Mine Detect

7/30/2019 Mine Detect

1/66

Designer Dogs:

Improving the Quality

of Mine Detection Dogs


2/66

i

Designer Dogs:

Improving the Quality of

Mine Detection Dogs


3/66

ii

GICHD

The designations employed and the presentation of the material in this publication do not imply theexpression of any opinion whatsoever on the part of the Geneva International Centre for HumanitarianDemining concerning the legal status of any country, territory or area, or of its authorities or armed

groups, or concerning the delimitation of its frontiers or boundaries.

For furt her i nformat ion please contact:

the Geneva International Centrefor Humanitarian Demining

7bis, avenue de la PaixP.O. Box 1300

CH-1211 Geneva 1

SwitzerlandTel. (41 22) 906 16 60Fax (41 22) 906 16 90

www.gichd.ch

GICHD, Designer Dogs: Improving the Quality of Mine Detection Dogs, GICHD, Geneva, December2001

This publication has been written for the GICHD by Dr. Ian G. McLean, Research Analyst([email protected]).

The Geneva International Centre for Humanitarian Demining (GICHD) supports the efforts ofthe international community in reducing the impact of mines and unexploded ordnance.The Centre is active in research, provides operational assistance and supports the

implementation of the Mine Ban Treaty.


4/66

iii

Contents

Foreword 1

Executive summary 3

1. Introduction 5

2. Economics and ethics 9

3. Why dogs? 11

4. Designing an MDD 131. Origins 13

2. Genetics 145. The essential characteristics of an MDD 17

1. Why nose to ground? 172. Why is intensity of focus high? 183. Why are the actions repetitive? 184. Why is reliability high? 185. Physical demands and pace of work? 196. What is important about the enthusiasm of the dog? 19

6. Essential characteristics of an MDD: alternative options 211. Trainability 21

a. Definition 21

b. Perspectives on trainability 242. Presentation 24

a. Issues for handlers and communities 27b. The physical appearance of the scent hounds 28

3. Size and shape 294. The cognition question intelligence 30

The need for intelligence 32

7. Behavioural opportunities: behavioural featuresof MDDs 35

1. Types of MDD operational today 352. Behavioural requirements in operational MDDs 37


5/66

iv

Acknowledgements

This analysis has benefited from discussions with a wide range of people, includingthe members of the Mine Dog Standards Advisory Group, delegates at the MineDetection Dog meeting in San Antonio Texas, and many others with an interest indogs. The time spent by respondents to the survey is gratefully acknowledged. Specialthanks go to T. Berndtsen (NPA, Sarajevo), R. Fjellanger (NOKSH, Norway), D. Hayter(Global Training, Texas), B. Lewis (NPA, Angola), P.J. Martre (NOKSH) and NPAMozambique for support of visits made to their organisations. The report was reviewed

and improved by E. Cameron, A. Gth, members of the GICHD study group(C. Aakerblom, H. Bach, P. Blagden), and by members of the Advisory Group onStandards. It is impossible to identify the sources of the many ideas and commentsmade here, except to note that views were wide-ranging and often strongly held,

but there was considerable overlap as well.

Photo credits: p. 5 (machine, dog), pp. 12, 16, 32, 36 (figure 13), all H. Bach; p. 27Asiatic wolf, Eyal Bartov/Oxford Scientific Films; p. 27 northern wolf, p. 43figure 14 DK; all other photos by I. McLean.

8. Genetic opportunities existing genetic resources in dogs 39

9. Other issues 431. The operational situation for the designer MDD 43

2. Designing a handler 443. Maintenance training 444. Gender and aggression 455. The views of practitioners 46

10. Conclusion 49

References 53

Index 59

Glossary 60


6/66

1

Foreword

In many post-conflict societies, mined areas are a major obstacle preventing thereturn of refugees and the re-use of land for civilian purposes. Mines have anenormous negative social and economic impact. Despite significant investment in

technology, humanitarian demining remains a slow, costly and labour-intensiveprocess.

The use of dogs in humanitarian demining is one of the most promising avenues tohelp affected societies. Because of their remarkable ability to detect hidden objects

using odour, dogs are being used increasingly in the search for landmines. Dogs cancontribute substantively to efficient programmes in many ways: they can be used forarea reduction, to find individual mines, or to assure quality control of minefields cleared

by equipment.

Unlike a rescue dog, or a dog that is trained to detect drugs, the mine-detecting dogneeds to be one hundred per cent reliable. This is why the training of a mine detectiondog is an extremely difficult and time-consuming training problem. Any proceduresthat might improve the efficiency of the training process, of breed quality, or of theoperational potential of individual dogs, have significant implications for operationalcosts and productivity.

But no purpose-bred detection dog currently exists. Procedures for selecting dogsappropriate to the task are poorly developed, and the opportunities presented bydifferent breeds have not been widely explored.

Since 1998, the Geneva International Centre for Humanitarian Demining (GICHD) hasbeen engaged in a broad-based, application-oriented programme designed to exploremany aspects of the use of dogs for demining. The final aim of this work is to improvethe efficiency, safety and cost-effectiveness of humanitarian demining.

Work has been conducted in close and permanent co-operation with field practitioners.Studies are being conducted by the research team in many different places, includingAfghanistan, Angola, Bosnia and Herzegovina, and Cambodia. We are convinced thatthe present study will contribute to ongoing improvements in the use of dogs for


7/66

2 Designer Dogs

demining through diversification in choice of breeds, further development of trainingprocedures, and improved safety.

The GICHD would like to thank the following donors who provided project funding:the Governments of Germany, Japan (through the United Nations Trust Fund), Norway,Sweden, the United Kingdom, and the United States of America (both directly, andthrough the United Nations Office for Co-ordination of Humanitarian Affairs).

Ambassador Martin DahindenDirector

Geneva International Centre

for Humanitarian Demining


8/66

3

The use of dogs as a mine detection device is expanding rapidly. Thiswork identifies and discusses the essential and optional characteristicsof a mine detection dog (MDD). Each characteristic is defined and/or

described, and its features detailed. An extensive review of the literature on dogbehaviour and origins is provided, and is linked to the identified characteristics.Recommendations are made about alternative ways to design an MDD, and arelinked to current practice. Implementation of a broadened perspective on theselection and training of MDDs raises a variety of issues, which are identified

and discussed. It is clear that any such change will require an experimentalapproach.To date, however, there has been almost no investment in research onthe development and understanding of MDDs as a mine detection device, andsuch investment is justifiable.

Identified essential characteristics of an MDD and its use are:

The dogs nose is always to the ground;

Intensity of focus is high the dog is concentrating on the sensory inputto its nose and may even be snuffling or huffing in order to improve theflow of air through its nose;

The actions are highly repetitive (usually involving walking a carefully

prescribed narrow path); Reliability the dog exhibits a series of carefully contrived actions(including staying on command) with 100 per cent reliability;

The work is physically demanding;

The dog is moving slowly;

The dog wants to work it is enthusiastic about the job.

Characteristics for which alternative perspectives are discussed include:

Trainability;

Presentation overall appearance, human cultural and social issues;

Size and shape endurance, health issues;

Executive summary


9/66

4 Designer Dogs

Cognition intelligence;

Handler issues.

The possibility of breeding for an MDD is discussed in relation to genetic andbehavioural issues. Despite the widespread use of dogs in a detection role inmodern times, there has been no breeding for a designer detection dog. Rather,the dogs commonly used for detection purposes were bred for other uses, eitheras general purpose working or hunting dogs, or specifically as scent hounds.The advantages and disadvantages of each of these general types of dog arediscussed in relation to their potential for use in a mine detection role. Acontradiction is identified: most analyses in the literature indicate that heritabilityof behavioural characteristics is low, yet stabilisation of behavioural characteristicshas been achieved in many breeds using artificial selection procedures. Thus,selection for the behavioural characteristics sought in an MDD may be possible,

but must be viewed as a long-term objective.

An alternative to breeding is to develop procedures for selecting young dogssuitable as MDDs, using behavioural predictors. The literature is clear on thispoint: the behaviour of young dogs gives only low to moderate predictability of

behavioural outcomes of those individuals as adults. Thus, attempts to selectgood MDDs by testing young dogs will remain an inefficient process (althoughit is certainly used).

A review of the views of practitioners resulted in diverse views being expressed,but most agreed that breeds other than the standard shepherds should be trialled.Some indicated a view that other breeds were likely to be better than the standardshepherds, although that view was not based on experience.

