Malcolm B. Willis BSc (Dunelm, 1956) PhD (Edin. 1960)
Hon Assoc RCVS (1996)
At the present time it is fashionable in some quarters to preach doom and gloom about pedigree dog breeds. Most of this is on the basis of inbreeding and much of it is circulating round the internet like some disease epidemic, or maybe a virus. Even editorials in the canine press have started asking questions. In some countries in Europe, notably The Netherlands and Germany, politicians are seeking to introduce rules about the breeding of dogs. I have little faith in politicians when they concentrate on politics and no faith whatsoever in such individuals passing judgement on biology, genetics and the dog on which they probably know nothing and are being advised by those with an axe to grind.
Although some types of dog (e.g. sight hounds) have existed in something like their present form for a possible 6,000 years, modern examples of Salukis, Afghan hounds or Greyhounds cannot trace their ancestry back to these early days. In the interim much intermingling will have occurred. Most modern breeds can trace their ancestry in pedigree terms back for some 100 to 125 years, some even less. The (English) Kennel Club was formed in 1873 and it is doubtful if reliable pedigrees exist before this date. On the basis that a generation in the dog is about 4.5 years we are talking of pedigrees that go back for some 22 to 28 generations.
All dogs descended from the wolf (probably various kinds) and began domestication some 10-12,000 years ago. It is popular to look at wolf behaviour and to conclude that dog behaviour has changed. Of course it has. Some behaviours seen in the wolf would have no purpose in a modern dog : burying food, for example, or vomiting up food for cubs. Some behaviours are not feasible because the conformation of the dog has been changed and this prevents some activities. Rightly or wrongly, we have tended to select dogs to retain their juvenile state and this has meant drop ears ( not as expressive as erect ones) while hairy faces and docked tails inhibit a dog's ability to convey messages to other dogs. That modern canine behaviour has changed from that of the wolf is not in itself a problem in the majority of cases. It could be in dogs that go feral but in most modern societies we are seeking to avoid that. The border collie herding sheep is following the hunting habit of the wolf but the exercise does not end up in a kill because man has brought about behavioural modification through selection. In contrast the fighting nature of some Pit Bull terriers is not the least bit wolf-like and represents the corrupting influence of man upon this tragic breed,
What one must be careful of is setting up the wolf as some kind of prototype to which dogs have to measure up and this is particularly true if we consider inbreeding. To discuss this we have to examine inbreeding.
The definition of Inbreeding
Inbreeding is popularly thought to be the mating of relatives. That is too general a definition. If we look at the pedigree of any individual of whatever species we find that ancestors double in each generation. We have two parents, four grandparents, 8 great grandparents etc. In generation 20 we actually have 1,048,576 acestors and twice that in the 21st generation. It must be obvious to anyone that, whatever species you examine, the time will come when there are more ancestors in a pedigree generation than there were individuals alive at that time. This can be readily demonstrated in any dog pedigree. It might be possible, in a numerically large breed to find nothing common within five generations but within ten there will be ancestors that appear several times and the further back one goes the more obvious this is. Take any Boxer pedigree and way back you will find Lustig v Dom. Take a BMD pedigree back for about 11 generations ( to the late 40's period) and you will find the Newfoundland dog Pluto v Erlengut. This will apply regardless of country of origin of the dog being examined. Though a few isolated lines might exist that are different they would be few and far between.
One cannot define inbreeding as simply mating relatives. The true definition is the mating of individuals more closely related than the average of the population from which they come. This means that a true definition of inbreeding could vary from breed to breed and from location (country) to location. However absolute values can be derived and assessed on that count.
Inbreeding is measured using Wright's Coefficient of Inbreeding which was first put forward in the 1920s. It can be expressed as a percentage e.g. 12.5% or as a proportion 0.125. It measures the increased homozygosity likely to occur in an individual. If you mate a Boxer to a Boxer you get Boxers which is no surprise because, of course, over the years many genes have become fixed in the breed. This means that all individuals carry the same combination of these genes.
For example, all BMD carry the non-merle version of the merle gene and are thus all mm at that locus. Similarly they are all a t a t and the greater majority are all BB though a very small number must be Bb and a very tiny fraction are bb. The fact that they are homozygous at these loci has no bearing whatsoever on the well-being of the dog and that will apply to a great many loci. Homozygosity is not a disease. Note also that although some inbreeding has been undertaken in the BMD the incidence of bb animals ( a colour problem not a disease) has not increased and most breeders have never seen one.
