.6 C4b 105 20.1 3.5 136 31.6 4.8 175 31.7 8 138.4 27.8 5.4 C5 65 17.5 2.5 51.5 12.9 2.5 71 24.6 4.5 62.5 18.3 3.2 C6 65.3 16.2 3 54.5 14.8 2.5 158.5 39.3 8 92.8 23.4 4.5 C7 49 13.6 2.3 38 11.5 2 64 19.4 3.5 50.2 14.8 2.6 C7.1c 21.8 6.3 1.3 27.5 9.6 2 53 10.5 2 34.1 8.8 1.8 C8 NA 2.2 1 NA 3.2 1 NA 7.2 2 NA 4.2 1.3 C8.1 NA 3.4 1 NA 2.6 1.3 NA 3.3 1 NA 3.1 1.1 C9 17.8 5.5 1 12.8 4.8 1 81 19 4 37.2 9.9 2 C9.1 23.8 7.5 1.3 24.5 8.4 1.5 54 11 2 34.1 9 1.6 C10 17 5.2 1 19 7 1 77 19 4 37.6 10.5 2 C10.1 15.3 3.7 1 22.5 8.2 1.5 15 4 1 17.6 5.3 1.2 C11 12.5 3.7 1 17.5 7.5 1.5 32 10.3 2 21 7.2 1.5 C11.1 16.3 3.9 1.3 16 4.6 1.3 15 3.3 1 15.7 3.9 1.2 Note. aOnly 3 Greyhounds completed up to this competency. bOnly 2 Greyhounds completed up to this competency. cOnly GHD4 completed this and the remaining competencies. There was no significant difference between the Border Collie’s and Labrador Retriever’s total reinforcements (F value = 0.16, p = 0.707) or total training minutes (F value = 0.07, p = 0.795). There was, however, a significant difference between the breeds’ total number of training sessions (F value = 21.11, p = 0.006; Table 5.5). There was also a significant difference between the number of training sessions the individual dogs required (F value = 78.11, p < 0.001). GHD4, the only Greyhound to complete training, required 2.2 times more training sessions than the Border Collies’ mean and 1.8 times more training sessions than the Labrador Retrievers’ mean. Both the Border Collies’ and Labrador Retrievers’ mean number of reinforcements and training sessions were quicker (shorter) than the average times. Only the Border Collies had a quicker mean training time

La Toya Jamieson 73 than average. The male dogs, however, required less training sessions (mean = 29) than the females (mean = 35.8) to achieve all competencies, however, this was not significant (F value = 0.06, p = 0.816). The only significant correlation between the dogs’ behaviour assessment scores and their overall training time was their Trainability scores (Person correlation = -0.689; p = 0.04).

Table 5.5. The dogs’ total number of reinforcements (R), training minutes (T), and training sessions (S) required to achieve all training competencies.

Total ID R T S BC1 479 88.95 22 BC2 644 163.95 27 BC3 486 112.8 28 BC4 628 173.4 33 Mean 559.25 134.7 27.5 SD 88.9 40.5 4.5 L1 667 156.75 28 L2 662 195.25 32 L3 564 160.5 33 L4 660 186.1 39 Mean 638.25 174.6 33 SD 49.5 18.9 4.5 GHD1a 1,170 202 48 GHD2a 774 198.7 41 GHD3a 723 118.6 35 GHD4 1,139 241.1 60 Mean 951.5 190.1 46 Overall 716.3 166.5 35.5 mean Note. The dark shading represents values that are below, or equal to, the average (overall mean). aDogs that did not complete training (GHD1 – GHD3), which were excluded from the study early for not demonstrating sufficient progress, were not shaded as their results were incomplete. There were no significant differences between the breeds sensitivity (F value = 0.08, p = 0.790), specificity (F value = 0.05, p = 0.832), PPV (F value = 0.00, p = 0.964) and NPV scores (F value = 0.15, p = 0.715). The Border Collies mean accuracy scores, however, were higher than the Labrador Retrievers (Table 5.6). The Border Collies had a larger variation within the breed for specificity and PPV scores, based on their standard deviations, whilst Labrador Retrievers had a larger within breed variation for sensitivity and NPV scores. The Greyhound mean was not calculated as only one individual was tested. This Greyhound (GHD4), however, had higher sensitivity scores than two of the Labrador Retrievers and one of the Border Collies, and higher specificity scores than three of the Labrador Retrievers and two of the Border Collies. The female dogs also had significantly

La Toya Jamieson 74 higher PPV (F value = 8.98, p = 0.03), NPV (F value = 8.01, p = 0.037) and sensitivity scores (F value = 7.72, p = 0.039) than the male dogs.

Table 5.6. The dogs’ odour discrimination accuracy scores from testing.

Dog Sensitivity Specificity PPV NPV BC1 100 95 73 100 BC2 100 100 100 100 BC3 94 100 100 99 BC4 69 92 52 96 Mean 90.7 96.9 81.3 98.8 SD 14.9 3.8 23.1 1.9 L1 100 96 73 100 L2 38 98 67 93 L3 100 100 100 100 L4 75 97 75 97 Mean 78.1 97.5 78.6 97.4 SD 29.5 1.9 14.7 3.5 GHD4a 94 98 88 99 Overall 87.5 97.6 82.7 98.5 mean Note. The shading represents scores that are above average (overall mean). aNo mean provided as GHD4 was the only Greyhound tested. There were no significant correlations between the dogs’ training times, number of reinforcements or number of training sessions, and their accuracy scores (sensitivity, specificity, PPV and NPV). There were also no significant correlations between the dogs’ behaviour assessment scores and their accuracy scores.

Discussion Our preliminary study demonstrated the variation between 12 dogs, from three different breeds, training success and training times. All of the Border Collies and Labrador Retrievers in this study completed training, whilst only one Greyhound did. It has previously been reported that Gun dogs (e.g. Labrador Retrievers) and Herding dogs (e.g. Border Collies) respond quicker to human communicative signals than breeds who work more independently from people (Gácsi et al. 2009). This may be a contributing factor for why the Greyhounds in this study performed poorly overall in training. It is also likely that the variation in the dogs’ early experiences, due to differences in their sourcing, impacted their cognitive and behavioural development (Rooney & Bradshaw 2004; van Rooy et al. 2014). Had the Greyhounds used in this study had a positive upbringing, similar to that of the other research dogs, they may have performed better in this study. This requires further research. The Border Collies, on average, achieved the most training competencies within the shortest time. We initially hypothesised that Border Collies and Labrador Retrievers would be

La Toya Jamieson 75 trained faster than Greyhounds. Our hypothesis was supported by our data, however, it was not anticipated that the Border Collies would be trained faster than the Labrador Retrievers. The fact that the Border Collies were trained quicker than the Labrador Retrievers, with Labrador Retrievers more traditionally being used for detection work (Maejima et al. 2007; Jezierski et al. 2014), is surprising and may warrant further investigation. Due to this study’s sample size, however, broadscale conclusions cannot yet be drawn, especially considering the large individual variation within each breed.

A dog’s detection accuracy is as important as how quickly they can be trained. When accuracy tested, the Border Collies had higher scores than the Labrador Retrievers. A mean could not be provided for the Greyhounds as only one was tested. This Greyhound, however, had a higher sensitivity score than two Labrador Retrievers and one Border Collie and a higher specificity score than three Labrador Retrievers and two Border Collies. When comparing against the average accuracy scores, this Greyhound performed above average for all scores whilst the Border Collies’ mean was above average for only sensitivity and NPV. None of the Labrador Retrievers’ mean accuracy scores were above average. These results were not anticipated considering these breeds traditional uses.

Previous breed comparisons have found similar, un-expected results when training dogs at scent detection. Hall et al. (2015) reported that Pugs significantly outperformed German Shepherds in both detection work learning acquisition and the dilution level at which these dogs could detect their target. This research also involved Greyhounds, with 9/10 failing the initial motivation criterion and the last Greyhound failing the motivation criteria in the sample dilution phase (Hall et al. 2015). Had our study had a similar motivation assessment it is likely that a large proportion of the research dogs would have failed, with no training being completed. Therefore, valuable information would not have been collected and the learning potential of individual dogs not being highlighted.

Within breed variation is commonly highlighted in the literature (Björnerfeldt et al. 2008; Barnard et al. 2017), including within breed variation for odour discrimination ability (Lazarowski et al. 2015). Our results also demonstrated that individual variation was prominent in all breeds. On average the Border Collies had a larger within breed variation for their training times, and their specificity and PPV scores. Labrador Retrievers, however, had larger within breed variation for their sensitivity and NPV scores. Greyhounds had the largest within breed variation for training success. This within breed variation is important to highlight, especially for working dog programs which exclusively use specific dog breeds.

La Toya Jamieson 76

It is often presumed that a dog’s intelligence and trainability is correlated with their working suitability or detection performance. In our study however, whilst Border Collies and Labrador Retrievers were trained quicker than Greyhounds, the dogs’ training times were not correlated with their detection performance. Breeds are often favoured for working purposes if they are perceived to be easily trainable, with this perception often not based on data (Hall et al. 2015; Jamieson et al. 2017). If trainability is not correlated with detection performance this may challenge the favouring of certain breeds for working purposes. It may therefore be beneficial to persevere with dogs that are slow to train as they may have the potential if given the opportunity. This may be challenging when ‘time is money’, however, it may improve success rates in working dog programs and minimise industry wastage. Further research with larger sample sizes are required before any of these potential conclusions are drawn.

The behaviour assessments in our study revealed both anticipated and surprising behaviour scores. The dogs’ behaviour scores, to a degree, are consistent with the literature, with Labrador Retrievers typically being described as highly sociable dogs, Border Collies being highly trainable and Greyhounds being difficult to train (Coren 1994; Helton 2010). The poor relationships between the dogs’ behaviour assessment scores, and their training times and detection accuracy were not anticipated. This may be a result of the study’s sample size and further research is warranted.

All breeds initially responded fearfully to the tiger sample. This has been mentioned previously for other dogs trained to locate large carnivores (Hurt et al. 2016). Allowances may therefore need to be made when training dogs for large carnivore detection, with slow and positive introductions to the odours being needed. It should be highlighted that the Labrador Retriever (L3) who performed the most fear related behaviours to the tiger samples, and for the longest time, was also the dog with the highest accuracy scores.

This preliminary research demonstrates the training and detection performance of 12 dogs from three different breeds that were historically bred for very different purposes. Whilst this study had only a small sample size the results demonstrated significant variation between the individuals, regardless of breed. Whilst our research cannot draw conclusions on breed trainability and working suitability, it demonstrates the performances of multiple individuals which may assist with future research and training. Based on our preliminary results, a dog’s breed may not the best predictor of their trainability or future working suitability.

La Toya Jamieson 77

Acknowledgements The first author acknowledges that she received an Australian Government Research Training Scholarship to complete this research. The authors acknowledge all the dog owners who loaned their dogs to this study, especially M. Wedgwood and the Collingwood family. Thank you to all the dedicated University of Queensland volunteers, and dog handlers who committed their time and expertise to this study. Thank you to the Clinical Studies Centre staff for their help, support and guidance throughout this project. Thank you to the dog trainers who contributed to the creation of the dog training plan (D. Marchiori, P. Clark, J. Palazzo-Orr, T. Robinson and S. McColl) and to D. Marchiori for assisting with the behaviour assessments. Thank you to D. Greenway for assisting with the statistical analyses.

References

Complete reference list is provided at the end of the thesis due to reference repetition throughout the chapters.

La Toya Jamieson 78

Chapter 6: You are not my handler! Impact of changing handlers on dogs’ behaviours and detection performance

Jamieson, La Toya J, Baxter, Greg S & Murray, Peter J 2018, ‘You are not my handler! Impact of changing handlers on dogs’ behaviours and detection performance’, Animals, vol. 8, pp. 176-87.

Contributor Statement of contribution Conceived/designed study (90%) Dog training and husbandry (100%) La Toya Jamieson Collected data (100%) Data analyses (100%) Wrote the article (90%) Designed study (5%) Greg Baxter Assisted with accuracy testing Wrote/edited the article (5%) Designed study (5%) Peter Murray Wrote/edited the article (5%)

“Zulu” and her familiar handler, thesis author

La Toya Jamieson 79

Abstract Dog-handler relationships can directly impact team success. Changing a dog’s handler may therefore compromise detection performance. However, there are currently few studies which support this. This research explored the performance and behavioural impact of changing a dog’s handler. Nine dogs trained at scent detection were accuracy tested with a familiar and unfamiliar handler. Both handlers were female with similar dog handling experience. The dogs were tested along brick lines containing target, non-target, and control samples. Testing was separated into four sessions, with each session having 36 samples. The dogs’ accuracy scores were then calculated and testing footage behaviour coded. The dogs had significantly higher sensitivity (p = 0.045) and negative predictive value (NPV) (p = 0.041) scores when handled by the familiar handler. With the unfamiliar handler the dogs performed more stress-related behaviours, and were distracted for a higher proportion of time (p = 0.012). Time spent distracted was negatively correlated to detection performance (correlation = −0.923, p < 0.001). With the unfamiliar handler the dogs’ performance did not improve throughout testing (p = 0.553). This research demonstrates how these dogs’ detection performances were impacted by changing handlers. Future research is required to determine if professional dog-handler teams are impacted similarly.

Introduction Domestic dogs (Canis lupus familiaris) and humans are closely bonded, with humans often replacing their dog’s conspecifics as their main social partner (Horn et al. 2013b). Due to this close relationship, dogs have been successful working partners with humans for centuries. Interactions between handlers and their animals, such as dogs and their handlers, influence their welfare and task performance (Hemsworth et al. 2009; Rohlf et al. 2010; Sorge et al. 2014; Zubedat et al. 2014). There has been much research on working dogs’ abilities, but further research is needed on factors impacting their success, such as the dog’s handler and their relationship (Johnen et al. 2013; Beebe et al. 2016; Hoummady et al. 2016).

Dogs have been used by humans for detection work, which can include narcotics and wildlife detection (Jezierski et al. 2014; Cristescu et al. 2015). Whilst a significant focus has been given to the detection dog, their handler and the dog-handler relationship has a strong influence on their working performance (Zubedat et al. 2014; Payne et al. 2015; Hoummady et al. 2016; Hurt et al. 2016; Diverio et al. 2017). Handlers must be able to recognise their dog’s subtle working behaviours and assist them (Hurt & Smith 2009; Vice et al. 2009). Failure to recognise these subtleties typically results in false negatives, where target samples are missed (Vice et al. 2009). Dogs may also trust human cues over their own olfactory senses (Szetei et al. 2003). These findings

La Toya Jamieson 80 demonstrate the importance of the handler on the dog’s performance and how easily they can impact their detection performance.

Dogs respond and behave differently to different people, depending on how familiar they are to each other (Kuhne et al. 2012; D’Aniello et al. 2015). For example, dogs demonstrate more ‘redirected behaviours’, including playing with inanimate objects and sniffing/licking the floor, and ‘appeasement gestures’, including blinking, averted head, and looking elsewhere, when interacting with a familiar person (Kuhne et al. 2012). Dogs will also typically respond quicker to the person they have a closer relationship with (Horn et al. 2013a). Hence, it has been postulated that the ideal system for explosives detection is a single dog and single handler team (Nolan & Gravitte 1977). However, research has typically focused on determining the importance of the dog handler’s experience, rather than the dog-handler relationship (Cooper et al. 2014).

Whilst some studies have demonstrated the abilities of dogs to work with multiple handlers (DeMatteo et al. 2009; Brook et al. 2012), no study directly compares dogs’ detection performances with an unfamiliar and familiar handler (Beebe et al. 2016). An unfortunate occurrence in the working dog community is the reluctance to share information (Minhinnick et al. 2016). It is therefore difficult to postulate how many working dogs experience a change of handler. Situations where this occurs include when dogs are trained and then sold to dog handlers; or when working dogs are owned by organisations or government agencies and are therefore used by multiple handlers. Belgian military dog teams, for example, are considered fully operational after a two- week settling in period with a new handler (Haverbeke et al. 2010). This transition period is relatively short considering it has been reported that these dogs are occasionally left in their kennels without handler interaction (excluding routine kennel cleaning and food distribution) for up to five consecutive days (Lefebvre et al. 2009). Changing working dog handlers may increase inconsistency or negatively affect the dog-handler bond, which is likely to generate conflict and compromise the team’s performance (Palmer & Custance 2008; Horn et al. 2013b). In circumstances where the dog’s role is potentially lifesaving (e.g. explosives detection), a compromised team can be catastrophic.

This preliminary study therefore aimed to compare dogs’ detection performances and behaviours when handled by a familiar and unfamiliar handler. These dogs weren’t operational working dogs, however, they had received extensive training at scent detection specifically for this research. The training principles used mimic professional detection dog training, and the dogs’ detection accuracy was rigorously assessed. We hypothesised that: (1) the dogs would have higher mean accuracy

La Toya Jamieson 81 scores with the familiar handler than the unfamiliar handler; and (2) the dogs would perform more stress-related behaviours, and be more distracted, when handled by the unfamiliar handler.