The analysis identifies four alternative theoretical constructs as approaches todesigning an MDD, and asks if breeds already in existence exhibit most or all ofthe characteristics identified in those constructs, or packages. One package isclearly the breeds already used: a general purpose working dog with hightrainability for a wide variety of tasks (German shepherds and Malinois). Thesecond package exploits available breed characteristics that minimise trainingissues and optimise a number of other features linked to the operational situationin which many MDDs work. This package is represented by a slow-moving scenthound, and a common breed in Sweden (the drever) is identified as potentiallythe best available example. The third example is termed the cognitive dog,and identifies problem-solving and intelligence as central characteristics of thepackage. Currently no appropriate example exists, as dogs with the requiredlevel of intelligence do not fit the package for other reasons. The fourth exampleis a bred detection dog. No such purpose-bred dog currently exists, but theprinciples and techniques for its creation are known and production should bestraightforward the only limitation is motivation and resources.

The analysis concludes by recommending that a broadening of perspectives onthe selection, use and training of MDDs could improve safety and operationalcapacity in current practice. The investment required is realistic and supportable,and should be made.


10/66

5

The abundance of landmines,unexploded ordnance, and otherhidden explosive devices in many

parts of the world has sparked a massiveworldwide effort targeting the locationand clearance of these devices (henceforthcalled mines). Motivated primarily byhumanitarian concerns, clearance

operations are usually designed to return

impact of abandoned land on localenvironmental issues are also relevant.The social and economic issues arereviewed elsewhere (GICHD, 2001).

Mine detection operations use a toolbox(Figure 1). Inside that box are threecompartments containing: (i) technical

devices (which include an array of

1. Introduction

Figure 1. The demining toolbox.

land to the human communities that once

used it, although concerns about effects ondomestic animals and wildlife and the

detection and clearance tools), (ii) human

deminers working manually, and (iii)mine detection dogs (MDDs). All three can

MachinesMachines

Humans prodding the g roundHuma ns prodding the g round DogsDogs


11/66

6 Designer Dogs

be thought of as detection tools withparticular strengths and weaknesses. Thechoice of which to use in a particularminefield will be influenced by too manyfactors to list here, and in reality, manydemining organisations routinely use allthree in varying degrees.

Although their potential for minedetection was probably recognised soonafter the first mine was laid in the ground,dogs have only recently gained entry tothe toolbox. The compartment in whichthey sit is by far the smallest, andinvestment in research and development

(R&D) on dogs as a clearance device hasbeen virtually non-existent. In contrast,the mechanical clearance devicecompartment contains an extraordinaryvariety of vehicles and tools for remotesensing, floating detection devices abovethe ground, undertaking remotely-controlled ground disturbance, metal orchemical detection, and so on (see Abinashet al., 2000, for numerous examples). Manyof these tools are technically complex,impossibly expensive to purchase and

operate, difficult and costly to maintain,and are unlikely ever to be deployed inreal minefields. Yet the ongoinginvestment in R&D on these devices costsmillions of dollars annually.

MDDs are not a clearance tool, and theyare not an alternative to technical devices(whether used for detection or clearance).All three compartments in the toolbox arecomplementary and make importantcontributions to the demining objective.

The main advantages that both dogs andtechnical devices offer are improvementsin efficiency and safety for their humanoperators, and support for the primaryclearance tool humans prodding theground.

MDDs are not cheap to either produce ormaintain. At first deployment, an MDDtrained by a western organisation willhave cost in the vicinity of US$30,000.Ongoing maintenance costs for anoperational dog are more difficult to

estimate, but include direct costs such ashandler salary, and time and equipmentused for maintenance training, andindirect costs such as veterinary andkennel staff, food and housing,administration, security, specialisedtransport and downtime (e.g. due tosickness). Clearly, investment in R&D thatwill improve the operational efficiency ofany of the above cost factors is desirable.

One advantage of MDDs over all but themost simple of technical devices is theirpotential for local production essentially no industrial infrastructure is

required (in contrast to, for example, ametal detector). Local salaries and landcosts in most situations where MDDs aredeployed tend to be much lower than inthe western countries that currentlyprovide most MDDs. Thus, productioncosts may be considerably reduced if thedogs are bred and trained in theoperational theatre, as is the case inAfghanistan (where Hakimi, 1999,estimated the cost of producing anoperational dog to be US$1,198). However,

the training of an MDD requires an in-depth knowledge of the psychologicalprinciples underlying learning, andconsiderable sensitivity about, andunderstanding of, dog behaviour. Thus, itdemands an educational infrastructurethat is to be unavailable in mostoperational theatres.

Clearly, any adjustments to the dog thatreduce overall training requirements willfacilitate the possibility of local

production.

Although dogs have only recently gainedacceptance for humanitarian minedetection work, the use of dogs asdetection devices has a long history. AfterWorld War I, the British created a caninewar medal for valour to acknowledge thecontribution made by dogs to activitiessuch as finding injured soldiers (Fogle,2000), and dogs were used for minedetection work both during and afterWorld War II. For example, Finland was


12/66

7

so impressed with MDDs used by othercountries during World War II that itinitiated a training programmeimmediately after the war, and evenproposed to begin training smaller dogs,such as terriers, spaniels and schnauzers(Edstrvm, 2001). Wild dogs use their noseto search for prey, and this natural skillhas been exploited and refined in breedingprogrammes to produce domestic breedswith extraordinary scent and behaviouralskills. Although originally used for trailingprey (i.e. tracking), such dogs are nowroutinely used as search devices forpeople, drugs , foodstuffs, and explosives

intensity and single-mindedness ofpurpose that are desirable characteristicsin a modern detection dog. Unfortunately,they tend to exhibit other behaviouralcharacteristics that make them lesssuitable for modern detection needs, andonly a small number of scent hound

breeds are used regularly for detectionwork (the beagle is probably the mostpopular).

Many modern detection dogs are actuallysourced from working dog breeds, suchas retrievers and shepherds. Such dogs can

be termed generalists in that they tend

Figure 2. A drugs detection dog at work.

(Figure 2), and small numbers have beentrained in highly-specialised roles such asthe detection of cancers or pollutants.

Most bred search or detection dogs are

classed as scent hounds. They exhibitthe general characteristics of anextraordinary nose and remarkablestamina, and can be termed specialistsin that they were bred for a specificpurpose. Many of the hounds are pack-hunting and fast-moving, such asfoxhounds or greyhounds. But many ofthe scent hounds were bred to moveslowly so that a hunter travelling on footcould keep up with the dog, and to workalone rather than as part of a pack or team.Such dogs naturally exhibit the search

to be all-purpose dogs with highintelligence. They exhibit behaviouralcharacteristics of trainability and bondingwith humans that facilitate learning a widevariety of skills. In general, they are more

desirable as working companions, andmore rewarding to work with for trainersand handlers, than are the scent hounds.Their versatility and good nature areattractive features, and it is not surprisingthat they are the dogs of choice in asupport role for humans. A little moresurprising is that they are currently thepreferred breeds for use as MDDs.

It is provocative to suggest that workingbreeds are preferred as MDDs because ofhistorical precedence and conservatism

Introduction


13/66

8 Designer Dogs

amongst dog trainers, but this is a likelyexplanation (this conclusion has beenarrived at independently many times, andwas recently proposed publicly by Almey,2001). Dog training is as much an art as ascience, and experience is an importantpredictor of training outcome.Additionally, dog-people routinelydevelop personal breed preferences foressentially the same reasons that peopleprefer particular makes of car, orparticular types of breakfast cereal. Thepreference may have little to do with thefunction of the product, and much to dowith personal history. Thus, a trainer who

has worked with shepherds for 20 years

will be a better trainer of shepherds thanof beagles, and is also likely to prefer thecompany of shepherds. Most of thecurrent trainers of MDDs have a longhistory of working with general-purposesupport and protection dogs usuallyshepherds and sometimes retrievers. Suchpeople are not only likely to chooseshepherds and retrievers for the new task(mine detection) because of personalpreference, they are also likely to be betterat training those breeds than others. Theevidence is easily found in operationalMDD situations, where the dogs, almostall of which are shepherds, can be seen to

be both effective and productive.


14/66

9

Effective and productive devices canalways be improved upon. In the(relatively unusual) situation where

anything less than 100 per cent isunacceptable, as is the case with minedetection, even small improvements may

be highly significant. Two standard waysto make improvements to any product are

(i) to further develop the current productby refining its features, or (ii) to reviseradically the design of the current productwhile retaining its desirable features.Clearly, the second course of action will

be undertaken less commonly and isnecessarily more experimental and costly.However, the second option is the primaryconceptual target of this analysis. The firstoption is being addressed in separatereports (GICHD dog substudies 3 and 4,in preparation) and will only be discussed

in passing here.

An underlying theme in this analysis is thenotion that the MDD is a product, in thesame way that a mechanical clearancemachine or a metal detector is a product.In this context, there is nothing specialabout an MDD, even though it is a livingorganism held in high esteem in somehuman cultures. The primary issue ofconcern here is mine detection. A properlytrained and motivated MDD is an efficientmine detection device, and nothing more.

Its purchase and use will depend oneconomic realities, which in turn will bedriven by the local context in which mineclearance is being undertaken.