The consequences of Inbreeding
Whether we inbreed or not it would make no difference to genes that are fixed in all members of the breed but in non-fixed genes it would lead to an increase in homozygosity and a decrease in heterozygosity. If we have a gene that we will call N with alternatives N and n then we have three possibilities NN, Nn and nn. Inbreeding will move us towards the NN and nn versions at the expense of Nn. Most deliterious traits tend to be recessive. This is because if an undesirable feature is dominant a dog which has it shows it and thus is usually selected against, unless it is very late onset. Thus most abnormalities and defects tend to be recessive ( the nn equivalent) while normal animals are NN or Nn (called carriers). If inbreeding takes us towards NN and nn then the first thing we will see is an increase in deliterious defects. But many rare genes will be lost on inbreeding, it depends on the population. Moreover once identified the nn animals and some of the Nn ones can be discarded from a breeding programme. Inbreeding does not operate in isolation but coupled with selection and this is certainly true of the dog in respect of inherited defects..
Inbreeding BMD, for example, would increase the risk of hypomyelinogenesis (trembler) but only very slightly because it is already rare. In contrast it might not increase cancer risks (though they are not necessarily simple traits) because some 40% of the breed already die of cancer of one kind or another. However if your line does not carry a specific defect inbreeding will not create it. For example, our own line began with a dog of exemplary character who was the son of a "trembler carrier". This meant that our dog had a 50:50 chance that he carried the trembler allele and a 50:50 chance that he did not. Close breeding (> 25% inbreeding) within the line has shown that he did not carry the trembler allele and thus it is absent from our line. Do not forget that breeders, even if they inbreed, also follow selection so that we are not just talking about inbreeding but inbreeding with selection which is a different ball game.
What we do know is that inbreeding brings about something called inbreeding depression. This depends on a formula which is:
This will look complex so let us examine this. The term F relates to the inbreeding coefficient. If the animal were a brother sister mating the F value would be 0.25 (25%) and if a half brother half/sister it would be 0.125. So 2F relates to twice the inbreeding coefficient in the population. The terms p and q relate to the frequencies in the population of the alternative genes. If, for example, half the genes in the breed were N and the other half were n then the p frequency (N) would be 0.5 and the q frequency (n) would also be 0.5. If only ten percent of the genes were n then p would equal 0.9 and q would equal 0.1. (A summation sign is needed because it must be done over all loci but is not given because my computer does not have it).
The term d relates to the degree of dominance. This is measured from the midpoint between the parents. Suppose wither height were a single gene trait and we mated a 66 cm animal to a 60 cm (having corrected for sex) then the mid point is 66+60/2 = 63. If this offspring from this mating averaged 63 cm there would be no dominance and d would be equal to 0.
If the offspring averaged 64 cm then this is 1cm above the midpoint and thus d = 1.
Clearly wither height is much more complex and the formula has the symbol to sum over all loci. However the term dpq is dependant upon there being some degree of dominance. If d = 0 then the term dpq is also equal to nothing and there would be no inbreeding depression. Similarly the pq part is higher when one is at intermediate rather than extreme values.
The belief that inbreeding always causes problems is too sweeping. If there is no dominance there is no depression for that trait. The critics of inbreeding rarely, if ever, tell you this.
Without going into too much mathematics it can be shown that in relatively highly heritable traits d tends to be small while in low heritability traits it can be large. Inbreeding depression as it is called is likely to be most obvious in what are called fitness traits ( fertility traits) and least obvious in traits like some aspects of conformation which tend to be quite heritable, often in excess of 0.40 (40%).
Inbreeding will not have much effect upon high heritability traits but could have in respect of low heritability traits. A breeder would rightly argue that maintaining fertility was desirable since without it the breeder cannot function. However what about prolificacy (litter size)? A pig breeder wants as large a litter as he can rear and thus inbreeding, if it reduced litter size, would be very damaging to a pig breeder. That is not true of a dog breeder.
A normal GSD litter is about 8 whereas a Pomeranian litter is closer to 2.