Methods Research Dogs Nine dogs extensively trained in scent detection work were used in this project (Table 6.1). These dogs were sourced from dog breeders, private owners, and an adoption centre. This research is part of a larger breed comparison training project where three dog breeds were used—Border Collies, Labrador Retrievers, and Greyhounds. After a literature review (Jamieson et al. 2017), Border Collies were selected as they are perceived to have the most suitable behavioural and physical traits for detection work, whilst Greyhounds were perceived to have the least. Labrador Retrievers were selected as they are one of the most commonly used breeds for detection work. These dogs were not professional detection dogs; however, they had received three months detection training, five days per week, before testing. This training used operant conditioning, with a significant focus on positive reinforcement, (e.g. Demant et al. 2011), to make the dogs associate their target scent with their reward (food). This training mimicked how professional working dogs are trained and is therefore comparable. Once the dogs were consistently making this association, non-target samples were included in the samples presented to improve their odour discrimination ability. The dogs’ training used a similar sample layout as described in ‘Testing layout’. These dogs were only tested if they achieved a high level of detection proficiency in training. Three dogs were therefore not included in this study as they did not reach this level of detection proficiency. Prior to testing, the dogs’ behaviours were assessed using the Match-Up II Shelter Dog Rehoming Program/Behavior Evaluation (Marder et al. 2013). From this assessment the dogs’ behaviour scores for ‘Friendliness’, ‘Fearfulness’, ‘Excitability’, ‘Aggressiveness’, ‘Playfulness’, and ‘Trainability’ were calculated. These scores were based on the dogs’ performance and frequency of related behaviours (e.g. sniffing, licking, or nudging a person were related to their ‘Friendliness’ score). Three behaviour assessment sub-tests were not completed as they weren’t relevant to this study. Adjustments were therefore made to the behaviour assessment totals. This research had the University of Queensland’s Animal Ethics Committees approval (approval number: SAFS/454/16) to house, train, and accuracy test all involved dogs.

La Toya Jamieson 82

Table 6.1. Detection dogs used during this project. Due to their extended training time, these nine dogs were separated into three groups: March–June (Group 1), June–September (Group 2), and September–November 2017 (Group 3).

Group Dog Sex Neuter status Age (years) Breed 1 1 Male De-sexed 6 Border Collie 1 2 Female Entire 2 Border Collie 1 3 Female Entire 2.5 Labrador Retriever 2 4 Female Entire 4 Border Collie 2 5 Male Entire 2 Labrador Retriever 2 6 Female Entire 2 Labrador Retriever 3 7 Male Entire 4 Border Collie 3 8 Female Entire 5 Labrador Retriever 3 9 Female Entire 3.5 Greyhound

Dog Handlers Both dog handlers were female, of similar height and build, with similar dog handling experience. Neither handlers were professional dog trainers/handlers, however, both had previously been instructed in dog training, either their own or other dogs, for several years. Handler 1 was the dogs’ trainer throughout this project. Handler 2 was introduced to the dogs on the first day of testing. Prior to testing, Handler 2 was sent information on each dog. This included basic information about the dogs, their personality, and tips on handling them. The morning of the first testing session, Handler 2 was also instructed by Handler 1 on how to handle each dog. Handler 2 had several practise runs along a mock test line-up with the dogs, whilst being instructed by Handler 1, for approximately 30 minutes. It was presumed in a real-world setting that a professional detection dog would not be transferred to another handler without this information being provided and this guidance given.

Testing Layout Accuracy tests were completed to measure each dog’s detection performance with both handlers. These tests were completed outdoors in a paddock used for grazing cattle, but none were present during testing. Tests were completed outdoors to better mimic the detection requirements of operational detection dogs. Clay house bricks (33×8×12 cm) with eight holes in each were laid out at a measured distance along a straight line. Each line comprised 18 bricks, which were separated into three groups of six. Each brick in a group was 2 m a part and there was 5 m between each group of six bricks. Target, non-target, and control samples were presented in the holes of these bricks. The target sample was Bengal Tiger (Panthera tigris tigris) scat, and the non-targets were cow (Bos taurus) and Brush-tailed Phascogale (Phascogale tapoatafa) scat. These samples were collected from captive facilities and a rural property, and were stored at −20◦C in a freezer. All

La Toya Jamieson 83 samples were contained in 5 mL plastic vials which, for storage and transport, had a screw cap. Control bricks and control samples were used in all brick lines, with the control samples being a brick with an open, but empty, vial. Of the 18 bricks, two were targets, nine were non-targets, four were controls, and three were empty bricks. The samples were randomly allocated to bricks, with eight brick lines being constructed over two testing days. The handlers did not know the order of the samples during testing. All dogs were tested along the same two brick lines in each session, with their testing order randomised.

Care was taken to minimise the likelihood of cross-contamination between the samples. Each species’ samples were only placed in their own specific bricks. Prior to testing these bricks were sterilised in boiling water, sun dried, and placed into new 50 L plastic storage tubs for storage and transportation. Labelled gloves, specific to each species, were worn when handling the samples or their bricks. The target and non-target samples were stored, transported, and handled separately, by separate field assistants.

Testing Procedures Prior to the dogs’ training, their testing order was randomly drawn. Five dogs (Dogs 1, 3, 6, 7, and 9) were tested firstly with Handler 1, whilst four dogs (Dogs 2, 4, 5, and 8) were tested firstly with Handler 2. After the first tests were complete (four sessions with each handler), the dogs had a rest day before being tested with the other handler. All tests therefore took four days to complete, with two testing sessions per day (morning and afternoon). Wind conditions, air temperature, and humidity levels were collected from the Australian Government’s Bureau of Meteorology weather station near to the testing site.

During testing, the other handler and dogs were a minimum of 50 m away from the team being tested, down wind and out of sight. The dog and handler team would walk along the brick line, with the handler repeating the cue ‘find’. A recording observer, who followed the dog team at approximately 5 m distance, was responsible for recording their indications and informing the handler if the indication was correct. The handlers were unknowing of the sample order. A person followed the recording observer at a 10 m distance from the dog-handler team and filmed all testing. The dogs indicated by sitting and facing the handler, or by stopping and turning to face the handler at the target brick. The dogs were rewarded with food and verbal praise. The dogs were not rewarded if they falsely indicated. To account for the changing and unpredictable wind conditions during testing, the dogs were allowed up to three attempts at each brick line. If the dogs did not locate the target by the third attempt the test ceased. By testing conclusion, each dog and handler team had been tested with 144 samples during four test sessions.

La Toya Jamieson 84

Behaviour Coding All tests were recorded using a GoPro Hero4 silver camera. The dogs’ behaviours were analysed and coded using Behavioral Observation Research Interactive Software (BORIS, Friard & Gamba 2016). Continuous behaviour sampling was completed. Stress-related behaviours were the main behaviours coded, which measured both their occurrence, frequency, and duration. Behaviours coded were lip licking, tail lowering, ears pinned back, yawning, whining, jumping, and shaking off (Handelman 2008). Behaviours were removed from future analyses if less than 50% of the dogs performed the behaviours at least once (e.g. whining). The proportion of time the dogs spent ‘distracted’ or ‘scenting’ was also calculated. ‘Distracted’ was defined as when the dog was ignoring the handler’s commands in order to smell or view other stimuli. ‘Scenting’ was defined as when the dogs were actively scenting/smelling along the brick line. At the conclusion of behaviour coding, time budgets were created.

Data Analyses From the dogs’ positive, negative, and false indications, their sensitivity, specificity, Positive Predictive Values (PPV), and Negative Predictive Values (NPV) were calculated. Sensitivity is a dog’s ability to locate their target and specificity is their ability to identify and not indicate to a non- target (Frederick & Bowden 2009). These are commonly used measurements of a detection dog’s accuracy (Porritt et al. 2015). A dog’s PPV assesses the correct proportion of their target-present indications and NPV assesses the correct proportion of their target-absent indications (Frederick & Bowden 2009).

General linear models were constructed to determine any relationships between the environmental conditions (wind speed, air temperature, and humidity) and the dogs’ performance. Statistical significance was set at < 0.05. Spearman correlations were completed to determine the relationship between the dogs’ ages and their accuracy scores. Spearman correlation was used as the data was not linear. Two-sample t-tests were completed to determine if there was a significant difference between the dogs’ performances with Handler 1 and Handler 2. General linear models were constructed to determine the impact of the training group and the testing session on the dogs’ performances. Two-sample t-tests were completed to determine the difference between the dogs’ behaviours when handled by the different handlers. Pearson correlations were completed to determine the relationship between the proportions of time spent ‘distracted’ and ‘scenting’ and the dogs’ performances.

La Toya Jamieson 85

Results The dogs had significantly higher sensitivity (p = 0.045) and NPV scores (p = 0.041) with Handler 1 (Table 6.2). When the dogs were with Handler 1 they also had higher mean PPV scores than when they were with Handler 2 (80.9 and 53.5, respectively), however this was not statistically significant (p = 0.114). There was no significant difference between the dogs’ specificity scores between the handlers. Of the nine dogs, three did not work for Handler 2, as demonstrated by their sensitivity and PPV scores of zero (Table 6.2).

Table 6.2. Sensitivity, specificity, PPV and NPV scores with Handler 1 and Handler 2. Mean scores and standard deviations (SD) are also provided. Dogs 4, 5, and 8 did not work for Handler 2 as demonstrated by their scores of zero for sensitivity and PPV.

Dog Handler Sensitivity Specificity PPV NPV 1 100 95.3 72.7 100 1 2 100 99.2 94.1 100 1 100 100 100 100 2 2 50 97.7 72.7 94 1 100 95.3 72.7 100 3 2 100 100 100 99.2 1 93.8 100 100 99.2 4 2 0 98.4 0 88.7 1 37.5 97.7 66.7 92.6 5 2 0 100 0 88.9 1 100 100 100 100 6 2 100 99.2 94.1 100 1 68.8 92.2 52.4 96 7 2 18.8 96.1 37.5 90.4 1 75 96.8 75 96.8 8 2 0 99.2 0 88.8 1 93.8 98.4 88.2 99.2 9 2 62.5 98.4 83.3 95.5 1 85.4 97.3 80.8 98.2 Mean 2 47.9 98.6 53.5 93.9 1 21.4 2.6 17.1 2.5 SD 2 44.8 1.2 44.1 4.9

The dogs also behaved differently with the two handlers. The dogs had their ‘tails lowered’ (p = 0.035) and were ‘distracted’ (p = 0.012) significantly more when handled by Handler 2. The proportion of time the dogs spent ‘distracted’ was significantly higher with Handler 2 (p = 0.012), which significantly affected their sensitivity (p = 0.004), PPV (p = 0.010), and NPV scores (p = 0.005). The proportion of time spent ‘distracted’ had a strong relationship with the dogs’ sensitivity

La Toya Jamieson 86

(Pearson correlation = −0.923, p < 0.001; Figure 6.1), PPV (Pearson correlation = −0.846, p < 0.001), and NPV (Pearson correlation = −0.925, p < 0.001) scores. Scatterplot of Sensitivity vs Distracted

100 Handler 1 2

80

y

t 60

i

v

i

t

i

s

n

e 40

S

20

0

0 10 20 30 40 50 60 70 80 90 Distracted

Figure 6.1. Strongly correlated negative relationship between the proportion of time (%) spent ‘distracted’ and the dogs’ sensitivity scores with Handler 1 and 2.

The dogs spent a significantly higher proportion of their time ‘scenting’ (p = 0.022) with Handler 1, which significantly influenced the dogs’ sensitivity (p < 0.001; Figure 6.2), PPV (p = 0.003), and NPV (p < 0.001) scores. Proportion of time spent ‘scenting’ was strongly correlated to the dogs’ sensitivity (Pearson correlation = 0.897; p < 0.001). The dogs’ sensitivity significantly improved during the testing sessions when handled by Handler 1 (p = 0.017). The dogs’ sensitivity did not improve through the testing sessions with Handler 2 (p = 0.553), and there was a very weak relationship between testing session and sensitivity (Pearson correlation = 0.029; p = 0.867). There was a weak relationship between the dogs’ age and their sensitivity (Spearman correlation = 0.099, p = 0.697), specificity (Spearman correlation = −0.383, p = 0.117), PPV (Spearman correlation = −0.209, p = 0.406), and NPV scores (Spearman correlation = −0.056, p = 0.825). There was no significant difference between the three testing groups, based on the dogs’ sensitivity (p = 0.088), specificity (p = 0.409), PPV (p = 0.157), and NPV (p = 0.080) scores. The dogs’ detection performance was also not significantly impacted by the environmental conditions (wind speed, p = 0.185; air temperature, p = 0.835; or humidity, p = 0.641). There were evident differences between

La Toya Jamieson 87

the dogs’ behaviour assessment scores (Table 6.3). There were also few apparent similarities between the dogs who worked well for both handlers. Scatterplot of Sensitivity vs Scenting

100 Handler 1 2

80

y

t 60

i

v

i

t

i

s

n

e 40

S

20

0

0 10 20 30 40 50 Scenting

Figure 6.2. Strongly correlated relationship between the proportion of time (%) spent ‘scenting’ and the dogs’ sensitivity scores with Handler 1 and 2.

Table 6.3. The dogs’ behaviour assessment scores for Friendliness (score/23), Excitability (score/23), Playfulness (score/17), Fearfulness (score/24), Aggressiveness (score/24), and Trainability (score/15). The dogs who performed well for both handlers are in bold.

Dog Friendliness Excitability Playfulness Fearfulness Aggressiveness Trainability 1 14 6 0 6 2 12 2 9 1 1 2 0 15 3 23 14 8 2 0 15 4 5 1 1 3 0 9 5 20 7 7 5 0 9 6 17 2 1 6 0 8 7 20 9 5 1 0 12 8 18 6 7 3 0 8 9 19 1 8 2 0 5 Mean 16.1 5.2 4.2 3.3 0.2 10.3

Discussion The variation between the dogs’ detection performances with Handler 1 and Handler 2 was significant. The dogs had significantly lower accuracy scores when handled by Handler 2. Whilst the dogs continued to improve throughout their testing with Handler 1, there was no improvement

La Toya Jamieson 88 with Handler 2. Due to the physical and dog handling experience similarities between the handlers, it can be assumed that this variation was at least in part influenced by Handler 1′s familiarity with the dogs. Familiarity may however, not be the only influential factor. As a result of training the dogs, it can be presumed that Handler 1 was also strongly bonded with the dogs, and they with her. This bond and frequent contact have previously been correlated with higher team performance (Haverbeke et al. 2008; Hoummady et al. 2016). As proposed by Horn et al. (2013a), familiarity may not be as influential as a strong dog-handler relationship on the dog’s performance or learning ability. This may also be the reason for the dogs’ lack of improvement with Handler 2. The fact that three of the nine dogs also refused to work for Handler 2 further supports this. The lack of improvement with Handler 2 is concerning for professional working dogs that are transferred between handlers. This is especially true when the transition period has time restrictions (e.g., two weeks until a detection team is fully operational (Haverbeke et al. 2010). It is likely that within a certain time the dogs would become familiar and eventually bond with Handler 2. However, there is no knowing the time this would take or the level of impact this would have on their performance. This acclimatisation time would also likely vary depending on the personalities of both the dogs and handlers. This requires further research.

There were little similarities between the dogs who performed well with both handlers. Of these three dogs, one was a male Border Collie and two were female Labrador Retrievers. We are not suggesting this was breed related, although further research is warranted. There were no consistent similarities in their behaviour scores. The only similarity between these dogs is their high scores with Handler 1. This indicates that perhaps dogs who are high performers will transfer more easily to a new handler. This was not consistent with all the dogs however, as there were other high performing dogs (Dogs 2 and 4) who performed very poorly for Handler 2. A larger sample size may be needed to highlight the characteristics of dogs that can adjust more quickly to new handlers, and this should be researched further.

Along with their differences in performance, the dogs also behaved differently with the two handlers. With Handler 2 the dogs demonstrated more stress-related behaviours and were significantly more ‘distracted’. The dogs’ lack of focus with an unfamiliar handler has been replicated in other studies (Horn et al. 2013a; Udell 2015). The dogs also spent less time ‘scenting’ with Handler 2, which was influential on the dogs’ performances. The dogs’ specificity scores were likely not impacted by their time spent ‘distracted’ or ‘scenting’ as they could achieve a score of 100 even if they weren’t scenting, simply by never indicating. This is demonstrated by Dogs 4, 5, and 8 all having sensitivity scores of 0 (meaning they never indicated to a target), but having

La Toya Jamieson 89 specificity scores of 98.4, 100, and 99.2, respectively. This highlights the importance of not evaluating dogs’ detection performances based solely on one calculation.

This study was limited by a small sample size. Due to the extended time commitment and resources needed to kennel and train the dogs, this could not be avoided. Even with this small sample size, the impact of the dog handler and the dog-handler relationship was clearly evident. This impact may not be as significant if professional working dogs and handlers were used, and this should be explored further. It is likely, however, that even professional working dogs are impacted by a change of handler, whether through their behaviours, and therefore their welfare, or working performance.