In western cultures at least, there has beenincreasing sensitivity about the supportrole of animals, and the extent to which

their needs must be considered whenhuman objectives are being pursued (e.g.Fraser, 1999). The use of animals in manyroles is now regulated by legislation, andanimal rights issues must be consideredin any operation that uses animals for itstask. Dogs used in a demining rolefrequently end up in countries in whichsuch protective legislation does not exist,and animal rights are not accorded therecognition they receive in westerncountries. Programmes may even be

handed over in entirety to local operatorsfor whom the notion of animal rights isincomprehensible. Clearly, local trainingon these issues is needed, both to fulfilllegislative requirements of the country inwhich the training organisation is based,and to ensure that the operational capacityof the dogs is protected.

Fortunately, the ethical issues arereasonably effectively addressed byeconomic and functional realities. Thequality of a product is reflected in its

2. Economics and ethics


15/66

10 Designer Dogs

original purchase price. Most users willtreat a quality product with care, and willmaintain it to a higher standard than acheaper, more disposable and replaceableoption. Thus, from an ethical perspective,the relatively high purchase cost of anMDD is a benefit, as it encourages the userto treat that dog with care and respect, andto optimise the maintenance of both itshealth and its work skills. Even withimproved breeding and trainingtechniques, it is unlikely that MDDs willever become a cheap item that can betreated as disposable, or mistreated,

because it is easily replaceable.

Any breeding programme developed inan attempt to produce a dog specificallyfor mine detection will necessarily face asignificant problem the production ofa large number of (possibly cross-bred)puppies, many of which are unsuitable forthe objective. Disposing of such dogs in aresponsible and ethically acceptablemanner may be difficult, and needs to beresolved as a logistical component of theprogramme. Currently, most MDDs are

sourced from breeding programmes forservice dogs, or from private breeders.These programmes produce dogs for amuch wider variety of roles and the rateof production of unwanted puppies is low.Tests of purpose-bred dogs for serviceroles suggest that reliability is improvedto some extent by a breeding programme(Goddard and Beilharz, 1983; Wilsson and

Sundgren, 1997; Slabbert and Odendaal,1999). A more focused breedingprogramme for MDDs might thereforeproduce dogs better suited to the task. Theproblem of unwanted puppies could beresolved by according top priority toproduction of MDDs in a programme thatproduces dogs for a broader range of uses.

Perhaps the biggest issue of concern forethics agencies is the question of what

becomes of MDDs that are no longerproductive. Ageing animals of anydomesticated species are most likely to beallowed to live out their lives if they have

developed a close relationship withparticular humans. Destruction orabandonment of relatively young animalsin the racing (horse or dog) business is acommon practice, in part because theanimals are only valued to the extent towhich they can race. Here, the animal issimply a product, and is likely to bedisposed of as soon as it is economicallyunproductive. The same problem maywell apply to MDDs, especially in thoseoperations where the dog-handler

relationship is restricted to a functionalworking relationship and the dog lives ina kennel.

Having recognised that the issues raisedabove are significant, and requirediscussion and resolution, they will beignored in the rest of this analysis.


16/66

11

Clearly, exploiting the odour givenoff by a mine in order to search forthat mine is a sensible option. Dogs

have been used for centuries in a widevariety of search and detection roles

because of their known olfactory skills, soit is not surprising that they would be usedin this relatively new role.

Dogs, however, are not the only animalswith an excellent sense of smell. Withrespect to searching for mines, there have

been recent attempts to train ferrets andthe African pouched rat (Figure 3). Animalspecies used in other roles by humans assniffing devices include ferrets (forrabbits) and pigs (for truffles). Males ofsome moths are able to detect the existenceof a female if just one molecule of theappropriate sex pheromone lands on his

sensors, and this skill is exploited in pestcontrol operations (Payne et al., 1986).Bees can locate sources of nectar fromdistances of hundreds of metres, and thisskill is exploited by the apiary industry.At least one attempt to develop bees asdetection devices is achieving somesuccess (Almey, 2001). Many animalsknown to have extraordinary smellingabilities are not exploited by humans inany way. For example, a bird withextremely sensitive olfactory skills (the

kiwi) spends its days (actually nights)

probing the ground and sniffing for prey such a lifestyle suggests it might be anideal candidate for development as anMDB (mine detection bird). It seems likelythat at least some of these examplesrepresent levels of odour sensitivity thatmatch or exceed the capabilities of dogs.

The problem is not one of locatingexamples, but of developing techniquesfor exploiting the known olfactorycapabilities of animals in a mine detectionrole, in a sensible time frame, and withminimum cost. For example, it is likelythat a few generations of careful selectionwith pigs would produce an animal ableto be used in mine detection work, butmany years would be required for such aprogramme. The image of genetically-engineered male moths flocking over a

buried mine because they believe it to bea female is alluring, but unlikely ever to

be more than a fantasy because of the costsand uncertainties involved. For both ofthese examples, the desired outcome couldnot be guaranteed even if the appropriateR&D was undertaken. On the other hand,such fanciful suggestions may havesimilar probabilities of success to some ofthe unlikely high-tech mine detectionsolutions that have proliferated in recentyears (numerous examples in Abinash et

al., 2000). In reality, it might be worth

3. Why dogs?


17/66

12 Designer Dogs

diverting a small portion of the massiveexpenditure on demining operations intoresearch on such flights of fancy. Theattempt with African pouched rats is justsuch an example.

Fundamentally, the answer to whydogs? is that this is a species that humansunderstand, and know how tocommunicate with. Because of a 12,000-year history of domestication (Morey,

Figure 3. The African pouched rat.Experiments with this species conducted by APOPO in Tanzania

suggest considerable potential as a tool for mine detection.

1994), dogs are motivated to work forhumans in a support role, and will do soreliably and consistently. Put simply, weknow how to explain the problem to a dog

better than for any other species. But itshould be kept in mind there is noimperative to use dogs, and imaginativeexperiments with other possible minedetection species should be supported andencouraged.


18/66

13

1. Origins

Although the modern science of geneticsdid not exist at the time, Darwin (1859)was the first to recognise that the knowntechnique of directing a genetic lineage

by a process of selection might have abiological basis. He therefore devoted

the first section of his famous bookto the process now called artificialselection.

Most breeders have specific designobjectives in mind when operating such aprogramme of selection, and theextraordinary diversity of dogs today isevidence of the potential inherent inthe genome of the original wolves fromwhich dogs are descended (Vila et al.,1997).

Many modern breeds can trace theirorigins to just one breeder, and even justone dog (e.g. the abundant Australiankelpie can be traced to a female calledKelpie; Fogle, 2000). Once a new breedhas been developed, the enthusiasts whoproduce it use stabilising selection to workto specific design characteristics (see

Jackson, 2000, for an overview of thetheory of dog breeding). If those standardsare linked to the show ring, then they mayhave no functional relevance whatsoever

and may not even be in the interests of thedog (Wolfensohn, 1981).

If the primary motivation driving selectionis functionality, as with many workingdogs, then the breed may be quite variablein form. But it will still exhibit the criticalfeatures that make it an effective working

dog. A frequently discussed modernexample is the American or pit bull terrier,which is so variable in form that attemptsin some countries to isolate or eliminatethis fighting breed have proved almostimpossible.

The relevance of these points to thedevelopment of an MDD is that, if thecharacteristics (or standards) of an MDDcould be agreed upon, then it should bepossible to produce a breed that offered

them. It is likely that slightly differentcharacteristics would be preferred indifferent places for example a short coatin hot countries and a wire coat in wetcountries. Alternatively, the reality for anMDD is that it could end up workingunder a wide variety of climaticconditions, and tolerance of variability inconditions might be the importantcharacteristic to breed for. Thus, acharacteristic could be a particularfeature, or something very general more a concept than a characteristic.

4. Designing an MDD


19/66

14 Designer Dogs

2. Genetics

In dogs, as in many other organisms,particular features may be controlled by arelatively simple genetic switch. Forexample, coat colour and type may becontrolled by either one or a small numberof genes. Such features tend to berelatively easy to manipulate geneticallyin order to stabilise their presence orabsence. Unfortunately, genetic control of

behavioural characteristics is routinelycomplex behaviour has been describedas the quintessentially complexphenotype, and is widely regarded as

be in g highly va riab le due to it sresponsiveness to environmentalinfluences (Wolf, 2001:117). For example,the stereotypical behavioural pattern ofpointing does not have a simple geneticswitch (MacKenzie et al., 1986). Generalcharacteristics that are essentiallyidentifiable concepts, such as trainability,intelligence, intensity or mobility, are theleast likely to have simple genetic controlsand are therefore potentially very difficultto manipulate genetically. Most published

research indicates that the heritability ofsuch features is low in dogs, although it isnot known whether this is because geneticvariation for them has been exhausted, or

because the genetic background to themis unresponsive to artificial selectionprocedures, (Scott and Fuller, 1965;Reuterwell and Rhyman, 1973; Goddardand Beilharz, 1983).