If one wants to pick a dog and a bitch to father the next generation then in the GSD case one is retaining 2 out of 8 or the best 25% whereas in the Pom one has to keep the whole litter and thus has no selection unless one produces a second litter. However a Pom breeder would need to have four litters to have the same selection rate as a GSD breeder and that will take longer in time to actually reach, increasing the generation interval. As a GSD breeder one wants a dog that is of sound character, is typical of the breed, free of major inherited problems and (if one is an exhibitor) that will win in the ring or win in the working side of the breed or both. If it is a companion animal then the winning aspect is irrelevant but the other features will apply. If one inbreeds and inbreeding had no effect upon high heritability traits but reduced litter size from 8 to 6 the litter would still be worth doing because a litter with one outstanding member would still be worth producing. A pig breeder wants uniformity and large numbers, a dog breeder ( and also a racehorse breeder) wants an exceptional animal and large numbers of mediocrity born is of some economic but little selectional merit. If we could each produce a litter of 6 with one world-champion and 5 average dogs that would be of more use to us and the breed than a litter of 8 average animals. I know few breeders who would be put off that idea.
A breed is a collection of animals that are related, stem from a specific origin and have various features in common. In the development of breeds there was a distinct tendency to inbreed so as to try to fix various attributes, mainly of phenotypic importance. When Longhorn cattle were the norm the development of the Shorthorn was enhanced by making it not only shorter in horn length but of different colour and shape with different attributes. In the development of the Orlov Trotter in Russia in the late eighteenth century inbreeding was rife and Kelley (1946) gives a pedigree of Lubensoi 111 which has an inbreeding coefficient of over 64% inside 7 generations, higher than any dog pedigree I have seen outside of colonies bred for research and thus not "real dogs".
Breeds of dog have been selected in many instances for specific roles. These may vary from being a lap dog through to all manner of work and companion animal status. Without a herding dog (Border Collie, Kelpie etc) a sheep farmer would be highly handicaped in European/American situations. In USA the role of livestock protection dogs (Maremma, Anatolian etc) is such that without them sheep farming would be impossible in some locations because of predator attack. But the traits that are found in a herder are not those found in a livestock protection dog. One dog cannot thus replace the other and both types are needed. This must not be lost sight of. Dogs have a specific purpose and are needed for that purpose even if it is merely as a companion to an old-age pensioner living alone.
The GSD is the police dog par excellence. British police forces have tried Rottweilers, Dobermanns, Bouvier des Flandres and Mallinois among others but for general police work the GSD is the dog that still wins over the others, even though few police forces breed their own dogs and they are thus dependent upon breeders who did not set out to breed police dogs.
Breeds not only aid man in working situations but they also undertake such tasks as being the eyes for those who cannot see. Dogs like Labradors and Golden retrievers, developed originally as gundogs, now have little or no opportunity to do this sort of work but they have developed other attributes such as guide dogs for the blind. This shows a tremendous versatility and ability to adapt yet some advisors to governments seem to want to be rid of the purebred dog. If crossbreeds and mongrels were the ideal then why have they not been used? Twenty years ago I visited South Africa where the police were crossing Bloodhounds and Rottweilers/Dobermanns for working purposes. I went back to SA a few years ago and the crossbred animals were a tiny minority with GSD of excellent police ability dominating the force's dogs.
Crossbreeding has a useful purpose in Beef, Sheep and Pig production and no agricultural scientist would argue with the wisdom of such an action. Indeed in such species the concept of pure breeds is almost redundant. But world wide modern dairy cattle are almost all black-and-white descendents of Dutch Holstein-Friesians and breeds like Jerseys, Guernseys and Ayrshires are almost gone. When I was a university student in the early 1950s the Dairy Shorthorn was the breed of choice in Britain but it is now a rarity. This did not come about by pressure from scientist but the fact that for the circumstances of the day the Holstein-Friesian outclasses all others. Crossbreeding is not undertaken in most dairy herds because the Holstein is so far ahead of the other breeds that crossing it would be a retrograde step. However inbreeding is hardly a problem even though many small bottlenecks have occured over history.