Conclusions This study has demonstrated the behavioural and performance impact of changing a dog’s handler. Our results may therefore not only have implications for detection dogs, but all dogs required to work closely with humans (e.g., assistance and herding dogs). Whilst each dog was impacted differently by this change, collectively the dogs responded negatively to the change of handler. It is unclear how long it would take for the dogs to adjust to a new handler and the best ways to manage this transition should be researched further. Whilst this research had a small sample size and only demonstrates how these specific dogs were affected by changing these specific handlers, our results and their possible outcomes are still significant and are further supported by the literature. This research highlights that whilst dogs are an incredible working partner for humans, they are not simple minded, easily transferable machines, and should not be managed as such.

Acknowledgments The authors acknowledge all the dog owners who lent their dogs to this study. Thank you to all the dedicated University of Queensland volunteers, and dog handlers who committed their time and expertise to this study. Thank you to D. Greenway for her recommendations for the statistical analyses. The first author would also like to acknowledge she received an Australian Government Research Training Scholarship to complete this research.

References

Complete reference list is provided at the end of the thesis due to reference repetition throughout the chapters.

La Toya Jamieson 90

Chapter 7: Can handlers be their dog’s lie detectors? Don’t believe those puppy eyes

Contributor Statement of contribution Conceived/designed study (100%) Behaviour coding (100%) La Toya Jamieson Data analyses (100%) Wrote the article (90%) Greg Baxter Wrote/edited the article (5%) Peter Murray Wrote/edited the article (5%)

“Arwen” correctly indicating to a sample

La Toya Jamieson 91

Abstract Detection dogs (Canis lupus familiaris) play numerous roles in human communities, however, like all detection systems they are not faultless. The indication of false positives and negatives reduces detection dog performance and can have serious consequences. The dog’s handler is crucial in minimising false indications. Analysing dogs’ smelling times, and smelling and indication behaviours, may assist handlers determine false from true indications. Ten dogs’ detection training sessions, totalling 1,089 minutes, were behaviour coded. The dogs’ smelling times were significant factors influencing their indications (p = 0.028). There were, however, significant differences between the dogs’ smelling times (p = 0.05). The dogs took significantly longer to perform a false positive indication than a true positive indication (p = 0.049). The dogs completed behaviours which were often specific, or performed more frequently, for true or false positives. For example, dogs completed ‘Paw-lifts’ significantly more for true positives than false positives (p = 0.016). Our research demonstrates that dogs’ smelling and indication times, and their associated behaviours, can assist handlers in discriminating between dogs’ true and false indications. Due to individual variation however, this information may need to be determined for individual dogs. Further research is required to determine the variation in sampling/smelling strategies dogs use in different scenarios and under varying environmental conditions. Our research, however, does suggest that handlers can become more proficient at evaluating the performance of their dogs if they become familiar with their dog’s particular mannerisms.

Introduction Dogs are used in a variety of working roles for scent detection, including narcotics (Jezierski et al. 2014), explosives (Furton & Myers 2001; Gazit & Terkel 2003; Goldblatt et al. 2009), human search and rescue (Greatbatch et al. 2015), cancer (Horvath et al. 2008; Sonoda et al. 2011), oestrus in animals (Johnen et al. 2015) and wildlife detection (Hurt & Smith 2009; Cristescu et al. 2015). Detection dogs have been shown to be the most accurate and efficient detection method for many applications, such as for wildlife detection (e.g. Long et al. 2007; Cristescu et al. 2015), and are widely used for civilian (e.g. search and rescue) and military operations (Sinn et al. 2010; Osterkamp 2011). Whilst there continue to be advances in detection technologies (Zhang et al. 2013), dogs are still more sensitive than artificial detection systems which are still fronting technological difficulties (Shelby et al. 2006; Bomers et al. 2012; Horvath et al. 2013; Liang et al. 2018). Detection dogs are not without fault, however, and can return both false positives (indicating presence when their target is absent) and false negatives (not indicating when their target is present). For this study a true positive is when the dog indicates presence when their target is present, and a true negative is when the dog does not indicate presence when their target is absent.

La Toya Jamieson 92

There is a large variation in detection dogs’ reported accuracy: 41% mean search success in detecting bladder cancer in urine (Willis et al. 2004); 76.4% search success for live humans (Greatbatch et al. 2015); and 91% search success for Desert tortoises (Gopherus agassizii; Cablk & Heaton 2006). Investigating an area where a dog has falsely indicated is fatiguing for the search team and a waste of resources, both time and financial (Greatbatch et al. 2015). Eliminating or reducing false positive and negative responses is therefore crucial in validating the detection dog methodology (Lehnert & Weeks 2016). Factors affecting detection dogs’ working ability include: the dog selected (Jamieson et al. 2017), their training (Hurt et al. 2016), environmental conditions (Wright & Thomson 2005; Greatbatch et al. 2015), and their handler, including the dog-handler relationship (Hurt & Smith 2009; Payne et al. 2015; Jamieson et al. 2018b).

The dog handler’s crucial role is reading their dog’s behaviours to not only best assist their search, but to determine the correctness of their indications (Greatbatch et al. 2015; Hurt et al. 2016). Dog handlers may reduce their team’s performance due to insufficient training, lack of handler motivation, handler error or introduction of response bias, leading to their dogs performing false positives (Lit et al. 2011; Browne et al. 2015). The most likely handler error causing false positives is rewarding their dog accidently on non-target samples (Hurt et al. 2016). In a real working environment it is difficult to know if the dog is indicating to the correct target based only on visual identification (Goldblatt et al. 2009), which makes rewarding the dog challenging. The likelihood of handlers inadvertently cueing their dog to falsely indicate is also related to stress levels (Zubedat et al. 2014), with team performance deteriorating when handlers are emotionally invested in their dog’s search outcomes (Browne et al. 2015). Dog handlers must understand their dog’s working behaviours and their subtle cues when nearing a target (Vice et al. 2009). Missed targets are often the result of dog handlers not recognising their dog’s behaviours, especially if their dog doesn’t indicate clearly (Vice et al. 2009). As dog handler experience and knowledge often impacts their dog’s performance (Lefebvre et al. 2007), methods to evaluate and assist novice dog handlers, in both training and field work, would be beneficial.

Sniffing (smelling) is the typical unit of active olfactory sampling (Youngentob 2005; Kepecs et al. 2006; Roux et al. 2006; Schoenfeld & Cleland 2006), and characteristically provides information for odour discrimination or identification (Uchida & Mainen 2003; Mainland & Sobel 2006). Smelling behaviours, including smelling frequency and duration, are directly related to scent detection and discrimination ability (Sobel et al. 2000; Verhagen et al. 2007; Concha et al. 2014). Previous research has been completed on dogs’ smelling times at detection tasks to determine if this was related to their indication (i.e. true positive and negative, and false positive and negative;

La Toya Jamieson 93

Concha et al. 2014). However, the dogs’ indication times (i.e. time taken to perform their indication behaviour) and smelling behaviours (i.e. behaviours completed when smelling samples, prior to indicating) weren’t investigated in that study. This information may further assist dog handlers in distinguishing whether their dog’s indications are correct. Improving our understanding of detection dog behaviour is crucial, especially in circumstances where incorrect indications can have negative consequences (e.g. explosives or disease detection).

This research therefore had several aims: (1) identify behaviours the dogs frequently performed prior to completing true and false positives, and true and false negatives; (2) measure time taken to smell target and non-target samples prior to indicating; and (3) determine if these smelling times were consistent when the handler did not know the sample order. We hypothesised that the dogs’ smelling times would be longer if they were uncertain about the sample (i.e. false positives and false negatives). Additionally the dogs would perform behaviours specific to the sample type.

Methods Research dogs Ten dogs were involved in this research, from three breeds (Table 7.1). Six were female and four were male, with a mean age of 3.4 ± 1.3 years. These dogs were involved in a detection training research project which compared the trainability of Border Collies, Labrador Retrievers and Greyhounds. The dogs were sourced from breeders and a rescue organisation. Whilst these dogs were not professional detection dogs they received extensive training for up to three months. The three month timeline was selected as it allowed the dogs time to learn to discriminate between their target sample and non-target samples, whilst still being financially feasible to house the dogs. The training completed attempted to mimic that received by professional detection dogs. The use of these dogs was approved by The University of Queensland’s Animal Ethics committee (approval permit: SAFS/454/16).

La Toya Jamieson 94

Table 7.1. The research dogs and their relevant information.

Dog Breed Sex Neuter status Age (years) 1 Border Collie Male De-sexed 6 2 Border Collie Female Entire 2 3 Border Collie Female Entire 4 4 Border Collie Male Entire 4 5 Labrador Retriever Female Entire 2.5 6 Labrador Retriever Male Entire 2 7 Labrador Retriever Female Entire 2 8 Labrador Retriever Female Entire 5 9 Greyhound Female Entire 3.5 10 Greyhound Male Entire 3

Training The training program utilised has previously been described in Jamieson et al. (2018b). The goal of training was to have the dogs independently locating their target odour (Bengal Tiger, Panthera tigris tigris, scat) amongst a range of non-target samples (cow, Bos taurus, and Brush-tailed phascogale, Phascogale tapoatafa, scat). This target odour was selected as we could guarantee it was novel and be sure the dogs would not encounter it again when returned to their owners. The dogs’ training sessions ran on average for five minutes, up to five days per week. All training sessions occurred outdoors in an open grassed area, under varying environmental conditions, to mimic field conditions. The same handler was used during all sessions and she was also the dogs’ trainer. The handler was not restricted in how she handled the dogs during this project. Volunteers were frequently used during training to randomise the order of, and deploy the training samples, thereby making the sample order unknown to the dog handler. During training the volunteers stood at least five meters away from the dog-handler team as they were working. The dogs were rewarded with both food and verbal praise upon indicating to a target sample.

Behaviour coding The dogs’ behaviours (Table 7.2) during their scheduled training sessions were coded using Behavioral Observation Research Interactive Software (BORIS; Friard & Gamba 2016). The person responsible for coding was the dogs’ trainer and was experienced in using BORIS. The focus of coding was to determine the mean time spent smelling the specific sample types (target and non- target samples) and the behaviours completed prior to the dogs’ indications. Smelling time was defined as when the dogs’ noses were over the sample containers, actively sniffing, and ended when their nose moved away from the sample (Concha et al. 2014). Due to the nature of this training all

La Toya Jamieson 95 sample types (i.e. true or false positives and negatives) were known to the person analysing the video footage. As the dogs were trained to recall to the sample, from a close distance (< 5 m) to re- indicate, these smelling behaviours weren’t included as the dogs had already been rewarded on that sample. Videos were slowed to a speed of 0.1 to improve the precision of coding, thereby maximising the validity of the results. Once coded, BORIS was used to create time budgets and plot behaviour events.

Table 7.2. Ethogram used for behaviour coding.

Behaviour Description True positive Time spent smelling a target sample – indication follows False positive Time spent smelling a non-target sample – indication follows True negative Time spent smelling a non-target sample – no indication False negative Time spent smelling a target sample – no indication Indication Time taken for a dog to perform indication behaviour on a target False indication Time taken for a dog to perform indication behaviour on a non-target Look to handler Dog turns head towards their handler Lower body Dog’s body posture sinks Lip lick Dog licks lips, not directly before or after receiving food Tail lowers Dog’s tail position drops in comparison to its’ normal tail carriage Tail lifts Dog’s tail position lifts in comparison to its’ normal tail carriage Ears lower Dog’s ears lower or pin towards head Paw-lift Dog lifts one front leg Jumping Dog lifts at least two legs simultaneously off the ground, not necessarily onto a person Side gaze Dog looks to handler without turning head Vocalisation Barks or whines Distracted Dog goes off task, ignores command or freezes and focuses elsewhere Pace quickens Dog’s walking speed quickens after smelling a sample Pace slows Dog’s walking speed slows after smelling a sample Refuse walk Dog won’t move away from sample after smelling

Data analyses All statistical analyses were completed using Minitab 18, with no data variables being transformed. Two-sample t-tests were completed to determine the difference between the dogs’ mean smelling and indication times for true and false positives, and true and false negatives. Two-sample t-tests were also completed to determine if the dogs’ behaviours prior to indicating were different depending on the sample (i.e. true and false positives, and true and false negatives). To determine if the dogs smelling times were significant factors influencing their indications General Linear Models were constructed. General Linear Models were also constructed to determine if the dogs’ smelling

La Toya Jamieson 96 times were different when the handler did not know the sample order (i.e. when volunteers set up the samples). Lastly, Pearson correlations were completed to determine if there was a relationship between the dogs’ smelling times and the average performance of false indications (false positives and negatives). Statistical significance was set at P < 0.05. The mean (X) results for smelling times are also provided.

Results 1,089 minutes of video, from 242 training sessions, were coded. The smelling times and behaviours of 1,648 true positives, 3,278 true negatives, 246 false positives and 389 false negatives were recorded. The dogs’ smelling times were significantly different between the different sample types

(ANOVA: F3 = 3.42, p = 0.028, Figures 7.1 and 7.2). There were, however, significant differences between the dogs’ smelling times (ANOVA: F2 = 4.13, p = 0.05). The dogs’ gender did not affect their smelling times. Whilst the dogs’ mean smelling times for true positives (X + SE = 0.51 ± 0.17 seconds, N = 10) was slightly higher than false positives (X + SE = 0.42 ± 0.19 seconds, N = 10), there was no significant difference between them (Paired t test: t10 = 1.09, p = 0.292). There were, however, significant differences in the dogs’ smelling times for true negatives and false negatives

(Paired t test: t10 = -2.43, p = 0.029).

Figure 7.1. Frequency of smelling times for true and false positives, and true and false negatives.

La Toya Jamieson 97

Boxplot of True positiv, False positi, True negativ, False negati

0.9

0.8

0.7

0.6

a

t

a 0.5

time(seconds)

D

0.4 Smelling 0.3

0.2

0.1 TrueTrue positivepositive FalseFalse positivepositive TrueTrue nnegativeegative FalseFalse nnegativeegative

Figure 7.2. Mean smelling times for true and false positives, and true and false negatives.

The dogs’ smelling times for true negatives were significantly impacted when the dog handler didn’t know the sample order (ANOVA: F2 = 5.8, p = 0.028). The smelling times for true positives

(ANOVA: F2 = 0.00, p = 0.96), false positives (ANOVA: F2 = 0.00, p = 0.979) and false negatives

(ANOVA: F2 = 0.78, p = 0.388) were, however, not significantly different when the handler didn’t know the sample order.

There was a very weak, non-significant relationship between the dogs’ false positive smelling times and the performance of false positives in their training sessions (Pearson correlation: rs = -0.061, N = 10, p = 0.867). Similarly, there was also a very weak relationship between the dogs’ false negative smelling times and the performance of false negatives (Pearson correlation: rs = -0.319, N = 10, p = 0.37). As with smelling times, the dogs’ gender also did not impact their average performance of either false positives (ANOVA: F1 = 2.28, p = 0.175) or false negatives (ANOVA:

F1 = 2.16, p = 0.185).

Similarly to the dogs smelling times, the time taken to complete their indication was also different depending on the sample they were smelling. The time it took for the dogs to perform a false positive indication was significantly longer than the time taken to perform a true positive indication

(Paired t test: t11 = -2.21, p = 0.049; Figure 7.3).

La Toya Jamieson 98

Histogram of True positive indicaiton, False positive indication 9 True positive 8 False positive

7

6

y

c

n 5

e

u

q

e 4

r

F

3

2

1

0 0.3 0.4 0.5 0.6 0.7 0.8 Indication time (seconds)

Figure 7.3. The dogs’ mean times for completing true and false positive indications. The bars that are a mixture of both shades (blue and pink) demonstrate both true positive and false positive indication time frequencies.

The behaviours the dogs completed prior to their indications appeared dependent on the individual (Table 7.3). The dogs that completed ‘Paw-lifts’ prior to indicating completed this behaviour for a significantly higher proportion of true positives (X + SE = 29.5 ± 19.9%) than false positives (X +

SE = 4.6 ± 6.1%; Paired t test: t7 = 3.17, p = 0.016). Therefore, out of all the true positive indications completed, the dogs completed ‘Paw-lifts’ prior to indicating for 29.5 ± 19.9% of them. There were also consistencies in the dogs’ behaviours after completing true negatives and false negatives. Dogs were ‘Distracted’ after a significantly higher proportion of false negatives (X + SE

= 28.4 ± 20.2%) than after true negatives (X + SE = 3.29 ± 3.09%; Paired t test: t9 = -3.88, p = 0.004). It should also be noted that the dogs were only required to smell a sample twice prior to indicating for 4.2% of the samples, and this only occurred for true positives. This means that the dogs only had to inspect very few samples multiple times before indicating. Therefore, the majority of these behaviours were completed during the dogs’ first inspection of the sample.