Despite these comments, heritability of ageneral behavioural feature (nervous-

ness) has been stabilised by artificialselection in dogs (Murphee et al., 1977),suggesting that the potential for geneticmanipulation of desirable behaviouralcharacteristics in an MDD is worthexploring (Goddard and Beilharz, 1983).In a recent unpublished study of about2,000 dogs, heritability of specific

behavioural features ranged up to aremarkable 0.6 on a scale of 0-1 (Bachem,2001). Clearly, there is more research to

be done on heritability issues.

It is now recognised that the behaviouralcharacteristics exhibited by adult animalsare a product of both evolutionary history(phylogeny) and personal experiences(ontogeny) (Adams et al., 2000). Forexample, Border collies have a longphylogenetic history of breeding forrounding up behaviour (usually ofsheep). Many Border collies raised as petshave never seen a sheep, yet they stillexhibit rounding up behaviour withoutany prompting usually targetting thefamily cat, the children, the chickens, oreven inanimate objects such as toys.With careful training, it is possible

(although difficult) to stop such a dog fromrounding up everything it encounters. Inthis example, individual experience(ontogeny) is used to counteractevolutionary history (phylogeny) in aprocess that could be called detraining(strictly, retraining). Detraining of bredskills is difficult, but any dog trainerknows that detraining of acquiredskills can be just as easily achieved asthe original training of those skills, ifpoor maintenance training techniques are

used.

The usual objective of training is todevelop and refine bred skills in order toimprove the efficiency and control of theirexpression. An artist may carve a beautifulobject from a piece of wood. But theoptions available as an end product will

be restricted by the original shape of thewood, the pattern of the grain, and theflow of colours. The dog trainer-as-artistmay similarly achieve remarkable

behavioural outcomes with a dog. But thebreeding of that dog will lend itself tosome outcomes, and not to others.

Two important reasons that Border colliesare the breed of choice for working withsheep are that they are highly motivatedto do the work, and they need relativelylittle training in order to become effectiveworking dogs. Both features are a productof breeding (or phylogeny). But bothrequire further personal development


20/66

15

(ontogeny) if the full potential of the dogis to be realised. Any breed of dog willrequire training in order to workeffectively with sheep, because everysituation in which sheep are being movedwill have its unique requirements. Ineffect, the dog must learn the languageused by the farmer to make thoserequirements clear. It is simply moreefficient to train a Border collie rather thansome other breed as a working sheep dog,

because both the ultimate quality of thedogs work and the time required to train

the dog, have been optimised by itsphylogeny.

If Border collies make better sheep dogsbecause of a history of select ion forcharacteristics that are desirable in a sheepdog, then a programme of selection for thecharacteristics desirable in an MDDshould have the same effect. What arethose characteristics, are they alreadyavailable, and how might they beobtained?

Designing an MDD


21/66

16 Designer Dogs


22/66

17

An observer of a good MDD at workquickly notices the followingfeatures:

The dogs nose is always to theground;

Intensity of focus is high the dogis concentrating on the sensory inputto its nose and may even be snufflingor huffing in order to improve theflow of air through its nose;

The actions are highly repetitive(usually involving walking acarefully prescribed narrow path);

Reliability the dog exhibits a seriesof carefully contrived actions(including staying on command)with 100 per cent reliability;

The work is physically demanding;

The dog is moving slowly;

The dog wants to work it isenthusiastic about the job.

1. Why nose to ground?

The answer to this question is not asstraightforward as it might seem. The doghas been trained to search for particularodours. If those odours are highly volatile,they might be more readily available inthe air above the ground than at the

surface. Or they might be more accessible

on vegetation (due perhaps to beingtransported from underground throughthe plants transport system). Through itsexperience searching for those odours, thedog will have learned where they are mostlikely to be found.

Research is still underway on the volatility

and mobility of molecules of TNT and theassociated chemicals likely to leak from anexplosive device, but the followingpreliminary comments can be made(summarised from Kjellstrm andSarholm, 2000, and Webb and Phelan,2000). TNT has extremely low volatility.Most of the associated chemicals in anexplosive compound (which are derivedfrom a mix of impurities, stabilisingchemicals, and breakdown products) havehigher volatility than TNT. The binding

properties of these molecules to soilparticles depend on soil particle size andcomposition, and soil moisture. Recentadjustments to soil moisture (e.g. rain aftera period of drought) can dramaticallychange volatility factors. The pattern ofdistribution of leakage products aroundan explosive device is erratic and maydepend on local topography (e.g. theremay be a plume along a moisturegradient). Overall, the volatility of allpotentially available odour sources is low

(in part because any odour sources with

5. The essentialcharacteristics of an MDD


23/66

18 Designer Dogs

high volatility have long since beenexhausted because of the time the minehas been in the ground). If volatility is lowfor an odour source hidden in the ground,then the most likely place to find chemicalresidues of that source will be at groundlevel, because most of the moleculesavailable to the dog are bound to dustparticles at the surface (George et al., 1999;Phelan, 2001). The dogs nose shouldtherefore be to the ground, and nose-to-ground is a learned response resultingfrom experience.

2. Why is intensity of focushigh?

Tired students quickly learn that it ispossible to read entire pages of textbookswithout comprehending any of theinformation contained on the page. It isvery easy to process the words. Itrequires considerably more concentrationto actually ingest the information thatthose words convey. The important pointmight be difficult to understand, or the text

might be poorly written (both are often thecase!). The central point may be containedin just one sentence somewhere on thepage, but there is nothing to identify thatsentence.

MDDs effectively face the same problem.They are searching for an elusive message a critical but tiny morsel of informationamongst a massive background of sensoryinput. As with the student, they do notknow which sentence (or which squarecentimetre of soil) contains the criticalmessage, so each square centimetre must

be scrutinised with equal intensity. Anyreduction in intensity of concentrationincreases the chance that the importantmessage will be missed, even though theground is apparently being covered. Thusintensity of concentration must always behigh, even though the critical message israrely encountered.

3. Why are the actionsrepetitive?

Safety concerns, and the need to ensurethat all of the ground is covered,necessarily require the dog to move in apredictable manner. The job actuallyinvolves endless repetition of a verysimple task in a rigorously controlled way.Search procedures that give the doggreater freedom of action (such as freeroaming throughout a search area) areunlikely to provide effective searchcoverage, even though they may make the

job more entertaining for the dog. An

important benefit of a repetitive searchpattern is that the dog becomesconditioned to undertaking the work in avery precise way including high levelsof search intensity and focus. Because thehandler is closely monitoring the link

between search zone and behaviour of thedog, they can choose to search again alonga path where the dogs concentrationlapsed.

4. Why is reliability high?

Any dog that behaves with less than 100per cent reliability in a minefield is adanger to itself and to its handler. In effect,the highly structured series of behavioursexhibited by the dog represents itsstandardoperating procedure (SOP). In adangerous situation, any variation on aSOP should only be made after carefulconsideration and thought, and probablysome experimentation. Such options arenot available to the dog.

As a training problem, reliability issues areusually referred to using the termobedience training. With respect to dogtraining, the GICHD has decided to avoidusing the words discipline andobedience, because their common usein the English language implies anassociation with unpleasant experiencesand might even justify training based on

punishment. Therefore, in this document,


24/66

19

the word reliability is used in place ofobedience.

5. Physical demands andpace of work?

Three features are relevant here: intensityof concentration, speed of action, and therepetitive nature of the task. All livingorganisms have a natural gait (or in mostcases, gaits), and those gaits naturally lendthemselves to particular speeds ofmovement. Some animals are actuallyphysically unable to move at some speeds

within their potential range because ofdesign constraints kangaroos are anexample because of energy conservationmechanisms built into their hopping gait.For horses, the natural gaits are walking,trotting, cantering and galloping. Formany medium to large sized dogs, thenatural walking pace of a human (5-6km/h) is an uncomfortable speed, leadingto pulling or dragging on a lead becausethe dog wants to walk (3-4 km/h) or lope(8-10 km/h). The loping run can be

maintained by dogs for hours, and it isoften the gait used by MDDs whilesearching for mines. Unfortunately, whilethe loping gait allows the dog to cover agreat deal of ground and work for longperiods, it is probably moving too fast toconduct an effective search over the entiresurface of the ground. To searcheffectively, the dog must move slowly byreducing its gait to a walk, or even a slowwalk. Dogs that do so naturally are likelyto be better MDDs.

Reasonably fit humans who have spent acouple of hours in a museum often reporton how exhausting the experience was,usually with some surprise because theydid not do much. The museumexperience includes all of the componentslisted above. The action of constantlystopping to review displays is repetitive,interpreting the displays requiresconcentration, and the gait and rate of

movement (a slow walk) are unnatural.