Various organisations around the world are intent on retaining rare pure breeds of farm livestock species in the knowledge (hope?) that one day they or their attributes may be needed. Such bodies are aware of inbreeding risks but they are intent on retaining the purity of race. In contrast canine breeds in Holland are threatened with political interference under the guise of preserving them but by imposing so many restrictions that they are taking away the role of the breeder and bringing in draconian rules even in breeds that are numerically large. I do not have Dutch registrations figures but in the UK we register about 20,000 GSD annually and the average lifespan is 10 years. That means that we have about 200,000 GSD in Britain at any one time. That is not a population at risk. In contrast we register 800 BMD with a lifespan of 7 years. Our population at any one time is thus 5,600 which is hardly risk making. Breeds that can amass a mere 100 individuals may have problems but we should deal with these according to their needs and not apply the same rules to major breeds with numerical strength.
Few canine breed studies on inbreeding have been undertaken but McCarthy and Blennerhasset (1972) made a study of a small sample of racing greyhounds in Eire and found an inbreeding coefficient of less than 1% per generation or less than 0.22% per year. Since man is permitted to marry a first cousin, leading to an inbreeding in any progeny of 6.25% the racing greyhound is hardly at risk from inbreeding. Indeed studies on most numerically large breeds would show that inbreeding over five generations would generally be around the 3-4% mark. My own analysis (Willis, 1989) of 276 Boxer champions in Britain showed a mean value of 4.2% with over half the animals being <1%. Extending pedigrees would increase these values slightly but breeders have no means of doing this. In his study of Seeing Eye dogs in America Pfaffenberger (1963) showed that inbreeding to the dog Frank of Ledge Acres increased the chance of successful guide dogs. It is true that Rehfeld (1970) in a beagle colony showed an increase in noenatal death with increasing inbreeding but he was operating at levels up to 78% , far higher than those seen in pedigree dogs.
Ubbink et al (1992) showed a highish level of inbreeding (median 6.4% to 12.5%) in Bouvier des Flandres in Holland and levels were slightly higher in some dogs with specific defects but not for all defects and the population was not large.
Nobody would dispute the fact that high levels of inbreeding can be damaging but in some cases the effect of inbreeding is not apparent at less than 20% (brother/sister and parent/offspring is 25%). Having said this the Chillingham Wild White Cattle herd, some 50 miles from where I live, has survived in genetic isolation under a king bull system for around 300 years and is still going strong but is numerically small (about 40 cows). In other words the high inbreeding coefficient on paper is not reflected in the realities on the ground.
In White Park Cattle in 1991 the effective population size was 79 and Alderson( 1992) argued that more than 15% of the variability could be lost in the next 50 generations. This is over a period of at least 250 years during which time Dr Alderson and the rest of us will no longer be involved! I do not think we can seriously hope to plan a breeding programme for 250 years ahead, nor should we try, however hungry we are for power and influence.
The idea that increased inbreeding will increase the incidence of defects is unlikely (Alderson & Bodo, 1992) if the defect has a high frequency. For example the long coat (an aesthetic defect in GSD) has an incidence to the extent that about 10% of GSD are born long-coated which means that about half the normal coated dogs are carriers and half are free of the allele. Nothing much is done or needs to be done because some owners like long coats and they have minimal biological disadvantage. Alderson & Bodo(1992) argue that inbreeding has not detracted from conservation programmes even when some animals have reached 20% inbreeding.
Let me close this section by referring to the eminent geneticist, the late Prof J.L.Lush of Iowa State, a world authority in the field of genetics. In his classic book Animal Breeding Plans (1945) he states " It seems reasonably certain that more opportunities for breed progress are lost by not inbreeding when inbreeding would be advisable than are lost by too much inbreeding." Some of the advisors to the Dutch KC, by no means in the class of Lush, would do well to re-read this classic book, dated though it is.
The dog has about 30,000 genes and there are about 400 known defects reported, many of them simple autosomal recessives or sex linked recessives and a few are dominant while some are polygenic. Around two dozen have been located by DNA testing and such defects can thus be rapidly eliminated from a population if testing is applied. For example PRA has a low incidence in the Irish setter and could be eliminated completely if all breeding stock were DNA assessed and those carrying the PRA gene culled from the breeding pool. Sometimes, as in copper toxicosis in the Bedlington terrier the gene is so widespread that a simple culling operation would be too drastic and a careful use of carriers to Normals undertaken to preserve other attributes.
Where DNA testing applies then breeding restrictions need only apply to the removal from breeding of the Nn and nn animals (NN being homozygous normal). In the case of other defects we must assess the nature of the defect and the stage at onset.