La Toya Jamieson 99

Table 7.3. Behaviours (B) performed during or directly after smelling the samples. P is the percentage of these behaviours that were completed for each indication type (e.g. True positive or true negative).

True positives True negatives False positives False negatives Dog B P B P B P B P Look to Look to Look to 87.5 7.9 86.6 Distracted 63.6 handler handler handler 1 Ears Refuse to Look to 1.2 6.6 18.2 Tail lifts 0.4 lower walk handler Distracted 0.8 Distracted 3.3 Side gaze 18.2 Look to Look to 93.7 3.3 Distracted 18.4 handler handler Look to 2 94.7 handler Look to Distracted 1.1 Distracted 1.2 10.5 handler Look to Look to 96.2 Look to Look to 57.1 3 handler 6.3 100 handler handler handler Paw-lift 18.9 Distracted 14.2 Look to Look to Look to 84.0 13.4 69.0 Distracted 38.7 handler handler handler 4 Paw-lift 6.3 Lip lick 2.4 Look to Distracted 6.0 28.8 Distracted 4.0 Paw-lift 2.4 handler Look to Look to Look to 2.7 Look to 18.2 5 67.6 handler 71.4 handler handler handler Distracted 0.5 Distracted 18.2 Look to Look to Look to Look to 91.1 11.1 64.8 25.5 handler handler handler handler 6 Paw-lift 25 Distracted 4 Distracted 17.4 Paw-lift 5.4 Tail-lift 7.2 Side gaze 0.6 Side gaze 10.5 Look to Look to Look to Look to 31.4 4.7 60.0 43.7 handler handler handler handler 7 Ears Pace Paw-lift 20.9 Distracted 1.7 10.0 6.3 lower slows Look to Look to Look to 93.2 13.9 100 Distracted 41.8 handler handler handler 8 Look to Paw-lift 21.0 Paw-lift 16.6 31.3 Distracted 4.0 handler Distracted 2.3 Distracted 8.3 Lip lick 1.5 Look to Tail-lift 93.3 4.8 Tail-lifts 78.4 Distracted 16.7 handler Refuse Look to 9 Paw-lift 58.1 21.6 12.5 walk handler Distracted 4.8 Jumping 21.6 Jumping 21.9 Tail-lift 8.3 Paw-lift 8.1 Paw-lift 56.6 Distracted 9.9 10 Tail-lift 26.4 Look to Tail-lift 16.7 Distracted 54.5 4.5 Jumping 1.9 handler

La Toya Jamieson 100

Discussion This study had several aims, which were all achieved, including identifying behaviours dogs frequently perform before completing indications and measuring the time taken to smell and indicate to target and non-target samples. Our research hypotheses were also supported by our results. This study therefore provides information on the differences in dogs’ smelling times and behaviours specific to different sample types. Whilst there were differences between the dogs’ smelling times, there were no differences between these times based on the dog’s sex. Previous studies which reported dogs’ smelling times similarly determined that the dogs’ sex was not influential (Riach et al. 2017).

The dogs consistently smelled true negative samples for less time than other samples. The differences in smelling times at specific tasks has been demonstrated in previous research on dogs (Concha et al. 2014), and rats and mice (Rinberg et al. 2006; Brooks et al. 2017). It has been hypothesised that dogs spend greater time smelling an unfamiliar scent to better understand it (Riach et al. 2017). This is similarly demonstrated when dogs stare at humans for longer when they are demonstrating unfamiliar facial expressions to better understand this visual information (Yong & Ruffman 2015). Temporal integration is the concept that increased sampling (smelling) time increases olfactory performance (Gold & Shadlen 2007). When rats have increased sampling time, either voluntary or forced, their odour discrimination performance improves (Rinberg et al. 2006; Abraham et al. 2012). This depends on the difficulty of the search situation, however, with some studies reporting that rats didn’t increase sampling time in difficult discrimination tasks (Uchida & Mainen 2003). Previous research has also reported a positive correlation between the proportion of time dogs spend ‘scenting’ and their accuracy at a detection task (Jamieson et al. 2018b). Our research, however, did not determine a significant correlation between the dogs’ smelling times and their accuracy (i.e. performance of false positives or negatives). The pace at which the dogs were travelling when smelling, however, was not recorded. Research on ring-tailed coatis (Nasua nasua) reported that their olfactory detection of food was significantly influenced by their travelling speed, with detection distance increasing as speed decreased (Hirsch 2010). Further research would be beneficial on the impact the dog’s working speed has on their detection accuracy and performance.

The inclusion of the dog handler further complicates the interpretation of the dogs’ smelling times. An urgency-gating model has been developed where a signal of urgency hastens reaching the decision threshold (Cisek et al. 2009). This type of modelling, which relies on rapid evidence accumulation and decision making, is likely similar in odour processing (Wesson et al. 2008). Dogs may receive a ‘signal of urgency’ from their handler when working in the field, due to time

La Toya Jamieson 101 pressures, which may cause them to modify their typical smelling strategy. As Brooks et al. (2017) highlights, the sampling strategy used by an animal in one task may not represent their strategy in another task. It is therefore possible that this study’s research dogs may have employed different smelling times or strategies when worked under different conditions or at a different detection task. Dog handlers should therefore monitor their dogs’ behaviours and smelling times in different training and working scenarios to determine any variation.

Previous research on dogs’ smelling times has reported that their dogs required multiple smelling episodes at true positives, false positives and false negatives (Concha et al. 2014). In our study, however, the dogs only required to smell samples twice on very few occasions, and only for true positives. This may be due to differences in the dogs training (inside or outside training sessions) or the sample presentation (scent carousel or scent line) in comparison to other studies (e.g. Concha et al. 2014). As the dogs in our study only needed multiple smells for true positives, this is another way to determine if the dog’s indication is correct, at least for true and false positives. The differences in the dogs’ indication times also provides another way to determine the accuracy of a dog’s indication. It may have taken the dogs longer to complete false positive indications for similar reasons animals increase their smelling times during difficult tasks - as they are uncertain, and therefore not confident, in their indication. The differences in the dogs’ smelling and indication times may assist handlers discriminate between their dogs’ true and false indications, thereby improving the handler’s evaluation of detection accuracy.

Previous research has highlighted the impact of having a handler blind to the sample order during testing, and the importance of having the handler blind to the sample order early during training (Johnen et al. 2013, 2015). Handlers who know the target sample location often unintentionally signal to their dog with body language (Lit et al. 2011). This research therefore included training sessions where the handler was blind to the sample order throughout the dogs’ training. It was anticipated that during these sessions the dogs would have different smelling times to compensate for the lack of unintentional handler cues. This however, was not demonstrated in our results, with the dogs’ smelling times not differing except for true negatives. True negative smelling times may have increased during these sessions as the dogs were less confident due to the lack of cues from their handler. As the dogs’ smelling times didn’t differ in these handler blind training scenarios, except for true negatives, this provides a good indication of the dog’s smelling times in real working scenarios. This highlights the relevance and practical application of our results.

The behaviours the dogs completed whilst smelling or completing their indication behaviour further revealed differences often specific to the sample type (i.e. target or non-target sample). Some

La Toya Jamieson 102 behaviours were completed most commonly during true positives, and only infrequently for false positives. The performance of these behaviours could therefore further help handlers distinguish whether their dog’s indication was correct, or if the dog was completing a false positive or negative. Some dogs in this study (e.g. dogs 1, 2 and 5), however, completed minimal behaviours outside of looking at their handler after smelling the samples, regardless of the sample type. For dogs who similarly have minimal variation in their behaviours regardless of sample type, the differences in their smelling and indication times will be crucial in assisting their handlers determine the correctness of their indications. Based on our results, we recommend that handlers frequently film and review their dogs training or working sessions to determine their dog’s subtle behaviours (including displacement behaviours) and standard smelling times. These videos should be watched at a slowed speed, ideally the speed used during this study (x 0.1) or frame-by-frame (Siniscalchi et al. 2011; Concha et al. 2014), to increase the likelihood of noticing subtle behaviours. This especially may assist transferring dogs from one handler to another and may improve communication between the dog-handler team.

No detection system can be 100% accurate. Whilst our research did not use professional detection dogs, the dogs used were highly trained. Further research is needed, however, on dogs’ sampling strategies in different search environments and under varying environmental conditions. Our current findings however, may assist handlers better understand their dogs smelling times and behaviours, thereby improving detection accuracy and working performance. Detection success is therefore not only dependent on how well a dog uses its’ nose, but how well the handler uses their eyes.

Acknowledgements The authors acknowledge all the dog owners who loaned their dogs to this study and the dedicated University of Queensland volunteers who committed their time to this study. The first author would also like to acknowledge she received the Australian Government Research Training Scholarship to complete this research.

References

Complete reference list is provided at the end of the thesis due to reference repetition throughout the chapters.

La Toya Jamieson 103

Chapter 8: General Discussion and Conclusions Detection dogs have many benefits which is reflected in their diverse use, including narcotics, explosives and wildlife detection. Wildlife detection dogs’ high accuracy levels and ability to efficiently survey an area has resulted in their use in multiple countries for various wildlife species. Detection dogs have been reported to be the most effective survey method for multiple species (Harrison et al. 2002; Cablk & Heaton 2006; Long et al. 2007a), especially cryptic or endangered species which may not have been located without their use (Cristescu et al. 2015; Orkin et al. 2016). Whilst this monitoring method is not without fault, the limitations can be overcome through proper team training and management, and survey design. Detection dog team success is impacted by several factors, most notably: the dog and handler selected; the dog-handler relationship; and detection team management, including their training (Rebmann et al. 2000; Svartberg & Forkman 2002; Sinn et al. 2010; Hurt et al. 2016; Jamieson et al. 2017; Jamieson et al. 2018a, b). Whilst these factors are frequently highlighted as important, minimal research has been published. Detection dog selection and management is therefore commonly a result of personal preference rather than scientific reasoning (Hall et al. 2015; Minhinnick et al. 2016). This is concerning due to the significant negative impact incorrectly managing these factors can have. Incorrect team selection and management can dramatically reduce team performance and welfare. My research therefore determined crucial factors that can improve dog and handler selection and working dog management. Whilst this research was specific to wildlife detection dogs, this information can be applied to, and beneficial for, other detection dog fields.

Research aims and discussion My thesis aims were successfully achieved through separating the previously listed influential factors into research projects. Through the completion of these projects a significant amount of knowledge was collected which can directly benefit working dog selection and management. These projects included investigating detection dog and handler selection; comparing dog breed’s training progression and accuracy assessments; determining dogs’ abilities to work with multiple handlers; and measuring dogs’ smelling times and behaviours during odour discrimination tasks. The highly applied nature of this research and the significance of the results can improve how wildlife detection dogs are managed, thereby improving this methodology. The following sections will report the major research findings; if they supported my hypotheses and how they related to my aims (outlined in Chapter 1). The significance and relevance of this research will be highlighted and areas for future research will be proposed.

La Toya Jamieson 104

Aim 1 – Wildlife detection dog benefits and suitable individuals A critical literature review identified the benefits and limitations of the wildlife detection dog methodology (Chapter 2). Collating this information and comparing it to alternative survey methodologies was crucial in highlighting the success and benefits of wildlife detection dogs. These benefits included high detection rates and survey efficiency in comparison to traditional methods. The main limitation of wildlife detection dogs were high survey costs and environmental factors (e.g. high air temperatures). These limitations are manageable however, and typically do not outweigh the overall benefits. This review highlighted the variation between dog team’s working success and efficiency. Because of this variation it is therefore important to evaluate each dog team prior to completing field surveys and not presume all teams perform similarly. Ensuring employed wildlife detection dog teams are working to their best capacity will maximise team success and survey results. Overall, the literature reports that wildlife detection dogs are a highly accurate and efficient method, with highly versatile applications.

Due to the relatively low success rate of working dog breeding programs, characteristics of suitable detection dogs were determined through reviewing the literature (Jamieson et al. 2017, Chapter 3). This review has made identifying a suitable detection dog a simpler process, with the recommendations provided based on scientific evidence and not personal preference. Dog breeds which typically possess these suitable traits were highlighted, and their different reported working accuracy was noted. Herding and gun dog breeds were highlighted as most likely possessing individuals with favourable characteristics, which included: a medium-sized, highly play/food motivated individual, that was obedient yet independent. It was highlighted that within breeds individual variation is prominent, therefore not all dogs within these breed types are likely suitable for detection work. The importance of the dog handler on team performance was also highlighted, which was further explored in Chapter 4 (Jamieson et al. 2018a) and Chapter 6 (Jamieson et al. 2018b). Collating this information will improve detection dog selection in the future, thereby improving the overall success of this method and its’ outcomes. Further research is needed to better develop assessments which determine individual dog’s suitability for working roles. This will likely improve the feasibility and welfare standards of working dog organisations. To increase the likelihood of detection dog success, however, the correct handler must also be selected (Jamieson et al. 2018a).

Aim 2 – Important dog handler traits Whilst dog handler importance is frequently highlighted in the literature, concerningly minimal research has been published on qualities and knowledge that are important for working

La Toya Jamieson 105 performance. My study therefore contributed to our knowledge of important wildlife detection dog handler traits and skills, which will assist with future handler selection and training (Jamieson et al. 2018a). From distributing questionnaires to wildlife detection dog handlers, the importance of specific traits and skills were highlighted. The dog handlers rated ‘ability to read dog body language’, ‘ability to trust in a dog’s indications’, ‘strong working ethic’, and ‘knowledgeable on dog behaviour’ as highly important qualities and skills (Jamieson et al. 2018a). It was anticipated that knowledge and experience in ecology would have been rated highly, given it was listed as important in the literature (Hurt et al. 2016). This was not revealed by my results. Knowledge on the dog’s target species was, however, listed as important. This information is crucial for future detection dog handler selection and training, as it highlights important skills which should be selected for and subjects where dog handlers may require further training.

On average, the wildlife detection dog handlers surveyed scored high in the Agreeableness domain, low in the Neuroticism domain, and average in the Extraversion, Conscientiousness, and Openness domains (Jamieson et al. 2018a). Similar results were reported in a previous study (Kaleta et al. 2011). There was, however, a large range in the handlers’ personality scores (Jamieson et al. 2018a). A person’s suitability for wildlife detection work may therefore be more a result of their dedication to training and their dog-handler bond, rather than their personality. This suggests that a personality assessment may not be a beneficial assessor of a person’s suitability for this work (Jamieson et al. 2018a). Whilst this research did not determine that there was a specific personality type best suited for this detection field, my findings did reveal important skills and knowledge required to work in this field. This knowledge is invaluable for selecting and training future wildlife detection dog handlers, thereby improving team success.

As the literature highlights the importance of the dog–handler relationship (Zubedat et al. 2014; Payne et al. 2015; Hoummady et al. 2016), it was anticipated that all surveyed handlers would be very emotionally attached to their dogs. This was not demonstrated in my findings, where only 57% of handlers stated they were ‘very emotionally attached’ to their dogs. Determining if these highly bonded dog-handler teams have higher detection rates than less bonded teams would be beneficial. Further research would also be beneficial in tailoring training programs to individual dog and handler teams, based on their personalities and current knowledge. Determining this information may improve team working success and welfare standards, which is crucial for this monitoring method.

Lastly my research determined the species detection dogs are currently locating in Australia and New Zealand. These included mammals, birds, reptiles and insects, from both native and pest

La Toya Jamieson 106 species (Jamieson et al. 2018a). This has significantly improved our knowledge on the use of wildlife detection dogs in Australia and New Zealand, and further highlights the versatility of this method. Our improved knowledge of how wildlife detection dogs are currently being used may inspire more ecologists and researchers to use this unique monitoring method for novel purposes.

Aim 3 – Breed comparison Based on my review of important detection dog characteristics (Jamieson et al. 2017), three dog breeds were selected to compare their trainability at detection work. Detection dog selection is often focused on breed rather than the individual. This is likely an important factor contributing to the poor success rates of many breeding and working dog programs. My experiment was therefore completed to determine if the breeds most commonly used for detection work were the most suitable. The breeds selected were: (1) the breed with the most ideal physical and behavioural characteristics – Border Collies; (2) the breed with the least ideal physical and behavioural characteristics – Greyhounds; and (3) the breed most commonly being used for detection work – Labrador Retrievers. All Border Collies and Labrador Retrievers in this study completed training, whilst only one Greyhound did. These findings supported my hypothesis that Border Collies and Labrador Retrievers would be trained quicker than Greyhounds. The dogs’ training times weren’t, however, correlated with their detection accuracy scores. This was highlighted by the one successful Greyhound having higher sensitivity scores than two Labrador Retrievers and one Border Collie, and higher specificity scores than three Labrador Retrievers and two Border Collies. These results challenge the favouring of certain dog breeds for working purposes and highlight the individuality within breeds.