Put simply, humans are not designed tomove at a slow and erratic walking pacewith frequent stops, and they quickly tirewhen doing so. Similarly, dogs also tirequickly when required to walk slowly andwith frequent pace changes. Thus, the rateof movement of an intensively searchingMDD may be slow, but the pace andactivity are physically demanding.

6. What is important aboutthe enthusiasm of the dog?

Little needs to be said here. Any creature

undertaking a physically demanding andrepetitive activity is likely to work moreeffectively and for longer if it enjoys whatit is doing, where enjoys means that theactivity is intrinsically rewarding. Furthercomments are made in section 4 of thefollowing chapter (the cognition problem).The usual training solution with MDDs isto turn the search activity into a game,where the dog works for a highly-desiredreward (usually a toy).

Thus the essential characteristics of anMDD are:

A good nose;

High motivation for work (ormotivation easy to maintain);

High natural fitness with goodendurance (or these are easy tomaintain);

Tolerance of repetitive actions =

motivation to undertake simpletasks;

Ability to move at a slow pacefor sustained periods;

Tolerance of local environ-mental conditions;

Resistance to local diseases.

The last two characteristics are featuresthat improve the operational reliability ofthe dog in a particular context.

The essential characteristics of an MDD


25/66

20 Designer Dogs

It is worth noting that none of thesecharacteristics must be bred into the dog all can be established and maintainedwith time, training, veterinary support,and careful management. However, any

The Jewish cemetery in Sarajevo from which several thousand mines were

removed with the help of MDDs.

dog that requires a significant investmentin the maintenance of an essentialcharacteristic is unlikely to be a successfulMDD.


26/66

21

In the listing of essential characteristicsabove, the aim was to identify featuresabout which there should be little

disagreement in the MDD community.This section discusses characteristics thatare more controversial. The aim is toexplore those characteristics fromalternative perspectives, leading to the

possibility of creating different packages essentially alternative best designsfor an MDD. Much of the discussion isnecessarily speculative and is likely togenerate strong opinions among membersof the MDD community (and among dogtrainers generally). In the section above,the observable end product (i.e.operational MDDs) was used as the targetfrom which the characteristics weresourced. In this section each characteristicis discussed in general terms,

independently of the end product. Finally,suggestions are made about alternativeways to achieve the end product.

1. Trainability

a. Definition

No matter how well designed the basicMDD is, it will need training. Thus, somenotion of trainability is an essential featureof an MDD. The notion of trainability is

an intrinsic feature of any dog, and a rare

point of agreement on this vexed issue isthat different dogs will have differenttrainability. Defining the notion oftrainability is difficult, but is fundamentalto an effective analysis of the concept. If adefinition can be developed, then animportant step has been made towardsoptimisation of trainability in the designer

MDD.

Listed in Table 1 (following page) areseven ways in which the notion oftrainability is used. Together, theyprovide a preliminary attempt at definingthe concept. Clearly, there is overlap

between these points, but they serve asuseful topics to help focus the discussion.

Concert musicians exhibit extremelyfinely-honed musical skills, built on a

combination of extraordinary musicalskills (phylogeny) and practice(ontogeny). We normally refer to suchpeople as being highly talented, but wedo not normally think of them as beinghighly trainable. In reality, it is training,or practice, that turns a talented musicianinto a concert performer. Most concertmusicians are specialists on a particularinstrument, or more rarely on a smallnumber of closely-related instruments. Butattend a folk festival and you will find

individual musicians playing on an

6. Essential characteristics ofan MDD: alternative options


27/66

22 Designer Dogs

amazing array of instruments withapproximately equal and consummateskill. Possibly none of those instrumentsare played with the subtlety and flairachieved by a concert musician, but we arestill likely to refer to such generalistmusicians as being highly talented. Insome ways the diversity of their musicalskills makes them even more awesomethan the concert musician. Is one of these

musicians more talented than the other?Is that an inappropriate distinction?

Most people would automaticallyconsider the national obedience championto be a highly trainable dog. Clearly, sucha dog is very well trained, but was it highlytrainable? We are unlikely to refer to thedog as talented, probably reserving thatconcept more for the trainer than for thedog. Linking back to the metaphor above,the dog is the instrument and the trainer

is the musician. Instruments vary inquality (or playability), but surely theydo not have talent in any meaningfulsense. In fact, the more talented the trainer,the less necessary it may be for the dog to

be highly trainable. A good musician canmake even a bad instrument sound prettygood.

The obedience champion is most likely tobe a dog that exhibits finely tuned skillsin a range of behaviours (points 5 and 6 inthe trainability items above). On these

items at least, it will be considered highlytrainable. It will also have a very strongrapport with its handler (a point notusually considered when notions oftrainability are being discussed).However, the notion of potential(expressed in the first two points above)requires further clarification before thedog can be assigned a trainability indexon those measures. Discussion with the

trainer could indicate that the dog wasdifficult to train even though the ultimateproduct was impressive. Thus its potentialwas high, but its willingness to learn wasmore questionable. Just as with a concertmusician, this is a dog that required a greatdeal of practice for it to express its fullpotential. If the musician is talented, then,surely, so too is the dog.

In the three paragraphs above word gameswere used to develop at least two

different, and somewhat contradictory,perspectives on trainability. Otherarguments could easily be developed, butthe point has been made that trainabilityis not an easy concept to either define oragree upon. It is reasonable to assume thatdog trainers using the word trainabilityroutinely talk at cross-purposes. Forexample, Hakimi (1999) suggested thatMDDs need to be both talented and smart,Dyck (1999) indicated that MDDs should

be intelligent (presumably=smart) andshow aptitude for this work

Table 1. Different uses of the word trainability

The potential for a dog to acquire trained skills in general terms;

The potential for the dog to acquire particular trained skills;

The ease with which trained skills can be built into the dog by a training

programme;

The diversity of skills that can be acquired by the dog;

The complexity of skills that can be exhibited by the dog;

The refinements (or subtleties) that can be achieved in the final training

outcome;

The need for different training techniques, or at least the need for adjusting

the way training techniques are used, with each dog.


28/66

23

(=talented?), and Smith (1999) indicatedthat potential MDDs should be matureand stable. The courage listed by Dyckwas not mentioned by either Hakimi orSmith, and none of these three authorsmentioned trainability per se, althoughthey were presumably referring to thisconcept when using the other words.Lockwood (1999) used the wordtrainability along with stability. Theauthors all made their comments withinone conference setting, so were talking toeach other at the time. If some notion oftrainability is to be used in relation to thedesign of an MDD, it is essential that the

notion be explored in some detail if anyclarity of discussion is to be obtained.

Consider this sentence: trainability canbe ca tego ri sed bo th as a breedcharacteristic (e.g. there is reasonablygeneral agreement that the hounds aremore difficult to train than the workingdogs) and an individual characteristic(even within a litter, dogs will vary intrainability).

In contrast to the comment above, trainersof beagles (a hound) for fruit andvegetable detection work in airportsconsider them to be relatively easy to train.Beagles have a bred obsession with thenose and love of food, thus the trainingtask is relatively straightforward. The

beagle simp ly has to be taught thatsignalling the presence of certain odourswill result in a food reward, and theappropriate signal to give is to sit down.The beagle does not have to be taught to

sniff at bags (instead of, for example,watching the people milling around), orto maintain intensity of sniffing behaviour,

because it does these things naturally. Itdoes not have to be tricked into thinkingthat searching for fruit and vegetables is agame (in order to maintain motivation),nor does it need to develop an obsessionwith a convenient reward (such as a ball).It will need socialisation to accustom it tothe noise and chaos of an airport arrivalhall, but its natural tranquility means thatit easily adjusts to that potentially stressful

environment. Despite being hounds,perhaps beagles should be placed at thetrainable end of the trainability scale.

On the other hand, most owners of beagleswill go on at length about the difficulty ofdoing responsiveness training, thetendency of the dog to run away, itsconstant scavenging and stealing of food,and its dislike of water. There is no doubtthat some training objectives are difficultto achieve with beagles. If those objectivesare in mind, beagles will routinely beplaced at the untrainable end of thetrainability scale.

How does the beagle rate with respect tothe seven factors extracted from the notionof trainability? On the general potential foracquiring trained skills, beagles rate low.On the potential for acquiring particulartrained skills, beagles rate high (these arethe skills that beagles exhibit naturallyanyway). On the ease with which trainedskills can be built into the dog, beagles ratevariablebecause some are easy and someare difficult. On diversity of skills, beagles

rate low. On complexity of skills, beaglesrate low. On refinements in the finaltraining outcome, beagles rate low. On theneed for adjusting training with eachindividual dog, it is enough to say that thegenerally good nature of beagles meansthat they are probably at the higher end ofthe scale.

Overall, beagles appear to rate low tovariable not a very promising outcome.Yet beagles are the breed of choice for a

critical detection role that potentiallyprotects entire agricultural economiesworth millions. Why? Because on just onetrainability measure, they rate high, andthat is the critical measure for a foodproducts detection dog.