Much as one does not want pituitary dwarfs in the GSD it is obvious by about 7 weeks of age and thus represents economic loss to the breeder not the buyer. Such dogs can be given away or culled and if the concept of culling offends then such easily offended people should not be in animal breeding. All species including man have defects and always will because of mutations and we have to learn to live with the problem. It is not the most important feature of dog breeding. Late onset defects like some types of PRA are undesirable and breeding programmes need to be adjusted to identify carriers if this is not feasible (as yet) by DNA testing.
Polygenic traits like behavioural flaws, hip dysplasia, epilepsy, elbow dysplasia etc are more complex and where grading/scoring schemes etc are available then ALL breeding stock in appropriate breeds must be graded/scored and the breed clubs at general meeting must decide what is an acceptable level of grade/score and then the appropriate kennel club should impose regulations. Scientific advice should be taken from those scientists with experience and their ideas used in drawing up rules but this does not require breeding restrictions or limits on sires which achieve the criteria.
In a field like hip dysplasia a level of acceptability should be drawn up for appropriate breeds (not the same for each breed) and only dogs/bitches used that attain this level. Then progeny test data should be collected and if a sire (or dam) is producing poorly they should be removed from the breeding pool. No rules of this type yet apply in Britain but progeny averages are published and do lead to the reduced use of specific animals that had good results themselves but which did not live up to these as sires. Most breeders of experience have enough common sense to know that there is nothing to be gained breeding from a poor producer and they do not need some scientist telling them how to breed if the scientist(s) concerned have no experience of breeding dogs other than for vivisection purposes. The Zuchtwerts produced by the SV in Germany by Dr Reiner Beuing are more useful than blanket restrictions on use without such data being available.
Limiting sire usage
Although most publicity is given to abnormalities of a generally Mendelian nature the really important traits like behaviour, shape, construction etc are polygenic. Genetic theory states quite clearly that response to selection per year is given by the formula:
R = h 2 S
where R is the response per year, h 2 is the heritability of the trait, S is the selection differential and t is the generation interval.
As an example let us take a trait like wither height which in the GSD has a heritability of about 0.63 (63%) and let us assume that mean wither height( ignoring sex) is 64cm. Let us assume that the standard deviation of wither height is 2 cms. Let us further assume that generation interval ( the mean age of animals when their offspring are born) is 5 years. The selection differential is the superiority of parents relative to the mean of the population from which they come.
Let us assume that we are going to seek to make the breed larger in wither height. I am not saying that this is a logical objective but I am using it to show what restriction of progeny numbers can do. Let us assume that breeders want to use the best 10% of sires which, in this case would be the tallest animals. Then, using standard tables, these 10% will exceed the mean by 1.755 standard deviations which is (2 x 1.755) or 3.51 cms. In other words the best 10% would average 67.51 cms.
Progress through sires (we have ignored dams) will be on a per year basis:
R = 0.63 x 3.51 = 2.21 = 0.442cm
We have increased wither height from 64 to 64.442 in one year but since we have not selected dams this is reduced to 0.221 or 64.221.
Suppose we had exactly the same parameters but now we are restricted to using the best 50% of sires. These 50% will exceed the mean by 0.80 standard deviations which is (2 x 0.8) = 1.6 cm
Progress through sires per year is now:
R = 0.63 x 1.6 = 1 .008 = 0.2016 cm
In this case we have increased wither height to 64.202 or (taking account of no dam selection 64.101cm).
This may not seem a lot but it is indicative of what will happen if sires are restricted and thus more sires are used and it will happen on every trait you select. The whole basis of genetics is selecting the best and mating the best to the best. If we limit the best then we have to use second best and progress is reduced. In hip score in Britain over the past 15 years the Newfoundland has improved from around 35 total score to 22 total score by greater use of better sires. If the best are limited then hip progress will be less. This will appply to just about anything.
Some scientists, whose credentials in dog breeding may need to be proven, want to impose rules on breeders through kennel clubs without appreciating what breeders are trying to do. I accept that not all breeders are skilled but in that case let breed clubs and kennel clubs put their house in order, not by restricting use of the best, but by making sure that the best really are the best and that, within reason, steps are taken to ensure not only that they have desirable qualities of character and working ability (if important) but also other constructional virtues. Remember also that the dog breeder is interested in multi-trait selection not just one feature but a whole range of, mainly polygenic, traits. No dog will be perfect in everything and some allowances need to be made but they must be made sensibly and with logic by those who know.