The large degree of individual variation within breeds, both during training and accuracy testing, was not anticipated. Previous studies have highlighted the degree of individual variation within breeds (Björnerfeldt et al. 2008; Barnard et al. 2017), including their odour discrimination ability (Lazarowski et al. 2015). Due to this study’s small sample size, however, broad conclusions cannot yet be drawn on specific dog breed’s trainability or working suitability. My results do, however, suggest that a dog’s breed is not the best predictor of their trainability or detection accuracy, which challenges current beliefs. Future research would be beneficial in comparing the performance of purposely bred detection dogs against those sourced from rescue organisations, and determining which avenue is the most effective and feasible option. As only the Greyhounds in this study were sourced from rescue organisations, this information could not be determined. This information would be highly valuable in assisting organisations determine where they should source working

La Toya Jamieson 107 dogs from, and both their benefits and limitations. Ensuring detection dog teams are properly managed after selection will directly impact their working performance.

Aim 4 – Impact of changing handlers Selecting suitable detection dogs and handlers does not guarantee working success - these teams must be compatible in order to be successful (Jamieson et al. 2017). How dog-handler teams are paired and managed will therefore directly impact team success. This aspect of working dog management, however, is under represented in the literature, which my research has addressed. Dogs trained at detection work were therefore assessed with both a familiar and unfamiliar handler to determine if this change would impact their detection accuracy or behaviours. Working dogs are commonly handled by multiple handlers, such as when transferred between handlers in the military (Haverbeke et al. 2010). Determining the impact this has on a dogs’ detection ability and welfare is crucial for working dog management.

The results support my research hypotheses, with the dogs having significantly higher detection accuracy and demonstrating significantly less stress-related behaviours with their familiar handler (Jamieson et al. 2018b). Whilst this change in the dogs’ behaviours and detection accuracy were suggested in the literature (Hoummady et al. 2016), no previous research had been published. My research has therefore significantly contributed to our knowledge and understanding of how dogs adapt to a change of handler. Whilst this study had only a small sample size, these results suggest that a single dog and handler team may result in better working performances and welfare. These findings challenge how some working dog organisations currently manage their dogs, with dogs often having multiple handlers throughout their working life. Determining how long it takes to have a dog working to their normal standard, once transferred to a new handler, would be highly beneficial future research. This may allow dogs to work for multiple handlers whilst maintaining the same detection accuracy and not compromising their welfare. More research would therefore be beneficial on managing these transitions, such as training programs which help develop dog-handler communication.

Aim 5 – Smelling times and behaviours Improving handler understanding of their dog’s smelling times and working behaviours can improve detection accuracy or help establish when a dog has acclimatised after transferring to another handler. These behaviours are often subtle (Vice et al. 2009) and can easily be missed or misinterpreted by a new or novice handler. Whilst a dog’s behaviour is often the point of focus during detection work, minimal research has been completed on dogs smelling or indication times and how this may be related to their detection accuracy (Concha et al. 2014). My project therefore

La Toya Jamieson 108 reviewed videos of dogs completing detection tasks under controlled conditions in the field and measured their smelling and indication times, and their associated behaviours. My results demonstrated that dogs’ smelling times and time spent performing their indication behaviour were different depending on whether the sample was a target or non-target. These results supported my hypothesis and demonstrate that smelling and indication times can be used to determine if a dog’s indication is true or false.

Individual variation in the dogs’ smelling and indication times, and their behaviours, were present. Handlers should therefore review their own training footage at a slowed speed to determine their dog’s smelling times and unique behaviours. This would be particularly beneficial for new dog- handler teams or novice handlers. Further research would be beneficial to determine how dogs smelling times adjust to different environmental conditions or habitats. My research has therefore demonstrated a relatively simple technique which may reduce the recording of false positives and negatives during field surveys, which is crucial for detection dog accuracy and survey success. This research has therefore directly contributed to improving our knowledge of dogs’ scenting behaviours, which can be applied to improving detection dog training and survey results.

Conclusions My research determined multiple ways to improve detection dog and handler selection, and improved our understanding of how different management practises impact dog team’s success and welfare. These projects explored several factors which had previously only been minimally researched, if at all. My research therefore has contributed significantly to the literature on working dogs. Whilst further research is required, my results highlight the importance of the dog-handler relationship and how improving the handler’s knowledge and skills can directly benefit team success. These results are therefore very applied and will benefit the use of detection dogs, particularly wildlife detection dogs.

In all aspects of my research individual variation was prominent. This was demonstrated in the dogs’ training times, even within breeds; their performance and behaviour with an unfamiliar handler; and their smelling times and behaviours at specific samples. This degree of individual variation is crucial to highlight as it can have repercussions for how detection dogs are being selected and managed. My research therefore directly challenges much of how working dogs are currently being selected and managed. Whilst further research is needed, my results highlight that working dogs should be treated on an individual basis and should not be defined solely by their breed. Adjusting working dog training and management to focus on the individual dog will improve training success and performance, thereby improving industry welfare and best practise standards.

La Toya Jamieson 109

Whilst our working partnership with dogs can be traced back centuries, that doesn’t mean how we view and manage our working dogs must remain the same. Changing these practises will not be a simple task. However, through continued research and education this is achievable and will benefit both working dogs and the people who rely on them.

La Toya Jamieson 110

References Abney, D 2009, Canine Tracking Guide: Training the All-purpose Tracker, i-5Publishing, Irvine.

Abraham, NM, Guerin, D, Bhaukaurally, K & Carleton, A 2012, ‘Similar odor discrimination behaviour in head-restrained and freely moving mice’, PLoS One, vol. 7, e51789.

Adamkiewicz, E, Jezierski, T, Walczak, M, Górecka-Bruzda, A, Sobczyńska, M, Prokopczyk, M, & Ensminger, J 2013, ‘Traits of drug and explosives detection in dogs of two breeds as evaluated by their handlers and trainers’, Animal Science Papers and Reports, vol. 31, pp. 205-17.

Adams, VJ, Evans, KM, Sampson, J & Wood, JLN 2010, ‘Methods and mortality results of a health survey of purebred dogs in the UK’, Journal of Small Animal Practice, vol. 51, pp. 512–24.

Alpak, H, Mutuş, R & Onar, V 2004, ‘Correlation analysis of the skull and long bone measurements of the dog’, Annals of Anatomy, vol. 186, pp. 323–30.

Appleby, DL, Bradshaw, JWS & Casey, RA 2002, ‘Relationship between aggressive and avoidance behaviour by dogs and their experience in the first six months of life’, Veterinary Record, vol. 150, pp. 434–38.

Arandjelovic, M, Bergl, RA, Ikfuingei, R, Jameson, C & Parker, M 2015, ‘Detection dog efficacy for collecting faecal samples from the critically endangered Cross River gorilla (Gorilla gorilla diehli) for genetic censusing’, Royal Society Open Science, vol. 2, pp. 1-14.

Arnott, ER, Early, JB, Wade, CM & McGreevy, PD 2014a, ‘Estimating the economic value of Australian stock herding dogs’, Animal Welfare, vol. 23, pp. 189-97.

Arnott, ER, Early, JB, Wade, CM & McGreevy, PD 2014b, ‘Environmental Factors Associated with Success Rates of Australian Stock Herding Dogs’, PLoS ONE, 9, e104457.

Barnard, S, Marshall-Pescini, S, Pelosi, A, Passalacqua, C, Prato-Previde, E, & Valsecchi, P 2017, ‘Breed, sex, and litter effects in 2-month old puppies’ behaviour in a standardised open-field test’, Scientific Reports, vol. 7, pp. 1-8.

Bartels, A, Martin, V, Bidoli, E, Steigmeier-Raith, S, Brühschwein, A, Reese, S, Köstlin, R & Erhard, M 2015, ‘Brachycephalic problems of pugs relevant to animal welfare’, Animal Welfare, vol. 24, pp. 327–33.

La Toya Jamieson 111

Batt, L, Batt, M, Baughley, J, & McGreevy, P 2008a, ‘The effects of structured sessions for juvenile training and socialization on guide dog success and puppy-raiser participation’, Journal of Veterinary Behavior: Clinical Applications and Research, vol. 3, pp. 199-206.

Batt, L, Batt, M, Baughley, J & McGreevy, P 2008b, ‘Factors associated with success in guide dog training’, Journal of Veterinary Behavior: Clinical Applications and Research, vol. 3, pp. 143-51.

Beauvais, W, Cardwell, JM & Brodbelt, DC 2012, ‘The effect of neutering on the risk of mammary tumours in dogs −a systematic review’, Journal of Small Animal Practice, vol. 53, pp. 314–22.

Beckman, JP 2005, ‘Carnivore conservation and search dogs: The value of a novel, noninvasive technique in the Greater Yellowstone Ecosystem’, in A Wondrak Biel (ed), Greater Yellowstone Public Lands: A century of discovery, hard lessons, and bright prospects, Yellowstone Centre for Resources, Wyoming, pp. 20-6.

Beebe, SC, Howell, TJ & Bennett, PC 2016, ‘Using scent detection dogs in conservation settings: A review of scientific literature regarding their selection’, Frontiers in Veterinary Science, vol. 3, pp. 1–13.

Bird, RC 1997, ‘An examination of the training and reliability of the narcotics detection dog’, Kentucky Law Journal, vol. 85, pp. 405–33.

Björnerfeldt, S, Hailer, F, Nord, M & Vilá, C 2008, ‘Assortative mating and fragmentation within dog breeds’, BMC Evolutionary Biology, vol. 8, pp. 28-39.

Bomers, MK, van Agtmael, MA, Luik, H, van Veen, MC, Vandenbroucke-Grauls, CM & Smulders, YM 2012, ‘Using a dog’s superior olfactory sensitivity to identify Clostridium difficile in stools and patients: proof of principle study’, British Medical Journal, vol. 345, e7396.

Bradshaw, JW, Goodwin, D, Lea, AM & Whitehead, SL 1996, ‘A survey of the behavioural characteristics of pure-bred dogs in the United Kingdom’, Veterinary Record, vol. 138, pp. 465–8.

Brook, SM, van Coeverdende Groot, P, Scott, C, Boag, P, Long, B, Ley, RE, Reischer, GH, Williams, AC, Mahood, SP, Hien, TM, Polet, G, Cox, N & Hai, BT 2012, ‘Integrated and novel survey methods for rhinoceros populations confirm the extinction of Rhinoceros sondaicus annamiticus from Vietnam’, Biological Conservation, vol. 155, pp. 59–67.

Brooks, J, Nicholas, J & Robertson, JJ 2017, ‘Task dependence of odor discrimination: choosing between speed and accuracy’, Journal of Neurophysiology, vol. 119, pp. 377-9.

La Toya Jamieson 112

Browne, CM, Stafford, KJ & Fordham, RA 2015, ‘The detection and identification of tuatara and gecko scent by dogs’, Journal of Veterinary Behavior: Clinical Applications and Research, vol. 10, pp. 496-503.

Brownell, DA & Marsolais, M 2002, ‘The brownell-Marsolais scale: a proposal for the qualitative evaluation of SAR/disaster K9 candidates’, Advanced Rescue Technology, vol. 5, pp. 57–67.

Buffington, CAT 2002, ‘External and internal influences on disease risk in cats’, Journal of the American Veterinary Medical Association, vol. 220, pp. 994–1002.

Cablk, ME & Heaton, JS 2006, ‘Accuracy and reliability of dogs in surveying for Desert Tortoise (Gopherus agassizii)’, Ecological Applications, vol. 16, pp. 1926-35.

Cablk, ME, Sagebiel, JC, Heaton, JS & Valentin 2008, ‘Olfaction-based detection distance: a quantitative analysis of how far away dogs recognize tortoise odor and follow it to source’, Sensors, vol. 8, pp. 2208-22.

Chambers, CI, Vojta, CD, Mering, ED & Davenport, B 2015, ‘Efficacy of scent-detection dogs for locating bat roosts in trees and snags’, Wildlife Society Bulletin, vol. 39, pp. 780-87.

Chesney, CJ 1997, ‘The microclimate of the canine coat: the effects of heating on coat and skin temperature and relative humidity’, Veterinary Dermatology, vol. 8, pp. 183–90.

Cisek, P, Puskas, GA & El-Murr, S 2009, ‘Decisions in changing conditions: the urgency-gating model’, Journal of Neuroscience, vol. 29, pp. 11560-71.

Clare, JDJ, Anderson, EM, MacFarland, DM & Sloss, BL 2015, ‘Comparing the costs and detectability of Bobcat using scat-detecting dog and remote camera surveys in Central Wisconsin’, Wildlife Society Bulletin, vol. 39, pp. 210-17.

Clarke, T, Cooper, J, Mills, D, 2013, ‘Acculturation: perceptions of breed differences in the behaviour of the dog (Canis familiaris)’, Human-Animal Interaction Bulletin, vol. 1, pp. 16–33.

Cobb, M, Branson, N, McGreevy, P, Lill, A & Bennett, P 2015, ‘The advent of canine performance science: Offering a sustainable future for working dogs’, Behavioural Processes, vol. 110, pp. 96– 104.

Concha, A, Mills, DS, Feugier, A, Zulch, H, Guest, C, Harris, R & Pike, TW 2014, ‘Using sniffing behaviour to differentiate true negative from false negative responses in trained scent-detection dogs’, Chemical Senses, vol. 39, pp. 749-54.

La Toya Jamieson 113

Coppinger, R & Coppinger, L 2001, Dogs: A Startling New Understanding of Canine Origins, Behaviour and Evolution, Scribner, New York.

Coppinger, R & Schneider, R 1995, ‘Evolution of working dogs’, in J Serpell (ed), The Domestic Dog: Its Evolution, Behaviour and Interactions with People, Cambridge University Press, Cambridge, pp. 21–47.

Cooper, R, Wang, C & Singh, N 2014, ‘Accuracy of trained canines for detecting bed bugs (Hemiptera: Cimicidae)’, Journal of Economic Entomology, vol. 107, pp. 2171–81.

Coren, S 1994, The Intelligence of Dogs: How Intelligent Is Your Dog, Headline Book Publishing, London.

Craven, BA, Neuberger, T, Paterson, EG, Webb, AG, Josephson, EM, Morrison, EE & Settles, GS 2007, ‘Reconstruction and morphometric analysis of the nasal airway of the dog (Canis familiaris) and implications regarding olfactory airflow’, Anatomical Record Advances in Integrative Anatomy and Evolutionary Biology, vol. 290, pp. 1325–40.

Craven, BA, Paterson, EG & Settles, GS 2010, ‘The fluid dynamics of canine olfaction: unique nasal airflow patterns as an explanation of macrosmia’, Journal of the Royal Society Interface, vol. 7, pp. 933–43.

Cristescu, RH, Foley, E, Markula, A, Jackson, G, Jones, D & Frère, C 2015, ‘Accuracy and efficiency of detection dogs: A powerful new tool for koala conservation and management’, Scientific Reports, vol. 5, pp. 8349–55.

Curb, L, Abramson, C, Grice, J & Kennison, S 2013, ‘The Relationship between Personality Match and Pet Satisfaction among Dog Owners’, Anthrozoös, vol. 26, pp. 395–404.

D’Aniello, B, Scandurra, A, Prato-Previde, E & Valsecchi, P 2015, ‘Gazing toward humans: A study on water rescue dogs using the impossible task paradigm’, Behavioural Processes, vol. 110, pp. 68–73.

Dahlgren, DKRD, Elmore, RD, Smith, DA, Hurt, A, Arnett, EB & Connelly, JW 2012, ‘Use of dogs in wildlife research and management’, in N Silvy (ed), Wildlife Techniques Manual, The Wildlife Society Inc., Washington, pp. 140-53.

De Bondi, N, White, JG, Stevens, M & Cooke, R 2010, ‘A comparison of the effectiveness of camera trapping and live trapping for sampling terrestrial small mammal communities’, Wildlife Research, vol. 37, pp. 456-65.

La Toya Jamieson 114 de Oliveira, ML, Norris, D, Ramĺrez, JFM., Peres, PHD, Galetti, M & Duarte, JMB 2012, ‘Dogs can detect scat samples more efficiently than humans: an experiment in a continuous Atlantic Forest remnant’, Zoologia, vol. 29, pp. 183-6.

Demant, H, Ladewig, J, Balsby, TJS & Dabelsteen, T 2011, ‘The effect of frequency and duration of training sessions on acquisition and long-term memory in dogs’, Applied Animal Behaviour Science, vol. 133, pp. 228–34.

DeMatteo, K, Rinas, M, Sede, M, Davenport, B, Argüelles, C, Lovett, K & Parker, P 2009, ‘Detection Dogs: An Effective Technique for Bush Dog Surveys’, Journal of Wildlife Management, vol. 73, pp. 1436–40.

DeMatteo, KE, Blake, LW, Young, JK & Davenport, B 2018, ‘How behaviour of nontarget species affects perceived accuracy of scat detection dog surveys’, Scientific Reports, vol. 8, p. 13830.

Diverio, S, Menchetti, L, Riggio, G, Azzari, C, Laboni, M, Zasso, R, Di Mari, W & Matteo Santoro, M 2017, ‘Dogs’ coping styles and dog-handler relationships influence avalanche search team performance’, Applied Animal Behaviour Science, vol. 191, pp. 67–77.