The point of this exercise is not to mountan attack on beagles, but to dissect thenotion of trainability into its componentparts and to explore how those parts might

be appl ied when consider ing breed-related trainability issues. Trainers of



29/66

24 Designer Dogs

MDDs appear to be in fairly generalagreement that the original training of anew MDD is difficult, and takesconsiderable time (routinely a year, andsometimes much longer). Very littleinformation is available on therelationship between notions oftrainability and the maintenance trainingrequired once an MDD is operational.However, with respect to improving theoverall efficiency of training for an MDDthroughout its working life, the questionof maintenance training may be even moreimportant than the question of howdifficult it is to train the dog in the first

place.

b. Perspectives on trainability

The above discussion is directed towardstwo general perspectives on the notion oftrainability.

Perspecti ve 1:

The designer MDD will need high generaltrainability because it must undertake avariety of unnatural tasks in a highly

constrained way. Training of such a dogwill probably be assisted if the dognaturally develops a strong rapport withthe trainer. Once trained, such a dog canprobably be relied on to workindependently, as long as motivation can

be maintained, and the dog can potentiallywork to almost any SOP. Put another way,the dog is trained to fit a desired SOP. If itis to be passed from one trainer/handlerto another, the issue of its tendency to relyon one person may require some

management. Maintenance training maybe quit e demanding beca us e of thecomplexity of the required tasks.

Perspecti ve 2:

The main training requirement will be todirect and control behaviours that the dogexhibits naturally. Here, trainability may

be rated low in most senses in which theword is used, and it may be necessary tofit the SOP to the dog. There is lesslikelihood of trainer/handler issuesneeding managing. Maintenance training

should be less demanding than underperspective 1.

Clearly, perspective 1 describes theshepherds that are normally used asMDDs. Breeds that might be suitableunder perspective 2 are considered

below.

2. Presentation

Presentation means physicalcharacteristics such as overall body sizeand shape, visual presentation (e.g. wolf-

like, puppy-like), design of structuralcomponents (e.g. ears, nose, feet), coatcolour and style, behaviouralcharacteristics, and so on.

Much of the amazing diversity of dogstoday can be explained by a combinationof particular functional requirements, theidiosyncracies of fashion preference indifferent times and cultures, and theserendipitous exploration of curiositiesthat appeared in certain genetic lineages.

For example, the unattractive undershotjaw and compressed muzzle of bulldogswas a functional requirement of bull-

baiting, the pulsating hair of Afghanhounds was a popular fashion accessoryin the 1970s, and the compressed face of apug was designed to create an animal inthe image of a human.

The curiosities and special-purposeexamples aside, certain features of dogsappear repeatedly across many breeds,

and the reasons for selection of thosefeatures are worth exploring.

Dogs were originally derived from wolves(Vila et al., 1997), probably with frequent

back-crossing to wild wolf populations.The possibility of occasional introductionof genes from other species such as jackalsis still controversial. In the evolutionaryprogression of dogs since theirdomestication, a process called neotenycan be found in both behaviour andphysical presentation across breeds


30/66

25

(Goodwin et al., 1997). Neoteny isdefined as the expression of juvenilecharacteristics in the adult form (Grandinand Deesing, 1998). It is not a rare oraberrant process and can be found inmany wild species and mostdomesticated species. For exampleaxolotls are the juvenile form ofsalamanders, yet axolotls are capable of

breeding and may never metamorphoseinto fully adult salamanders this isequivalent to tadpoles breeding without

becoming frogs or caterpillars breedingwithout becoming butterflies. The flatface (strictly a compressed primate

Cairns et al. (1996; see especially thechapter by Gottlieb). Put simply, effectssuch as lengthened or shortened bonesare achieved by adjustments in the ratesof bone development, and the achievableoptions are restricted by biology. Themorphing software mimics this processremarkably closely (Figure 4). Theexample began with a line drawing of aGerman short-haired pointer originallyproduced by Konrad Lorenz in about 1935(Lorenz, 1952). If the bones of the face areextended, a sight hound such as a borzoior greyhound appears (i.e. a breed inwhich there has been selection for long

muzzle) of humans is a neotenous

feature, and humans are possibly themost domesticated of any species.

The evolutionary process leading tochanges in the bone structure of a dog can

be mimicked using morphing software.The process involves the application ofprinciples of artificial selection to realdogs, but the derived possibilities areconstrained by the mathematics and

biochemical processes of development.The behavioural perspective ondevelopmental issues can be found in

bones and a lithe body). If the face is

shortened (mimicking the process ofneoteny), a series of recognisable breedsappears ranging down to the malformedpug.

Why would dogs be neotenous? Whenhumans chose to domesticate the originalwolf, they were confronted with theproblem of living alongside an animal thatwas a predator, a competitor for food, anda threat to children if not to adults. Eventoday, a wolf howl heard in the wildernessthrills and excites the senses by causing a

Figure 4. Developmental change and neoteny in dogs. A line drawing of a Germanshort-haired pointer was manipulated using a morphing programme to mimic

the process of artificial selection. As the shape of the head is adjusted,well-known breeds appear.



31/66

26 Designer Dogs

rush of adrenalin an endocrineresponse linked to fear. It is thereforelikely that the original domesticatedwolves were quickly modified to makethem less threatening, effectively makingthem less wolf-like. Two obvious changeswere decreased size and decreasedaggression (or increased subordinate

behaviour). Wolves signalling subordinatestatus make themselves appear less wolf-like by decreasing their size (lowering the

body, flattening the fur, turning back theears) and hiding their weapons (turningthe head, lowering the lips) (Coren, 2000).The artificially selected changes to wolves

to turn them into something lessthreatening for humans were morepermanent representations of the sameeffects, and can be seen today in the villagedogs that abound throughout Africa andAsia (Figure 5). These dogs are small,extremely subordinate to humans evenwhen seriously provoked, and tend tohave floppy ears and multi-coloured orred fur. Visually, they give the appearanceof a submissive dog even when in acompletely relaxed state, as this one is.

baby features). Baby-like featuresappeal to the parental and protectiveinstincts of humans for many reasons;most relevant here is that a cute animal isunlikely to be perceived as threatening.

What ar e t he impli cati ons for t he design

of an MDD?

Many of the human cultures in whichMDDs are required to operate are not dog-friendly. The village dogs of Africa andAsia are not integrated as pets into thehuman families with which they associate,and are frequently the subjects of abuse

and disgust. The dogs may not be feared,but nor are they valued or respected. AnMDD that must be integrated into such acommunity will face a doubledisadvantage: (i) the local population mayreact negatively to it, and (ii) any personchosen as a handler will come from acultural background in which dogs areignored or hated. Experiences inAfghanistan and Africa indicate that it ispossible to overcome the disgust andloathing that such people feel initially

Figure 6. A King Charles spaniel,a breed commonly referred to as cute.

Figure 5. A typical village dog,photographed in Mozambique,

Africa.

A second pattern of neotenous change canbe detected in many breeds that weredeveloped more as fashion accessoriesthan for roles in hunting, fighting orguarding. These dogs, such as the KingCharles spaniel, are often characterised ascute (Figure 6). They tend to be small instature, with high foreheads, big eyes, ashort muzzle and juvenile behaviour (i.e.

towards dogs, and that MDDs and theprofession of handling them can beaccepted by the community. But peoplechosen from such communities fortraining as handlers will not bring anunderstanding of dog behaviour to the jobas a background skill. Nor are they likelyto comprehend the notion of the dog-human bond, which is an important issue


32/66

27

with many breeds. With experience andtraining, they will develop betterunderstanding, especially if they becomeinterested in and enthusiastic about dogsin general terms. But the latter outcomecannot be guaranteed, and realistically, isunlikely.

a. Issues for handlers andcommunities

Four points are worth considering here inrelation to the designer MDD.

If potential handlers are initiallyafraid of the dog with which they must

work, then their ability to develop aneffective working relationship with thatdog may be compromised. Overcomingfear of dogs is possible, but much more isneeded. The handler may be required todevelop a close partnership with that dog,including living with it up to 24 hours aday. Dominance can be a significant issuewith male shepherds in particular, andany fear exhibited by the human may beexploited by the dog to manipulate therelationship.

If there is a requirement for the dogto be integrated into local communities,then that dog should be as non-

threatening as possible. If the dog appearsnon-threatening, then it is more likely tobe accepted, or at least to be tolerated, bythe community. The designer MDD could

be made non-threatening by exploitingknown neotenous design trends alreadyavailable among dog breeds. But it mightequally be more acceptable if it simplylooks nothing like the village dogs. Forexample, German shepherds areobviously dog-like, but they look nothinglike a typical African/Asiatic village dog.

Their distinctiveness might be enough toconvince the community that Germanshepherds are something other than thedogs they know, and therefore to betreated as something other than a dog (andmore acceptable).