The dog has survived as the most variable domestic animal for the best part of 10-12,000 years. It does not need messing about with. Because there are problems with numerically small species like Siberian tigers, gorillas or condors does not mean that the same things applies to the dog. If scientists want to deal with small population genetics then concentrate on these wildlife and leave the domestic dog alone. Dog breeders want advice not instructions.
If kennel clubs are concerned that American pit bulls are dangerous, as indeed some are, then limit their breeding and if need be get rid of the breed but do not let knee jerk reactions ruin pedigree dogs as quite clearly the impending rules will. They will also lead to more unregistered dogs and the decline in power of kennel clubs. Dog breeders do not need kennel clubs as much as kennel clubs need dog breeders and it is high time that some European kennel clubs started to rethink and educate their political masters accordingly. The last man who tried to impose "genetic theories" (in this case for man) was Hitler and we don't need reminding about the damage he caused.
Alderson, G.L.H. (1992) A system to maximise the maintenance of genetic variability in small populations. ( In genetic conservation of domestic livestock. Vol 2. Ed. Alderson & Bodo) CABI, Wallingford.
Alderson, G.L.H. & Bodo I. (1992) Review of Species and Breed studies. ( In Genetic conservation of domestic livestock.Vol 2. Ed. Alderson & Bodo) CABI, Wallingford.
Kelley, R.B. (1946) Principles and methods of animal breeding. Angus & Robertson, Sydney.
Lush, J.L. (1945) Animal breeding plans . Iowa State College Press (3rd ed).
McCarthy, J.C. & Blennerhasset, T. (1972) A preliminary estimate of the degree of inbreeding in Irish racing greyhounds. Dept Agriculture Journal 69 : 3-9.
Pfaffenberger, C.J. (1963) The new knowledge of dog behavior . Howell, New York.
Rehfeld, C.E. (1970). Definition of relationships in a closed beagle colony. Am.J.Vet.Res . 31 : 723-32.
Ubbink, G.J., Knol, B.W. & Bouw, J. (1992) The relationship between homozygosity and the occurrence of specific diseases in Bouvier Belge des Flandres dogs in the Netherlands. Vet. Quart . 14: 137-40.
Willis, M.B. (1989) Genetics of the Dog. Witherbys, London.
CV of the author
Qualifications are given in the title in detail to show that the author is an academic.
Dr Willis was born in 1935 in Yorkshire, England. He has a BSc in Agriculture from Durham University and a PhD from Edinburgh in Genetics and Animal Breeding where he studied under Prof. Alan Robertson FRS. From 1960-65 he was geneticist at the Milk Marketing Board of England and Wales. From 1965-72 he was Head of Animal Science at the Instituto de Ciencia Animal, University of Havana, Cuba and from 1972 to 1998 was Senior Lecturer in Animal Breeding & Genetics at the University of Newcastle-upon-Tyne, England. For two years he was also Postgraduate sub-dean of the Faculty of Agriculture & Biological Sciences. Now retired he still teaches one day a week.
Dr Willis bought his first GSD in 1953, first judged in 1959 and gave CCs from 1978. He has also breeds BMD with his wife Helen from 1985 and gives CCs in that breed. He has judged GSD in 9 different countries and BMD in 5 at championship level. He has visited several other countries to study BMD and GSD.
He is the author of, among other books, : Genetics of The Dog (1989), The GSD: a genetic history (1991) and The BMD today (1998). He has written chapters in numerous books on a variety of breeds and two books on cattle. His latest chapter (with Dr K.Houpt of Cornell) is on the Genetics of Behaviour in The Genetics of the Dog (eds Ruvinsky & Sampson, 2001). Dr Willis was awarded Honourary Associateship by the Royal College of Veterinary Surgeons in 1996, the highest honour the RCVS can give a non-veterinarian. He was also awarded the Gold medal of the GSD Breed Council of Australia in 1988 for services to the breed. He is Chairman of the GSD Breed Council in Britain ( a federation of over 40 clubs) and has been since it began in 1986. He is also a breed surveyor for the GSD Breed Council. He is President and Chairman of the Northern BMD Club. He has lectured on dog breeding in numerous countries, advised police forces and writes regularly for the magazine Dogs Today .
c m.b.willis, 2002