Duffy, DL, Hsu, Y & Serpell, JA 2008, ‘Breed differences in canine aggression’, Applied Animal Behaviour Science, vol. 114, pp. 441–60.

Duggan, JM, Heske, EJ, Schooley, RL, Hurt, A & Whitelaw, A 2011, ‘Comparing detection dog and livetrapping surveys for a cryptic rodent’, Journal of Wildlife Management, vol. 75, pp. 1209– 17.

Early, JB, Arnott, ER, Wade, CM & McGreevy, PD 2014, ‘Manual muster: a critical analysis of the use of common terms in Australian working dog manuals’, Journal of Veterinary Behaviour: Clinical Research, vol. 9, pp. 370–4.

Elbroch, ML & Evans, J 2012, ‘Animal scats, urine, and other scent-marking behaviours’, in ML Elbroch, M Kresky & J Evans (eds), Field Guide to Animal Tracks and Scat of California, University of California Press, Berkeley, USA, pp. 99-116.

Fadel, FR, Driscoll, P, Pilot, M, Wright, H, Zulch, H & Mills, D 2016, ‘Differences in trait impulsivity indicate diversification of dog breeds into working and show lines’, Scientific Reports, vol. 6, pp. 1–10.

Feaver, J, Mendl, M & Bateson, P 1986, ‘A method for rating the individual distinctiveness of domestic cats’, Animal Behaviour, vol. 34, pp. 1016–25.

La Toya Jamieson 115

Fels, L & Distl, O 2014, ‘Identification and validation of quantitative trait loci (QTL) for canine hip dysplasia (CHD) in German Shepherd dogs’, PLoS ONE, vol. 9, e39620.

Fischer-Tenhagen, C, Johnen, D, Heuwieser, W, Becker, R, Schallschmidt, K & Nehls, I 2017, ‘Odor perception by dogs: Evaluating two training approaches for odor learning of sniffer dogs’, Chemical Senses, vol. 42, pp. 435-41.

Fischer-Tenhagen, C, Wetterholm, L, Tenhagen, BA & Heuwieser, W 2011, ‘Training dogs on a scent platform for oestrus detection in cows’, Applied Animal Behaviour Science, vol. 131, pp. 63- 70.

Fleming, JM, Creevy, KE & Promislow, DEL 2011, ‘Mortality in north American dogs from 1984 to 2004: an investigation into age-, size-, and breed-related causes of death’, Journal of Veterinary Internal Medicine, vol. 25, pp. 187–98.

Frank, H & Frank, MG 1985, ‘Comparative manipulation-test performance in ten-week old wolves (Canis lupus) and Alaskan Malamutes (Canis familiaris): a Piagetian interpretation’, Journal of Comparative Psychology, vol. 99, pp. 266–74.

Fratkin, JL, Sinn, DL, Patall, EA & Gosling, SD 2013, ‘Personality consistency in dogs: a meta- analysis’, PLoS ONE, vol. 8, e54907.

Frederick, RI & Bowden, SC 2009, ‘The test validation summary’, Assessment, vol. 16, pp. 215–36.

Friard, O & Gamba, M, 2016, ‘BORIS: A free, versatile open-source event-logging software for video/audio coding and live observations’, Methods in Ecology and Evolution, vol. 7, pp. 1325–30.

Fryxell, JM, Sinclair, ARE & Caughley, G 2014, Wildlife ecology, conservation, and management, John Wiley & Sons, Chichester, West Sussex.

Fuller, AK, Sutherland, CS, Royle, JA & Hare, MP 2016, ‘Estimating population density and connectivity of American mink using spatial capture-recapture’, Ecological Applications, vol. 26, pp. 1125-35.

Furton, KG & Myers, LJ 2001, ‘The scientific foundation and efficacy of the use of canines as chemical detectors for explosives’, Talanta, vol. 54, pp. 487-500.

Gácsi, M, McGreevy, P, Kara, E & Miklósi, Á 2009, ‘Effects of selection for cooperation and attention in dogs’, Behavioral and Brain Functions, vol. 5, pp. 31–8.

La Toya Jamieson 116

Gagnon, S & Doré, FY 1992, ‘Search behavior in various breeds of adult dogs (Canis familiaris): object permanence and olfactory cues’, Journal of Comparative Psychology, vol. 106, pp. 58–68.

Galis, F, Van der Sluijs, I, Van Dooren, TJM, Metz, JAJ & Nussbaumer, M 2007, ‘Do large dogs die young?’, Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, vol. 308, pp. 119–26.

Gazit, I & Terkel, J 2003, ‘Domination of olfaction over vision in explosives detection by dogs’, Applied Animal Behaviour Science, vol. 82, pp. 65-73.

Glen, AS & Veltman, CJ 2018, ‘Search strategies for conservation detection dogs’, Wildlife Biology, vol. 2018, pp. 1-9.

Glen, AS, Anderson, D, Veltman, CJ, Garvey, PM & Nichols, M 2016, ‘Wildlife detector dogs and camera traps: a comparison of techniques for detecting feral cats’, New Zealand Journal of Zoology, vol. 43, pp. 127-37.

Glen, AS, Russell, JC, Veltman, CJ & Fewster, RM 2018, ‘I smell a rat! Estimating effective sweep width for searches using wildlife-detector dogs’, Wildlife Research, vol. 45, pp. 500-4.

Goddard, ME & Beilharz, RG 1984, ‘The relationship of fearfulness to, and the effects of sex, age and experience on exploration and activity in dogs’, Applied Animal Behaviour Science, vol. 12, pp. 267–78.

Gold, JI & Shadlen, MN 2007, ‘The neural basis of decision making’, Annual Review of Neuroscience, vol. 30, pp. 535-74.

Goldblatt, A, Gazit, I & Terkel, J 2009, ‘Olfaction and explosives detector dogs’, in WS Helton (ed), Canine ergonomics: The science of working dogs, CRC Press, Boca Raton, pp. 135-74.

Goodwin, KM, Engel, RE & Weaver, DK 2010, ‘Trained dogs outperform human surveyors in the detection of rare spotted knapweed (Centaurea stoebe)’, Invasive Plant Science and Management, vol. 3, pp. 113-21.

Göth, A, McLean, IG & Trevelyan, J 2003, ‘How do dogs detect landmines?’, Mine Detection Dogs, vol. 18, pp. 195-208.

Graham, LT & Gosling, SD 2009, ‘Temperament and personality in working dogs’, in WS Helton (ed), Canine Ergonomics: The Science of Working Dogs, CRC Press, London, pp. 63–81.

La Toya Jamieson 117

Greatbatch, I, Gosling, R & Allen, S 2015, ‘Quantifying Search Dog Effectiveness in a Terrestrial Search and Rescue Environment’, Wilderness & Environmental Medicine, vol. 26, pp. 327-34.

Greer, KA, Canterberry, SC & Murphy, KE 2007, ‘Statistical analysis regarding the effects of height and weight on life span of the domestic dog’, Research in Veterinary Science, vol. 82, pp. 208–14.

Gsell, A, Innes, J, de Monchy, P & Brunton, D 2010, ‘The success of using trained dogs to locate sparse rodents in pest-free sanctuaries’, Wildlife Research, vol. 37, pp. 39-46.

Gutzwiller, KJ, 1990, ‘Minimizing dog-induced biases in game bird research’, Wildlife Society Bulletin, vol. 18, pp. 351–6.

Hagell, SE 2010, Conserving Forest Connectivity for the Central American Spider Monkey (Ateles geoffroyi) in Southwestern Nicaragua, PhD Thesis, Northern Arizona University.

Hall, NJ, Glenn, K, Smith, DW & Wynne, CDL 2015, ‘Performance of Pugs, German Shepherds, and Greyhounds (Canis lupus familiaris) on an odor-discrimination task’, Journal of Comparative Psychology, vol. 129, pp. 237-46.

Hall, NJ, Smith, DW & Wynne, CDL 2014, ‘Effect of odor preexposure on acquisition of an odor discrimination in dogs’, Learning & Behaviors, vol. 42, pp. 144–52.

Handelman, B 2008, Canine Behavior—A Photo Illustrated Handbook, Dogwise Publishing, Washington, USA.

Harrison, RL 2006, ‘A comparison of survey methods for detecting bobcats’, Wildlife Society Bulletin, vol. 34, pp. 548-52.

Harrison, RL, Barr, DJ & Dragoo, JW 2002, ‘A comparison of population survey techniques for swift foxes (Vulpes velox) in New Mexico’, American Midland Naturalist, vol. 148, pp. 320-37.

Hart, BL & Eckstein, RA 1997, ‘The role of gonadal hormones in the occurrence of objectionable behaviours in dogs and cats’, Applied Animal Behaviour Science, vol. 52, pp. 331–44.

Haverbeke, A, Laporte, B, Depiereux, E, Giffroy, J & Diederich, C 2008, ‘Training methods of military dog handlers and their effects on the team’s performances’, Applied Animal Behaviour Science, vol. 113, pp. 110–22.

La Toya Jamieson 118

Haverbeke, A, Messaoudi, F, Depiereux, E, Stevens, M, Giffroy, J & Diederich, C 2010, ‘Efficiency of working dogs undergoing a new Human Familiarization and training program’, Journal of Veterinary Behavior: Clinical Applications and Research, vol. 5, pp. 112–9.

Helton, WS 2009, Canine Ergonomics: The Science of Working Dogs, CRC Press, Boca Raton.

Helton, WS, 2010, ‘Does perceived trainability of dog (Canis lupus familiaris) breeds reflect differences in learning or differences in physical ability?’, Behavioural Processes, vol. 83, pp. 315– 23.

Hemsworth, JL, Barnett, JL & Coleman, G.J 2009, ‘The integration of human-animal relations into animal welfare monitoring schemes’, Animal Welfare, vol. 18, pp. 335–45.

Herron, ME, Shofer, FS & Reisner, IR 2009, ‘Survey of the use and outcome of confrontational and non-confrontational training methods in client-owned dogs showing undesired behaviors’, Applied Animal Behaviour Science, vol. 117, pp. 47–54.

Hirsch, BT 2010, ‘Tradeoff between travel speed and olfactory food detection in ring-tailed coatis (Nasua nasua)’, Ethology, vol. 116, pp. 671-9.

Hollerbach, L, Heurich, M, Reiners, TB & Nowak, C 2018, ‘Detection dogs allow for systematic non-invasive collection of DNA samples from Eurasian lynx’, Mammalian Biology, vol. 90, pp. 42- 6.

Homan, HJ, Linz, G & Peer, BD 2001, ‘Dogs increase recovery of passerine carcasses in dense vegetation’, Wildlife Society Bulletin, vol. 29, pp. 292-6.

Horn, L, Range, F & Huber, L 2013a, ‘Dogs’ attention towards humans depends on their relationship, not only on social familiarity’, Animal Cognition, vol. 16, pp. 435–43.

Horn, L, Range, F & Huber, L 2013b, ‘The importance of the secure base effect for domestic dogs—Evidence from a manipulative problem-solving task’, PLoS ONE, vol. 8, e65296.

Horvath, G, Andersson, H & Nemes, S 2013, ‘Cancer odor in the blood of ovarian cancer patients: a retrospective study of detection by dogs during treatment, 3 and 6 months afterward’, BMC Cancer, vol. 13, p. 396.

Horvath, G, Järverud, GAK, Järverud, S & Horváth, I 2008, ‘Human Ovarian Carcinomas Detected by Specific Odor’, Integrative Cancer Therapies, vol. 7, pp. 76-80.

La Toya Jamieson 119

Horváth, Z, Dóka, A & Miklósi, A 2008, ‘Affiliative and disciplinary behavior of human handlers during play with their dog affects cortisol concentrations in opposite directions’, Hormones & Behaviors, vol. 54, pp. 107–14.

Hoummady, S, Péron, F, Grandjean, D, Cléro, D, Bernard, B, Titeux, E, Desquilbet, L & Gilbert, C 2016, ‘Relationships between personality of human-dog dyads and performance in working tasks’, Applied Animal Behaviour Science, vol. 177, pp. 42–5.

Houpt, KA & Willis, MB 2001, ‘Genetics of behaviour’, in A Ruvinsky & J Sampson (eds), The Genetics of the Dog, CAB International, Wallingford, pp. 371–400.

Hurt, A & Smith, DA 2009, ‘Conservation dogs’, in WS Helton (ed), Canine Ergonomics: The Science of Working Dogs, CRC Press, London, pp. 175–94.

Hurt, A, Woollett (Smith), DA & Parker, M 2016, ‘Training considerations in wildlife detection’, in J Ensminger, T Jezierski & LE Papet (eds), Canine Olfaction Science and Law, CRC Press, London, pp. 139-53.

Jakovcevic, A, Elgier, AM, Mustaca, AE & Bentosela, M 2010, ‘Breed differences in dogs’ (Canis familiaris) gaze to the human face’, Behavioural Processes, vol. 84, pp. 602–7.

Jamieson, LJ, Baxter, GS & Murray, PJ 2017, ‘Identifying suitable detection dogs’, Applied Animal Behaviour Science, vol. 195, pp. 1–7.

Jamieson, LJ, Baxter, GS & Murray, PJ 2018a, ‘Who’s a good handler? Important skills and personality profiles of wildlife detection dog handlers’, Animals, vol. 8, pp. 222-36.

Jamieson, LJ, Baxter, GS & Murray, PJ 2018b, ‘You are not my handler! Impact of changing handlers on dogs’ behaviours and detection performance’, Animals, vol. 8, pp. 176-87.

Janutta, V, Hamann, H & Distl, O 2006, ‘Complex segregation analysis of canine hip dysplasia in German Shepherd Dogs’, Journal of Heredity, vol. 97, pp. 13–20.

Jezierski, T, Adamkiewicz, E, Walczak, M, Sobczynska, M, Gorecka-Bruzda, A, Ensminger, J & Papet, E 2014, ‘Efficacy of drug detection by fully-trained police dogs varies by breed, training level, type of drug and search environment’, Forensic Science International, vol. 237, pp. 112-8.

Johnen, D, Heuwieser, W & Fischer-Tenhagen, C 2013, ‘Canine scent detection – fact or fiction?’, Applied Animal Behaviour Science, vol. 148, pp. 201-8.

La Toya Jamieson 120

Johnen, D, Heuwieser, W & Fischer-Tenhagen, C 2015, ‘How to train a dog to detect cows in heat – Training and success’, Applied Animal Behaviour Science, vol. 171, pp. 39-46.

Johnson, JA 2014, ‘Measuring thirty facets of the Five Factor Model with a 120-item public domain inventory: Development of the IPIP-NEO-120’, Journal of Research in Personality, vol. 51, pp. 78–89.

Jones, AC & Gosling, SD 2005, ‘Temperament and personality in dogs (Canis familiaris): A review and evaluation of past research’, Applied Animal Behaviour Science, vol. 95, pp. 1–53.

Jones, KE, Dashfield, K, Downend, AB & Otto, CM 2004, ‘Search-and-rescue dogs: an overview for veterinarians’, Journal of the American Veterinary Medical Association, vol. 225, pp. 854-60.

Kaleta, T, Pisula, W, Fiszdon, K & Kondrakiewicz, P 2011, ‘Individual differences in police dog handlers’, Polish Psychological Bulletin, vol. 42, pp. 52–5.

Kapfer, JM, Munoz, DJ & Tomasek, T 2012, ‘Use of wildlife detector dogs to study Eastern box turtle (Terrapene carolina carolina) populations’, Herpetological Conservation and Biology, vol. 7, pp. 169–75.

Kemp, TJ, Bachus, KN, Nairn, JA & Carrier, DR 2005, ‘Functional trade-offs in the limb bones of dogs selected for running versus fighting’, Journal of Experimental Biology, vol. 208, pp. 3475–82.

Kepecs, A, Uchida, N & Mainen, ZF 2006, ‘The sniff as a unit of olfactory processing’, Chemical Senses, vol. 31, pp. 167-79.

Kerley, LL & Salkina, GP 2007, ‘Using scent-matching dogs to identify individual amur tigers from scats’, Journal of Wildlife Management, vol. 71, pp. 1349-56.

King, T, Marston, LC & Bennett, PC 2012, ‘Breeding dogs for beauty and behaviour: why scientists need to do more to develop valid and reliable behaviour assessments for dogs kept as companions’, Applied Animal Behaviour Science, vol. 137, pp. 1–12.

Kis, A, Turcsán, B, Miklósi, Á & Gácsi, M 2012, ‘The effect of the owner’s personality on the behaviour of owner-dog dyads’, Interaction Studies, vol. 13, pp. 373–85.

Kohn, MH & Wayne, RK 1997, ‘Facts from feces revisited’, Trends in Ecology and Evolution, vol. 12, pp. 223-7.

La Toya Jamieson 121

Kotrschal, K, Schoeberl, I, Bauer, B, Thibeaut, A.-M & Wedl, M 2009, ‘Dyadic relationships and operational performance of male and female owners and their male dogs’, Behavioural Processes, vol. 81, pp. 383–91.