The third point makes the aboveargument more difficult to apply.Unfortunately, German shepherds exhibita feature that could make them lessacceptable to humans in general terms.

Figure 7. Similarities in appearance between German shepherd (lower right) and thenorthern European wolf (upper right), and Malinois (low left)

and the Asiatic wolf (upper left).



33/66

28 Designer Dogs

German shepherds are a rarephenomenon a breed that wasspecifically selected to look more like thelarge northern wolves that inspire a thrillof fear when heard or seen (Pullainen,1967; Goodwin et al., 1997). It is morecoincidental, but equally unfortunate, thatthe other commonly used MDD (theversion of the Belgian shepherd called theMalinois) looks very like the smaller andleaner Asiatic wolf. If the (unnatural?) fearof wolves felt by many humans is arelevant issue in the overall operationalrequirements for an MDD, then it isdifficult to imagine a worse choice of

breeds than these two (Figure 7).

Fourth, if the MDD can operateeffectively with little requirement for thehandler to have a sensitive understandingof dog behaviour, then a significant andpotentially insurmountable trainingproblem will be avoided.

b. The physical appearanceof the scent hounds

A good nose is an essential requirement

for an MDD, and the scent hounds havethe best noses. Any review of theappearance of scent hounds leads quicklyto the conclusion that they were not bredfor looks. They tend to have largependulous ears, a thick muzzle, and wetfloppy lips (Figure 8), and they present aface that may have character, but isunlikely to be thought of as attractive(except, of course, by enthusiasts). Allthree of these features assist in the primaryobjective of the scent hound to detectan elusive odour. The flopping ears andlips create air movement that helps todraw odours into the nose. The thickmuzzle is a consequence of massiveinternal surface area of the nasal lining.The drooling mouth indicates that thenasal lining is being kept moist to optimiseodour capture. If these features improvethe ability of scent hounds to detect scents,then they are likely to be desirable featuresof an MDD.

Figure 8. The face of a scent hound (abassett hound), indicating the broadnose, floppy lips and pendulous ears.

Most scent hounds are short-haired,although one group (the griffons) isusually wire-haired. Short hair offers theconvenience of easy maintenance, andtolerance of hot climates both usefulfeatures in an MDD. Wire hair offers

waterproofing for wet climates. Thesefeatures were selected in the hounds inorder to minimise maintenance needs forhandlers.

The colour of scent hounds is variable,with many being multi-coloured. Ingeneral, there is no obvious feature ofcolour that might be optimised in anMDD. Such dogs usually work close to ahandler, so there is no need for contrastcolouration to facilitate keeping them in

view. Black dogs might be slightly moresusceptible to heat than dogs of othercolours, but this is likely to be a marginaldifference. The possible aim of developinga dog that looks unlike the village dog willnot be helped by the available colourvariation in the scent hounds, becausevillage dogs are also quite variable incolour and patterning.

Scent hounds offer considerablevariability in length of leg, speed ofmovement, and willingness to work


34/66

29

slowly and carefully. However, dogswith unnatural shapes (the basset anddachshund are extreme examples) tendto exhibit a variety of health problems,particularly of the back and hips, thatmake them unlikely candidates as MDDs.Overall, any attempt to design a slow-moving dog would be best achieved usinga dog of normal proportions in whichoverall size and/or gait was adjusted.

3. Size and shape

On minimum size, the requirement for a

good nose means that scale factors arelikely to work against the designer MDDbeing very small. Small noses are simplyunlikely to be as effective as large noses

because of the relationship between odoursensitivity and surface area of the nasallining noted above as a general feature ofthe scent hounds.

However, other economies of scale workto the advantage of smaller dogs. Smallerdogs require smaller kennels and less

food, and in a big operation the savingscould be substantial. Because of their

lower body mass overall, smaller dogsare less likely to suffer from the back and

joint problems that most very large dogsexperience, potentially reducingveterinary and downtime costs. Smallerdogs are also less likely than larger dogsto be perceived as threatening by thecommunity or by potential handlersunfamiliar with dogs.

In contrast to some commonly-heldbeliefs, small dogs do not necessarilyrequire less exercise or a smaller area inwhich to play than larger dogs, and theycan be just as aggressive as larger dogs.

Some small dogs, especially some terriers,are powerhouses of energy with an ego tomatch (Figure 9). Most of the very smalldogs were bred either as toy dogs (a widevariety of breeds) or as hunting (vermin)dogs (the terriers, chihuahua, dachshund).Although many of these dogs are verygood-natured and bond well withhumans, it is rare to find one that handlersrefer to as trainable. Thus, even if asmall dog with a good nose could befound, other factors are likely to work

against its suitability as an MDD.

Figure 9. A typical terrier in action!


35/66

30 Designer Dogs

Should the designer MDD be large?Hakimi (1999) indicated that large sizewas preferred, apparently because heequated size with stamina. In reality,stamina is more related to breeds than toindividual size of dog very large dogscan have impressive power, but theytend to tire quickly relative to workingdogs in the middle size range. Any dogwill have greater stamina if its fitnessand agility skills are maintained at a highlevel, but the associated maintenancecosts (mostly handler time) will be higherfor a large German shepherd than fora working breed such as a kelpie,

because of the higher natural fitness of thekelpie.

Apart from scale factors in nose design,the only benefit of large size is likely to bethe ease with which a handler can keeptrack of the dog. Large dogs are easier tosee in tall vegetation.

Two important disadvantages are: (i) thatmany large, heavier-bodied dogs suffer tosome extent from back and hip problems,

and (ii) the largest breeds tend to live fora relatively short time. Clearly, breeds ofvery large size should be avoided.

In reality, with health factors in mind itseems sensible to design an MDD inapproximately similar dimensions to wilddogs. Wild animals represent the bestpossible test of a design, and most wilddogs range in size from foxes (about 10kg, ignoring a few tiny species) to northernwolves (up to 40 kg, but more usually

around 25-30 kg). Thus, the designer MDDshould be in the 10-30 kg range, with apossible bias towards the smaller end ofthe range in order to facilitate economiesof scale in large operations.

With respect to shape and height, a tall,slim-bodied dog might be acceptable forhealth reasons, but such dogs tend to befast moving not a desirablecharacteristic of an MDD. A dog withvery short legs might move slowly, butis likely to suffer from back and joint

problems. Overall, it seems most sensibleto retain the shape as well as the size ofwild dogs.

4. The cognition question intelligence

A simplistic but useful view of how ananimal interacts with the world is toconsider two perspectives: (i) the worldpresents problems that need solutions (thecognitive framework), and (ii) the worldprovides rewards linked to actions (theoperant framework). A straightforwardreview of these concepts can be found inOFarrell (1992). The first option makessense from a human perspective problem-solving has even been cited as acharacteristic that distinguishes humansfrom other animal species. Certainly, thequestion of whether animals undertakeproblem-solving is controversial (Nicol,1995), although the problem-solvingskills of various dogs have been compared(Scott and Fuller, 1965), as have wolves

and malamute dogs (Frank and Gialdini,1985); wolves were better thanmalamutes). The problem with thisempirical approach has been one ofdesigning experimental protocols thatdemonstrate problem-solving by animalswhile rejecting the two alternative (andapparently simpler) explanations that: (i)the animals are either responding in astimulus-response way to cues in theexperimental environment (i.e. theoperant framework), or (ii) it is species

differences that explain the results and thecomparison is meaningless. Thesetheoretical problems are of no relevancehere, but they provide the background foralternative perspectives on the best designfor an MDD.

When dealing with the world cognitively,an animal receives information, andassesses the meaning and quality of thatinformation against a background ofother similar experiences. It then uses the

combination of current information and


36/66

31

background experience to decide on abehavioural outcome, i.e. to make adecision about what to do in the currentsituation.

When dealing with the world operantly, ananimal responds to a current stimulus ineither a negative or a positive way,depending on relationships between thestimulus and associated rewards. Noinformation processing is involved.However, the animal can learn thatdifferent stimuli are linked (ABC),and can tune its response to stimuli thatoccur reliably in its environment. To give

a relevant example, a trained MDD haslearned the sequence: detect odoursitdownreceive ball. No understandingof the relationship between odour and

ball is either implied or required, butbecause the ball represents a positivereward for the dog, it will express thestimulus-response sequence of detectodour sit down consistently andreliably.

Dyck (1999) referred to MDDs as requiring

courage. The word was undefined in hispresentation, but the implied meaningappeared to be a mix of robustness,determination and exploration acourageous dog had good stamina andmotivation, and was interested in itsenvironment.

More usually, the notion of courage islinked to bravery and risk-taking. Theimplication here is that the risk wasunderstood and the danger was accepted

because of the need to undertake someimportant task. Thus, rescue personnel arefrequently called on to be courageous,

because they must enter a danger zone.Their courage is tempered by theirtraining, which gives them a highly-refined ability to assess risk and judgemargins of safety, and their SOPs, whichhave been designed with safetyconsiderations in mind.

Could an MDD be courageous in thissecond sense? In effect, the implication is

that the dog understands the dangersrepresented by a minefield, and acceptsthe associated risks, as human mineclearance personnel do. Presumably, itwould also act to minimise risk bydeveloping appropriate SOPs.

This scenario is not as outrageous as itmight initially seem. The Border collie isan extremely intelligent dog, and mostpeople that work sheep with Bordercollies will argue that the dogunderstands the problem. Once the dogunderstands where the handler wants thesheep to go, it will assess the behaviour of

the flock, predict the behaviour of strays,make decisions about when to run andwhen to stop, judge when to move to theleft or to the right, and generally get the

job done with a minimum of commands.The words in italics here are the languageof the cognitive framework. At least withrespect to handling sheep, a good Bordercollie is potentially operating at a cognitivelevel similar to that of the handler. Watchsheep being worked, and it is not hard toconclude that disagreements between the

handler and dog are often because the dogis right!

The intelligence of Border collies has beenexploited in many other service contexts e.g. as search and rescue dogs and asassistance dogs. There have even been afew attempts to train them as MDDs (oneis currently operational in Cambodia,Figure 10). They have quite a good nose,a well-developed wish to please thehandler, and they love to work. Yet they

appear only rarely in any of these serviceroles, perhaps because too much down-time is involved, and Border collies needto work all the time.

a. The need for intelligence?

It is extremely unlikely that a dog alreadyexists which could be trained cognitivelyas an MDD, similarly to how a Bordercollie learns to work with sheep (Bordercollies and/or standard poodles are themost likely candidates here because of



37/66

32 Designer Dogs

their intellectual skills). Of course, anydog able to understand the problem islikely to immediately refuse the work,for the same reasons that many humans

would consider mine clearance to be anundesirable profession. Strictly, thatproblem would be overcome by breeding the bred intellectual MDD would bemotivated to search for mines in the sameway that Border collies are motivated toexhibit rounding up behaviour. For Bordercollies, understanding the problemreally means that they have the intellectualability to understand and predict the

behavio ur of sheep (s ee a similarargument made about the intellectual

skills of sheep in Kendrick, 1997). It doesnot mean that they understand why thefarmer is moving the sheep from one fieldto another. In the same sense, theintellectual MDD does not need tounderstand the dangers represented bymines. Rather, it needs to understand thatmines occur unpredictably, have acharacteristic odour, and should be treatedwith caution when found. In other words,it will have the ability to predict the

behaviour of mines, if such a concept

makes any sense.

Although unlikely, the possibility ofundertaking cognitive training with anMDD is raised here for two reasons. First,in the interests of completeness the

concept may have some application in thefuture even if it is unrealistic now. Second,

because it focuses attention on the valueor importance of intellectual skills in thedesigner MDD.

The usual training approach with MDDsdoes not utilise the cognitive or problem-solving perspective described above. Noattempt is made to train MDDs tounderstand the problem. Rather,searching for that elusive odour is

developed as a game. The game has somevery specific rules involving how andwhere to walk, required intensity ofconcentration, and appropriate behaviourwhen the odour is found. From the dogspoint of view, the aim of the game is tomanipulate the handler to provide adesired object, such as a toy with whichthe dog is obsessed as a result of previoustraining. If food rewards are used, then theaim is to manipulate the handler intoproviding a highly desirable treat. Here,the approach is strictly operant. The

Figure 10. A Border collie at work as an MDD in Cambodia.


38/66

33

cognitive potential of the dog could evenwork against the training objective if thedog is clever enough to understand thatthe handler can be tricked into providingthe reward.

Intelligence is not a requirement for ananimal functioning in an operantenvironment. Demonstrations ofconditioned responses can be found forvirtually all animals (down to unicellularorganisms) and even some cellularpreparations (Turkann, 1989), suggestingthat stimulus-response learning is afundamental feature of nervous systems.

However, dogs are intelligent creaturesrelative to most animals, and intelligencemay make a difference to the quality orsubtlety of learning that can be achievedin a stimulus-response environment.

Both Dyck (1999) and Hakimi (1999)identified intelligence as an essentialfeature of an MDD, probably becauseintelligence was perceived as being linkedto trainability. However, it seems likelythat these authors were not talking about

dogs in general when they noted therequirement for intelligence in an MDD.They were talking about the breedsthey knew best as MDDs Germanshepherds and Malinois (henceforthcalled the shepherds). In this context, theyare probably right. Intelligence is animportant requirement if individuals ofthese breeds are to be effective as MDDs.Why? Because the shepherds are derivedfrom general purpose working dog lines.They were not bred for the purpose of

mine detection (or, more generally, asdetection dogs). Their training as MDDswill therefore require learning of acomplex and unnatural array of skills withvery high reliability. That trainingobjective is more likely to be achievedwith an intelligent dog, because, onaverage an intelligent dog is likely tohave higher trainability for unnaturaltasks. The shepherds have an additionalcharacteristic that lends itself to thetraining problem: they bond strongly withhumans and have a strong desire to please.

That bond makes them more willing toundertake whatever task a handler asksof them, and with very high reliabilityonce they understand and have practicedthe task. With these breeds, the operantcontext of learning (a game) is just one ofseveral incentives that generatesmotivation and reliability.

It does not automatically follow that therequirement for intelligence in theshepherds as MDDs predicts intelligenceas a requirement in any breed used as anMDD. If intelligence were so important asa general feature, then surely there would

be more Border coll ies and standardpoodles used as MDDs. Lockwood (1999)and Almey (2001) provided oneexplanation for their absence: a biastowards shepherds in the training

background and experience of the peoplethat train MDDs. Lockwood was carefulto emphasise that the comment was notmeant critically it makes sense fortrainers to work with the breeds theyknow, and there is no doubt that theshepherds make effective MDDs. A

second possibility is that Border colliesand poodles have been tried, and foundto be unsuitable for other reasons, such ashigh activity levels or an inadequate nose.There is almost no evidence for this secondsuggestion.

Here is an alternative perspective onintelligence in the designer MDD. Themore motivated a dog is to perform a task,the less need there is to train that dog toundertake the task. Many dogs currently

kept as pets were actually bred for specificpurposes, and frequently express bred

behaviours in inappropriate contextsbecause the ap propri ate context isunavailable to them. A pointer that pointsleaves in the garden, or a Border collie thatrounds up the children, is amusing. But afighting dog that bites is dangerous andunacceptable. Blocking the expression of

bred behaviours can be very difficult, anddogs are destroyed every year simply

because they express bred behaviours ininappropriate contexts. The relevant point



39/66

34 Designer Dogs

here is that such behaviours representvery high motivation for a task. If a dogcan be bred to exhibit most of the

behavioural requirements needed in anMDD with determination and reliability,then there will be little need to traineither motivation or a complex array ofunnatural tasks, both of which requireextensive training in the shepherds. Theonly training problem with the bredMDD will be to direct naturally expressed

behaviours in to a part icular format ,presumably using operant techniques.There would be little or no need for sucha dog to be intelligent in the sense that

shepherds need to be intelligent.

Few if any breeds have been createdspecifically for dangerous service roles,

despite the mythology surrounding theSt. Bernard. However, some breeds doundertake potentially dangerous servicework with dedication and enthusiasm.The Newfoundland has the (unfortunate!)tendency to pull anybody from the waterwhether or not they need to be rescued.Stock guarding breeds, such as themaremma, Briard and Balkan shepherds,will fight to the death to protect theircharges. Such dogs are not usually notedfor either their trainability or intellectualskills it is their breeding that dedicatesthem to the role. If such breedingoutcomes can be achieved in those

contexts, then why not for minedetection? Perhaps a dog already existswhich exhibits most of the bredcharacteristics needed in an MDD.


40/66

35

The discussion so far has focused ongeneral features of an MDD. Suchfeatures form the background to the

specific problem one of creating a dogthat expresses a carefully constructed

beha vioura l reperto ir e in a re liab lemanner.

1. Types of MDDoperational today

Currently at work today, there are fourbehavioural types of MDD:

Unleashed(also called free-running)dogs search for mines without beingunder the direct physical control of ahandler;

Long- leashdogs (Figure 11) are

linked to the handler by a line, butmay work at distances up to 10metres away so are operating semi-independently;

Short-leashdogs (Figure 12) workalongside a handler while controlledon a short line (a typical leash);

REST(Remote Explosives ScentTracing) or MEDDS (MechemExplosive and Drug DetectionSystem) or EVD (Explosive VapourDetection) dogs (Figure 13) may be

leashed or unleashed, but they have

a very specific task: to check a seriesof samples for the presence of odourfrom mines.

Although each of these dogs must learn adiffe

Mine Detect

Documents

Transcript of Mine Detect