Krausman, PR 2002, Introduction to wildlife management: the basics, Prentice Hall, New Jersey.

Krebs, CJ 2006, ‘Mammals’, in W Sutherland (ed), Ecological Census Techniques, 2nd ed., Cambridge University Press, Cambridge, United Kingdom, pp. 351-69.

Kuhne, F, Hößler, JC & Struwe, R 2012, ‘Effects of human-dog familiarity on dogs’ behavioural responses to petting’, Applied Animal Behaviour Science, vol. 142, pp. 176–81.

Kydd, E & McGreevy, P 2017, ‘Sex differences in the herding styles of working sheepdogs and their handlers’, PLoS ONE, vol. 12, e0184072.

Lazarowski, L, Foster, ML, Gruen, ME, Sherman, BL, Fish, RE, Milgram, NW & Dorman, DC 2015, ‘Olfactory discrimination and generalization of ammonium nitrate and structurally related odorants in Labrador retrievers’, Animal Cognition, vol. 18, pp. 1255-65.

Lefebvre, D, Diederich, C, Delcourt, M & Giffory, J 2007, ‘The quality of the relation between handler and military dogs influences efficiency and welfare of dogs’, Applied Animal Behaviour Science, vol. 104, pp. 49-60.

Lefebvre, D, Giffroy, J & Diederich, C 2009, ‘Cortisol and behavioural responses to enrichment in military working dogs’, Journal of Ethology, vol. 27, pp. 255–65.

Lehnert, MP & Weeks, ENI 2016, ‘Trained dogs in insect detection’, in J Ensminger, T Jezierski & LE Papet (eds), Canine Olfaction Science and Law, CRC Press, London, pp. 321-33.

Leigh, KA & Dominick, M 2015, ‘An assessment of the effects of habitat structure on the scat finding performance of a wildlife detection dog’, Methods in Ecology and Evolution, vol. 6, pp. 745-52.

Ley, JM, Bennett, PC & Coleman, GJ 2009, ‘A refinement and validation of the monash canine personality questionnaire (MCPQ)’, Applied Animal Behaviour Science, vol. 116, pp. 220–7.

Liang, Z, Tian, F, Yang, SX, Zhang, C, Sun, H & Liu, T 2018, ‘Study on interference suppression algorithms for electronic noses: a review’, Sensors, vol. 18, pp. 1179-1205.

Lit, L, Schweitzer, JB & Oberbauer, AM 2011, ‘Handler beliefs affect scent detection dog outcomes’, Animal Cognition, vol. 14, pp. 387-94.

La Toya Jamieson 122

Long, RA, Donovan, TM, Mackay, P, Buzas, JS & Zielinski, WJ 2007a, ‘Comparing detection dogs, cameras, and hair snares for surveying carnivores’, Journal of Wildlife Management, vol. 71, pp. 2018-25.

Long, RA, Donovan, TM, Mackay, P, Zielinski, WJ & Buzas, JS 2007b, ‘Effectiveness of scat detection dogs for detecting forest carnivores’, Journal of Wildlife Management, vol. 71, pp. 2007- 17.

Lorenzo, N, Wan, T, Harper, RJ, Hsu, Y, Chow, M, Rose, S & Furton, KG 2003, ‘Laboratory and field experiments used to identify Canis lupus familiaris active odor signature chemicals from drugs explosives, and humans’, Analytical and Bioanalytical Chemistry, vol. 376, 1212–24.

Maejima, M, Inoue-Murayama, M, Tonosaki, K, Matsuura, N, Kato, S, Saito, Y, Weiss, A, Murayama, Y & Ito, S 2007, ‘Traits and genotypes may predict the successful training of drug detection dogs’, Applied Animal Behaviour Science, vol. 107, pp. 287–98.

Mainland, J & Sobel, N 2006, ‘The sniff is part of the olfactory percept’, Chemical Senses, vol. 31, pp. 181-96.

Manteca, X & Deag, JM 1993, ‘Individual differences in temperament of domestic animals review of methodology’, Animal Welfare, vol. 2, 247–68.

Maples, JL, Guan, L, Carter, NT & Miller, JD 2014, ‘A test of the International Personality Item Pool Representation of the Revised NEO Personality Inventory and Development of a 120-Item IPIP-Based Measure of the Five-Factor Model’, Psychological Assessment, vol. 26, pp. 1070–84.

Marder, AR, Shabelansky, A, Patronek, GJ, Dowling-Guyer, A & Segurson D’Arpino, S 2013, ‘Food-related aggression in shelter dogs: A comparison of behaviour identified by a behaviour evaluation in the shelter and owner reports after adoption’, Applied Animal Behaviour Science, vol. 148, pp. 150-6.

Marshall-Pescini, S, Passalacqua, C, Barnard, S, Valsecchi, P & Prato-Previde, E 2009, ‘Agility and search and rescue training differently affects pet dogs’ behaviour in sociocognitive tasks’, Behavioural Processes, vol. 81, pp. 416–22.

McGarrity, ME, Sinn, DL, Thomas, SG, Marti, CN & Gosling, SD 2016, ‘Comparing the predictive validity of behavioural codings and behavioural ratings in a working-dog breeding program’, Applied Animal Behaviour Science, vol. 179, pp. 82–94.

La Toya Jamieson 123

Meek, PD, Ballard, G & Fleming, PJS 2015, ‘The pitfalls of wildlife camera trapping as a survey tool in Australia’, Australian Mammalogy, vol. 37, pp. 13-22.

Mehrkam, LR & Wynne, CDL 2014, ‘Behavioral differences among breeds of domestic dogs (Canis lupus familiaris): current status of the science’, Applied Animal Behaviour Science, vol. 155, pp. 12–27.

Mehus-Roe, K 2009, The Original Dog Bible: The Definitive Source for All Things Dog, 2nd ed., BowTie Inc, New York.

Miklósi, Á 2015, Dog Behaviour, Evolution, and Cognition, Oxford University Press, Oxford.

Mills, LS, Citta, JJ, Lair, KP, Schwartz, MK & Tallmon, DA 2000, ‘Estimating animal abundance using noninvasive DNA sampling: promise and pitfalls’, Ecological Application, vol. 10, pp. 283- 94.

Minhinnick, S, Papet, LE, Stephenson, CM & Stephenson, MR 2016, ‘Training fundamentals and the selection of dogs and personnel for detection work’, in J Ensminger, T Jezierski, LE Papet (eds), Canine Olfaction Science and Law, CRC Press, London, pp. 155–71.

Mirkó, E, Kubinyi, E, Gácsi, M & Miklósi, Á 2012, ‘A preliminary analysis of an adjective based dog personality questionnaire developed to measure some aspects of personality in the domestic dog (Canis familiaris)’, Applied Animal Behaviour Science, vol. 138, pp. 88–98.

Moore, GE, Burkman, KD, Carter, MN & Peterson, MR 2001, ‘Causes of death or reasons for euthanasia in military working dogs: 927 cases (1993–1996)’, Journal of the American Veterinary Medical Association, vol. 219, pp. 209–14.

Murphy, JA 1995, ‘Assessment of temperament of potential guide dogs’, Anthrozoös, vol. 13, pp. 224–8.

Murphy, JA 1998, ‘Describing categories of temperament in potential guide dogs for the blind’, Applied Animal Behaviour Science, vol. 58, pp. 163–78.

Nadeau, CP & Conway, CJ 2012, ‘Field evaluation of distance-estimation error during wetland- dependent bird surveys’, Wildlife Research, vol. 39, pp. 311-20.

Nakash, J, Osem, Y & Kehat, M 2000, ‘A suggestion to use dogs for detecting Red Palm Weevil (Rhynchophorus ferrugineus) infestation in date palms in Israel’, Phytoparasitica, vol. 28, pp. 153- 5.

La Toya Jamieson 124

Nolan, RV & Gravitte, DL 1977, Mine-detecting Canines: A Summary Report, Army Mobility Equipment Research and Development Command, Fort Belvoir, VA, USA.

Nussear, KE, Esque, TC, Heaton, JS, Cablk, ME, Drake, KK, Valentin, C, Yee, JL & Medica, PA 2008, ‘Are wildlife detector dogs or people better at finding desert tortoises (Gopherus agassizii)?’, Herpetological Conservation and Biology, vol. 3, pp. 103-15.

O’Kelly, HJ, Evans, TD, Stokes, EJ, Clements, TJ, Dara, A, Gately, M, Menghor, N, Pollard, EHB, Soriyun, M & Walston, J 2012, ‘Identifying conservation successes, failures and future opportunities: Assessing recovery potential of wild ungulates and tigers in eastern Cambodia’, PLoS ONE, vol. 7, e40482.

Ohl, F, Arndt, SS & van derStaay, FJ 2008, ‘Pathological anxiety in animals’, The Veterinary Journal, vol. 175, pp. 18–26.

Oldenburg, C, Schoon, A & Heitkönig, IMA 2016, ‘Wildlife detection dog training: A case study on achieving generalization between target odor variations while retaining specificity’, Journal of Veterinary Behavior: Clinical Applications and Research, vol. 13, pp. 34-8.

Oliveira, MLD, Norris, D, Ramírez, JFM, Peres, PHDF, Galetti, M & Duarte, JMB 2012, ‘Dogs can detect scat samples more efficiently than humans: an experiment in a continuous Atlantic Forest remnant’, Zoologia (Curitiba), vol. 29, pp. 183-6.

Orkin, JD, Yang, Y, Yang, C, Yu, DW & Jiang, X 2016, ‘Cost-effective scat-detection dogs: Unleashing a powerful new tool for international mammalian conservation biology’, Scientific Reports, vol. 6, e34758.

Osterkamp, T 2011, ‘K9 water searches: scent and scent transport considerations’, Journal of Forensic Science, vol. 56, pp. 907-12.

Ostrander, E & Wayne, RK 2005, ‘The canine genome’, Genome Research, vol. 15, pp. 1706–16.

Packer, RMA, Hendricks, A & Burn, CC, 2012, ‘Do dog owners perceive the clinical signs related to conformational inherited disorders as ‘normal’ for the breed? A potential constraint to improving canine welfare’, Animal Welfare, vol. 21, pp. 81–93.

Palika, L 2007, The Howell Book of Dogs: the Definitive Reference to 300 Breeds and Varieties, Wiley Publishing, New Jersey.

La Toya Jamieson 125

Palmer, R & Custance, D 2008, ‘A counterbalanced version of Ainsworth’s strange situation procedure reveals secure-base effects in dog-human relationships’, Applied Animal Behaviour Science, vol. 109, pp. 306–19.

Papet, LE & Minhinnick, S 2016, ‘Training a statistically superior scent discrimination canine – Where trainer wisdom meets scientific validation’, in J Ensminger, T Jezierski & L E. Papet (eds), Canine Olfaction Science and Law, CRC Press, London, pp. 173-95.

Payne, EM, Arnott, ER, Early, JB, Bennett, PC & McGreevy, PD 2015, ‘Dogmanship on the farm: Analysis of personality dimensions and training styles of stock dog handlers in Australia’, Journal of Veterinary Behavior: Clinical Applications and Research, vol. 10, pp. 271-8.

Pongrácz, P, Miklósi, Á, Vida, V & Csányi, V 2005, ‘The pet dogs ability for learning from a human demonstrator in a detour task is independent from the breed and age’, Applied Animal Behaviour Science, vol. 90, pp. 309–23.

Porritt, F, Mansson, R, Berry, A, Cook, N, Sibbald, N & Nicklin, S 2015, ‘Validation of a short odour discrimination test for working dogs’, Applied Animal Behaviour Science, vol. 165, pp. 133– 42.

Prato-Previde, E, Marshall-Pescini, S & Valsecchi, P 2008, ‘Is your choice my choice? The owners’ effect on pet dogs’ (Canis lupus familiaris) performance in a food choice task’, Animal Cognition, vol. 11, pp. 167–74.

Rauth-Widmann, B, 2006. Your Dog’s Senses −Understand How He Perceives His World, 2nd ed., Cadmos Equestrian, United Kingdom.

Rebmann, A, David, E & Sorg, MHH 2000, Cadaver Dog Handbook: Forensic Training and Tactics for the Recovery of Human Remains, CRC Press, Boca Raton, Florida.

Reed, SE, Bidlack, AL, Hurt, A & Getz, WM 2011, ‘Detection distance and environmental factors in conservation detection dog surveys’, Journal of Wildlife Management, vol. 75, pp. 243-51.

Reindl-Thompson, SA, Shivik, JA, Whitelaw, A, Hurt, A & Higgins, KF 2006, ‘Efficacy of scent dogs in detecting black-footed ferrets at a reintroduction site in South Dakota’, Wildlife Society Bulletin, vol. 34, pp. 1435-9.

Riach, AC, Asquith, R & Fallon, MLD 2017, ‘Length of time domestic dogs (Canis familiaris) spend smelling urine of gonadectomised and intact conspecifics’, Behavioural Processes, vol. 142, pp. 138-40.

La Toya Jamieson 126

Rinberg, D, Koulakov, A & Gelperin, A 2006, ‘Speed-accuracy tradeoff in olfaction’, Neuron, vol. 51, pp. 351-8.

Robertson, HA & Fraser, JR 2009, ‘Use of trained dogs to determine the age structure and conservation status of kiwi Apteryx spp. populations’, Bird Conservation International, vol. 19, pp. 121-9.

Rocznik, D, Sinn, DL, Thomas, S & Gosling, SD 2015, ‘Criterion analysis and content validity for standardized behavioural tests in a detector-dog breeding program’, Journal of Forensic Science, vol. 60, pp. 213–21.

Rohlf, VI, Bennett, PC, Toukhsati, S & Coleman, G 2010, ‘Why do even committed dog owners fail to comply with some responsible ownership practices?’, Anthrozoös, vol. 23, pp. 143–55.

Rohlf, VI, Bennett, PC, Toukhsati, S & Coleman, G 2012, ‘Beliefs underlying dog owners’ health care behaviors: Results from a large, self-selected, internet sample’, Anthrozoös, vol. 25, pp. 171– 85.

Rolland, RM, Hamilton, PK, Kraus, SD, Davenport, B, Gillett, RM & Wasser, SM 2006, ‘Faecal sampling using detection dogs to study reproduction and health in North Atlantic right whales (Eubalaena glacialis)’, Journal of Cetacean Research and Management, vol. 8, pp. 121-5.

Rooney, NJ & Bradshaw, JWS 2004, ‘Breed and sex differences in the behavioural attributes of specialist search dogs—a questionnaire survey of trainers and handlers’, Applied Animal Behaviour Science, vol. 86, pp. 123–35.

Rooney, NJ, Gaines, SA, Bradshaw, JWS & Penman, S, 2007, ‘Validation of a method for assessing the ability of trainee specialist search dogs’, Applied Animal Behaviour Science, vol. 103, pp. 90–104.

Rooney, NJ, Morant, S & Guest, C 2013, ‘Investigation into the value of trained glycaemia alert dogs to clients with type 1 diabetes’, PLoS One, vol. 8, e69921.

Roux, SG, Garcia, S, Bertrand, B, Cenier, T, Vigouroux, M, Buonviso, N & Litaudon, P 2006, ‘Respiratory cycle as time basis: an improved method for averaging olfactory neural events’, Journal of Neuroscience Methods, vol. 152, pp. 173-8.

Ruefenacht, S, Gebhardt-Henrich, S, Miyake, T & Gaillard, C 2002, ‘A behaviour test on German Shepherd dogs: heritability of seven different traits’, Applied Animal Behaviour Science, vol. 79, pp. 113–32.

La Toya Jamieson 127

Russell, JC, Beaven, BM, MacKay, JWB, Towns, DR & Clout, MN 2008, ‘Testing island biosecurity systems for invasive rats’, Wildlife Research, vol. 35, pp. 215-21.

Saetre, P, Strandberg, E, Sundgren, P, Pettersson, U, Jazin, E & Bergström, TF 2006, ‘The genetic contribution to canine personality’, Genes, Brain & Behavior, vol. 5, pp. 240–8.

Savidge, JA, Stanford, JW, Reed, RN, Haddock, GR & Yackel Adams, AA 2011, ‘Canine detection of free-ranging brown treesnakes on Guam’, New Zealand Journal of Ecology, vol. 35, pp. 174-81.

Schaefer, PS, Williams, CC, Goodie, AS & Campbell, WK 2004, ‘Overconfidence and the Big Five’, Journal of Research in Personality, vol. 38, pp. 473–80.

Schoeberl, I, Wedl, M, Bauer, B, Day, J, Moestl, E & Kotrschal, K 2012, ‘Effects of owner-dog relationship and owner personality on cortisol modulation in human-dog dyads’, Anthrozoös, vol. 25, pp. 199–214.

Schoenfeld, TA & Cleland, TA 2006, ‘Anatomical contributions to odorant sampling and representation in rodents: zoning in on sniffing behavior’, Chemical Senses, vol. 31, pp. 131-44.

Scott, JP & Fuller, JL 1965, Genetics and the social behaviour of the dog, The University of Chicago Press, Chicago.

Serpell, JA & Duffy, DL 2014, ‘Dog breeds and their behaviour’, in A Horowitz (ed), Domestic Dog Cognition and Behaviour, Springer Berlin Heidelberg, Berlin, pp. 31–58.

Serpell, JA & Hsu, Y 2005, ‘Effects of breed, sex, and neuter status on trainability in dogs’, Anthrozoös, vol. 18, 196–207.

Serpell, JA & Jagoe, JA 1995, ‘Early experience and the development of behaviour’, in J Serpell (ed), The Domestic Dog: Its Evolution, Behaviour and Interactions with People, Cambridge University Press, United Kingdom, pp. 79–102.

Shelby, RA, Myers, LJ, Schrader, KK & Klesius, PH 2006, ‘Detection of off-flavour in channel catfish (Ictalurus punctatus Rafinesque) fillets by trained dogs’, Aquaculture Research, vol. 37, pp. 299-301.

Silveira, L, Jácomo, ATA & Diniz-Filho, JAF 2003, ‘Camera trap, line transect census and track surveys: a comparative evaluation’, Biological Conservation, vol. 114, pp. 351-5.

La Toya Jamieson 128

Siniscalchi, M, Sasso, R, Pepe, AM, Dimateo, S, Vallortigara, G & Quaranta, A 2011, ‘Sniffing with the right nostril: lateralization of response to odour stimuli by dogs’, Animal Behaviour, vol. 82, pp. 398-404.

Sinn, DI, Gosling, SD & Hillards, S 2010, ‘Personality and performance in military working dogs: Reliability and predictive validity of behavioural tests’, Applied Animal Behaviour Science, vol. 127, pp. 51-65.

Slabbert, JM & Odendaal, JSJ 1999, ‘Early prediction of adult police dog efficiency—a longitudinal study’, Applied Animal Behaviour Science, vol. 64, pp. 269–88.

Slotta-Bachmayr, L & Schwarzenberger, F 2007, ‘Faecal cortisol metabolites as indicators of stress during training and search missions in avalanche dogs’, Veterinary Medicine Austria, vol. 94, pp. 110–7.

Smith, DA, Ralls, K, Davenport, B, Adams, B & Maldonado, JE 2001, ‘Canine Assistants for Conservationists’, Science, vol. 291, p. 435.

Smith, DA, Ralls, K, Hurt, A, Adams, B, Parker, M, Davenport, B, Smith, MC & Maldonado, JE 2003, ‘Detection and accuracy rates of dogs trained to find scats of San Joaquin kit foxes (Vulpes macrotis mutica)’, Animal Conservation, vol. 6, pp. 339–46.

Smith, DA, Ralls, K, Cypher, BL & Maldonado, JE 2005, ‘Assessment of scat-detection dog surveys to determine kit fox distribution’, Wildlife Society Bulletin, vol. 33, pp. 897-904.

Sobel, N, Khan, RM, Hartley, CA, Sullivan, EV & Gabrieli, JD 2000, ‘Sniffing longer rather than stronger to maintain olfactory detection threshold’, Chemical Senses, vol. 25, pp. 1-8.

Sonoda, H, Kohnoe, S, Yamazato, T, Satoh, Y, Morizono, G, Shikata, K, Morita, M, Watanabe, A, Kakeji, Y, Inoue, F & Maehara, Y 2011, ‘Colorectal cancer screening with odour material by canine scent detection’, Gut, vol. 60, pp. 814-9.

Sorge, RE, Martin, LJ, Isbester, KA, Sotocinal, SG, Rosen, S, Tuttle, AH, Wieskopf, JS, Acland, EL, Dokova, A, Kadoura, B, Leger, P, Mapplebeck, JCS, Mcphail, M, Delaney, A, Wigerblad, G, Schumann, AP, Quinn, T, Frasnelli, J, Svensson, CI, Sternberg, WF & Mogil, JS 2014, ‘Olfactory exposure to males, including men, causes stress and related analgesia in rodents’, Nature Methods, vol. 11, pp. 629–32.

La Toya Jamieson 129

Stevenson, DJ, Ravenscroft, KR, Zappalorti, RT, Ravenscroft, MD, Weigley, SW & Jenkins, CL 2010, ‘Using a wildlife detector dog for locating Eastern Indigo Snakes (Drymarchon couperi), Herpetological Review, vol. 41, pp. 437-42.

Svartberg, K 2002, ‘Shyness-boldness predicts performance in working dogs’, Applied Animal Behaviour Science, vol. 79, pp. 157–74.

Svartberg, K 2005, ‘A comparison of behaviour in test and in everyday life: evidence of three consistent boldness-related personality traits in dogs’, Applied Animal Behaviour Science, vol. 91, pp. 103–28.

Svartberg, K 2006, ‘Breed-typical behaviour in dogs − historical remnants or recent constructs?’, Applied Animal Behaviour Science, vol. 96, pp. 293–313.

Svartberg, K & Forkman, B, 2002, ‘Personality traits in the domestic dog (Canis familiaris)’, Applied Animal Behaviour Science, vol. 79, pp. 133–55.

Szetei, V, Miklósi, Á, Topál, J & Csányi, V 2003, ‘When dogs seem to lose their nose: An investigation on the use of visual and olfactory cues in communicative context between dog and owner’, Applied Animal Behaviour Science, vol. 83, pp. 141–52.

Turcsán, B, Kubinyi, E & Miklósi, Á 2011, ‘Trainability and boldness traits differ between dog breed clusters based on conventional breed categories and genetic relatedness’, Applied Animal Behaviour Science, vol. 132, pp. 61–70.

Uchida, N & Mainen, ZF 2003, ‘Speed and accuracy of olfactory discrimination in the rat’, Nature Neuroscience, vol. 51, pp. 351-8.

Udell, MAR 2015, ‘When dogs look back: Inhibition of independent problem-solving behaviour in domestic dogs (Canis lupus familiaris) compared with wolves (Canis lupus)’, Biology Letters, vol. 11, p. 20150489. van Rooy, D, Arnott, ER, Early, JB, McGreevy, P & Wade, CM 2014, ‘Holding back the genes: limitations of research into canine behavioural genetics’, Canine Genetics & Epidemiology, vol. 1, pp. 7–17.

Vas, J, Topál, J, Gácsi, M, Miklósi, Á & Csányi, V 2005, ‘A friend or an enemy? Dogs’ reaction to an unfamiliar person showing behavioural cues of threat and friendliness at different times’, Applied Animal Behaviour Science, vol. 94, pp. 99–115.

La Toya Jamieson 130

Verhagen, JV, Wesson, DW, Netoff, TI, White, JA & Wachowiak, M 2007, ‘Sniffing controls an adaptive filter of sensory input to the olfactory bulb’, Nature Neuroscience, vol. 10, pp. 631-9.

Vice, DS, Engeman, RM, Hall, MA & Clark, CS 2009, ‘Working Dogs: The last line of defense for preventing dispersal of Brown Treesnakes from Guam’, in WS Helton (ed), Canine Ergonomics: The Science of Working Dogs, CRC Press, London, pp. 195-204.

Vynne, C, Skalski, J, Machado, R, Groom, M, Jacomo, A, Marinho-Filho, J, Neto, M, Pomilla, C, Silveira, L, Smith, H & Wasser, S 2011, ‘Effectiveness of scat-detection dogs in determining species presence in a tropical savanna landscape’, Conservation Biology, vol. 25, pp. 154-62.

Wasser, SK, Davenport, B, Ramage, ER, Hunt, KE, Parker, M, Clarke, C & Stenhouse, G 2004, ‘Scat detection dogs in wildlife research and management: application to grizzly and black bears in the Yellowhead Ecosystem, Alberta, Canada’, Canadian journal of zoology, vol. 82, pp. 475-92.

Wasser, SK, Hayward, LS, Hartman, J, Booth, RK, Broms, K, Berg, J, Seely, E, Lewis, L & Smith, H 2012, ‘Using detection dogs to conduct simultaneous surveys of Northern Spotted (Strix occidentalis caurina) and Barred Owls (Strix varia)’, PLoS ONE, vol. 7, e42892.

Watkins, AF, McWhirter, JL & King, CM 2010, ‘Variable detectability in long-term population surveys of small mammals’, European Journal of Wildlife Research, vol. 56, pp. 261-74.

Wesson, DW, Carey, RM, Verhagen, JV & Wachowiak, M 2008, ‘Rapid encoding and perception of novel odors in the rat’, PLoS BIOLOGY, vol. 6, e82.

Wielebnowski, NC 1999, ‘Behavioral differences as predictors of breeding status in captive cheetahs’, Zoo Biology, vol. 18, pp. 335–49.

Wikenros, C, Jarnemo, A, Frisén, M, Kuijper, DPJ & Schmidt, K 2017, ‘Mesopredator behavioural response to olfactory signals of an apex predator’, Journal of Ethology, vol. 35, pp. 161-8.

Willis, MB, 1995, ‘Genetic aspects of dog behaviour with particular reference to working ability’, in J Serpell (ed), The Domestic Dog: Its Evolution, Behaviour and Interactions with People, Cambridge University Press, United Kingdom, pp. 51–64.

Willis, CM, Church, SM, Guest, CM, Cook, WA, McCarthy, N, Bransbury, AJ, Church, MR & Church, JC 2004, ‘Olfactory detection of human bladder cancer by dogs: proof of principle study’, British Medical Journal, vol. 329, pp. 1-6.

Wilsson, E & Sundgren, P 1997, ‘The use of a behaviour test for the selection of dogs for service and breeding, I: Method of testing and evaluating test results in the adult dog, demands on different

La Toya Jamieson 131 kinds of service dogs, sex and breed differences’, Applied Animal Behaviour Science, vol. 53, pp. 279–95.

Wright, GA & Thomson, MGA 2005, ‘Odor perception and the variability in natural odor scenes’, in JT Romeo (ed), Chemical Ecology and Phytochemistry of Forest Ecosystems, Elsevier Ltd, Oxford, pp. 191-226.

Yong, MH & Ruffman, T 2015, ‘Is that fear? Domestic dogs’ use of social referencing signals from an unfamiliar person’, Behavioural Processes, vol. 110, pp. 74-81.

Youngentob, SL 2005, ‘A method for the rapid automated assessment of olfactory function’, Chemical Senses, vol. 30, pp. 219-29.

Zhang, L, Tian, F, Dang, L, Li, G, Peng, X, Yin, X & Liu, S 2013, ‘A novel background interferences elimination method in electronic nose using pattern recognition’, Sensors and Actuators A: Physical, vol. 201, pp. 254-63.

Zhong, H, Levalle, RR, Moghaddam, M & Nof, SY 2015, ‘Collaborative intelligence −definition and measured impacts on internetworked e-work’, Management and Production Engineering Reviews, vol. 6, pp. 67–78.

Zubedat, S, Aga-Mizrachi, S, Cymerblit-Sabba, A, Shwartz, J, Fiko Leon, J, Rozen, S, Varkovitzky, I, Eshed, Y, Grinstein, D & Avital, A 2014, ‘Human-animal interface: The effects of handler’s stress on the performance of canines in an explosive detection task’, Applied Animal Behaviour Science, vol. 158, pp. 69–75.

La Toya Jamieson 132

Appendices

La Toya Jamieson 133

La Toya Jamieson 134

Questionnaire: Characteristics of the ‘ideal’ wildlife detection dog-handler and personality profiles of Australian and New Zealand detection dog-handlers Section 1 – Important dog-handler traits (1) Please indicate which of the following qualities and characteristics you agree are important for a wildlife detection dog-handler to possess, along with the qualities you agree you personally possessed when you first began wildlife detection dog work. The ratings are classed as follows:

Strongly Disagree Neither disagree nor Agree Strongly agree disagree agree 1 2 3 4 5

Qualities/Characteristics of dog-handler Important dog-handler Qualities you possessed traits (please rate 1-5) (please click the boxes)

High level of physical fitness/stamina ☐

Knowledgeable of canine olfactory physiology ☐

Skilled in dog handling ☐

Experienced in dog training ☐

Knowledgeable on dog behaviour ☐

Ability to read dog body language ☐

Theoretical background in ecology ☐

Practical ecological experience ☐

Sound knowledge of target species ☐

Ability to read wind direction ☐

Navigational skills ☐

Team player ☐

Strong leader ☐

Ability to trust in a dog’s indications ☐

Strong working ethic ☐

Skilled in report writing ☐

(2) Do you believe there are any other important characteristics or traits for a detection dog-handler to possess?

Personality:

La Toya Jamieson 135

Previous experiences/knowledge:

Section 2 – Personality assessment The following are a list of questions which are relevant to evaluating your personality type. Within the rows are statements about one’s personality, with all statements beginning with “I…”. Please indicate, by an ‘X’ in the appropriate box, how the statement describes you.

EXAMPLE ONLY Strongly Moderately Neither agree Moderately Strongly disagree disagree nor disagree agree agree Worry about things X

Please respond to the below statements with how you currently see yourself and not how you would like to see yourself. Please ensure you respond to every statement.

agree

disagree

Strongly Strongly DISAGREE Moderately DISAGREE Neither nor Moderately AGREE Strongly AGREE Worry about things

Make friends easily

Have a vivid imagination

Trust others

Complete tasks successfully

Get angry easily

Love large parties

Believe in the importance of art

Use others for my own ends

Like to tidy up

Often feel blue

Take charge

Experience my emotions intensely

Love to help others Keep my promises

La Toya Jamieson 136

Find it difficult to approach others

Am always busy

Prefer variety to routine

Love a good fight

Work hard

Go on binges

Love excitement

Love to read challenging material

Believe that I am better than others

Am always prepared

Panic easily

Radiate joy

Tend to vote for socially liberal political candidates

Sympathize with the homeless

Jump into things without thinking

Fear for the worst

Feel comfortable around people

Enjoy fantasy

Believe that others have good intentions

Excel in what I do

Get irritated easily

Talk to a lot of different people at parties

See beauty in things that others might not notice

Cheat to get ahead

Often forget to put things back in their proper place

Dislike myself

Try to lead others

Feel others’ emotions

Am concerned about others

Tell the truth

Am afraid to draw attention to myself

La Toya Jamieson 137

Am always on the go

Prefer to stick with things that I know

Yell at people

Do more than what’s expected of me

Rarely overindulge

Seek adventure

Avoid philosophical discussions

Think highly of myself

Carry out my plans

Become overwhelmed by events

Have a lot of fun

Believe that there is no absolute right or wrong

Feel sympathy for those who are worse off than me

Make rash decisions

Am afraid of many things

Avoid contact with others

Love to daydream

Trust what people say

Handle tasks smoothly

Lose my temper

Prefer to be alone

Do not like poetry

Take advantage of others

Leave a mess in my room

Am often down in the dumps

Take control of things

Rarely notice my emotions

Am indifferent to the feelings of others

Break rules

Only feel comfortable with friends

Do a lot in my spare time

La Toya Jamieson 138

Dislike changes

Insult people

Do just enough work to get by

Easily resist temptations

Enjoy being reckless

Have difficulty understanding abstract ideas

Have a high opinion of myself

Waste my time

Feel that I’m unable to deal with things

Love life

Tend to vote for socially conservative political candidates

Am not interested in other people’s problems

Rush into things

Get stressed out easily

Keep others at a distance

Like to get lost in thought

Distrust people

Know how to get things done

Am not easily annoyed

Avoid crowds

Do not enjoy going to art museums

Obstruct others’ plans

Leave my belongings around

Feel comfortable with myself

Wait for others to lead the way

Don’t understand people who get emotional

Take no time for others

Break my promises

Am not bothered by difficult social situations

Like to take it easy

Am attached to conventional ways

La Toya Jamieson 139

Get back at others

Put little time and effort into my work

Am able to control my cravings

Act wild and crazy

Am not interested in theoretical discussions

Boast about my virtues

Have difficulty starting tasks

Remain calm under pressure

Look at the bright side of life

Believe that we should be tough on crime

Try not to think about the needy

Act without thinking

Section 3 – Personal information (a) Age:

(b) Gender: Male Female (c) Country of residence: Australia New Zealand (d) Are you a professionally employed dog-handler or a volunteer (e) Years working as a dog-handler: (f) What dog breed do you most frequently use for conservation detection work:

(g) What target species do you handle/train the most dogs to detect:

(h) How emotionally attached are you to the dogs you’re currently handling (please click one below): NOT EMOTONALLY ATTACHED MILDLY EMOTIONALLY ATTACHED MODERATELY EMOTIONALLY ATTACHED VERY EMOTIONALLY ATTACHED (i) Do you believe your stress levels influence your dog’s behaviour? YES NO

END OF QUESTIONNAIRE.

La Toya Jamieson 140

Thank you very much for completing this questionnaire. Please e-mail your completed questionnaire to [email protected] A reminder: all of your personal details and answers are strictly confidential. A summary of this study’s findings will be sent to you via e-mail at the completion of the research. If you have further questions regarding this research or the research outcomes please contact me on the previously mentioned e-mail.

La Toya Jamieson 141