STATEMENT (Evidence Act 1977, section 92) MAGISTRATES COURT OF QUEENSLAND BEENLEIGH Logan City Council Local Law No. 4 Appeal against destruction order BETWEEN DINO DA FRE Complainant AND LOGAN CITY COUNCIL Respondent STEPHEN COLLIER of University of New England, Armidale, in the State of New South Wales, states:- Expert witness statement 1. I have been asked to provide expert opinion evidence in this matter about certain scientific issues which have relevance to the proposition that dog breed can be determined by identification from physical characteristics when the actual breeding of a dog is unknown. 2. I understand that:- (a) (b) a witness giving evidence in a proceeding as an expert has a duty to assist the court; and the duty overrides any obligation the witness may have to any party to the proceeding or to any person who is liable for the expert's fee or expenses 1. 3. I state that:- (a) (b) (c) the factual matters in this statement are, so far as I know, true; and I have made all enquiries considered appropriate for the matters with which the statement deals; and the opinions stated in this statement are genuinely held as my own opinions; Signed: (Dr Stephen Collier)
Statement of Dr Stephen Collier Page 2 (d) this statement contains reference to all matters I consider significant or the issues discussed; and (e) in preparing this statement, I have complied with the duties of an expert witness as set out above. Qualifications and experience 4. I hold the degrees of M.A. in physical anthropology from Simon Fraser University in Vancouver, Canada awarded in 1983, and a PhD in palaeoanthropology from the University of New England in Armidale, NSW, awarded in 1991. 5. I am a palaeoanthropologist (or physical anthropologist) with a research PhD in the subject, and have been a lecturer in the discipline since 1991. While this discipline has little to do with dogs, apart from issues of their domestication, it is centred on the understanding of the relationship of physical form, behaviour and classification to categories like species. Detailed analysis and assessment of physical traits and their combinations is the essence of making decisions on species and sex assignment. This also holds for analysis of populations within species, as in determining the race and sex of modern human skeletal material, and I have extensive experience in this sort of work in my PhD project and in subsequent research. 6. Such methods rely primarily on statistical analysis of combinations of measurements of physical traits where material is complete enough, in combination with morphological trait assessment. One method of classification, cladistic analysis, utilises presence of selected traits to make conclusions of relationships between groups, but the traits must be carefully selected and weighted according to their assumed diagnostic power. Traits common to all groups under consideration are of no taxonomic use. As a simple example, to distinguish between the chimpanzee and the human, one would not use the hand, the basic form of which is common to all anthropoid primates, but may use the teeth, which show clear shape differences between those two species. 7. My interest in dogs began as a hobby, but I quickly became aware of the prohibited breeds issue, and in particular that issue in relation to the American Pit Bull Terrier. 1 In this matter, I am not charging fees or expenses.
Statement of Dr Stephen Collier Page 3 This was in 1989, shortly before the breed became notorious as a result of the British Dangerous Dogs Act, and the subsequent Commonwealth import ban on the breed. 8. I had met several breeders of the dogs by then, and many of the dogs, and the things being said of the breed by various agencies were at odds with my experiences and scientific understanding. In the last 15 years I have attended American Pit Bull Terrier club shows, regularly read the club journals from NSW Victoria and New Zealand, done extensive reading on the breed, its relatives and history, and visited internationally known APBT experts in Britain and the USA. I have also maintained contact with scientists of dog behaviour overseas, and with organisations like the US Humane Society. 9. Since 1990 I have researched the behaviour of the breed by analysis of available data on dog attacks from Australia and elsewhere. I have used these data as a basis for several submissions to governments and Councils, and have written two articles on the subject, that to date are unpublished. 10. I was consulted by the NSW government during its deliberations on the Companion Animals Act 1998, and have had meetings with two Ministers for Local Government. Subsequent to that, I was asked to address the NSW council rangers conference on identification of bull terrier breeds. At that time there was a lot of confusion of the Bull Terrier and American Pit Bull Terrier. In 2003 I attended the Urban Animal Management conference in Caloundra, and took part in discussions of the Queensland breed identification system. 11. The matters in paragraphs 7 to 10 have given me an understanding of some of the issues of dispute or controversy which exist with respect to identification of dogs whose actual breed is unknown as being American pit bull terriers. Scope of report 12. In this report, I have applied my professional qualifications and experience (paragraphs 4 to 6) to analyse three specific issues relevant to the question as to whether purporting to identify a dog of a particular breed by reference to physical characteristics, where the actual breed of its parents and more remote ancestors is unknown, is a process which:-
Statement of Dr Stephen Collier Page 4 (a) has any validity in terms of known science; or (b) is otherwise capable producing conclusions about breed which have any demonstrable level of factual reliability. 13. In my opinion, the answer to both 12(a) and (b) is unequivocally negative. Scientific basis for dog breed identification 14. The breed or mix of breeds (for a cross bred dog) of a dog is determined by its genetics - it is a direct result of the breed or breeds of its parents and more distant ancestors. Those breed origins will be reflected in the DNA of the dog in question. 15. However, currently it is not possible to determine a dog s breed by analysis of its DNA. 16. Theoretically it is possible to do this. In an article in Volume 304, 21st May 2004, issue of Science Parker et al reported a DNA study of 414 American dogs of 85 breeds, which was successful in correctly assigning 99% of individuals to their own breeds. The team developed a genetic profile for each breed using a variety of DNA markers, and tested each dog s DNA characteristics against the profile. This technique offers great promise in the task of DNA identification of individual dogs, but is not a practical technique that is currently available. The breed profiles were constructed from only 4-5 dogs per breed, so may not be representative of the whole breed even in one country like the USA. 17. However, apart from the fact that the technique is still under development and not generally available even in America, it is possible that, say, Australian dogs of those breeds may have DNA profiles sufficiently different to thwart the technique. The study did not succeed in separating some very closely related breeds, the Tevuren and Groenendael Belgian Shepherds, so it may not be able to distinguish the very closely related American Pit Bull Terrier, American Staffordshire Terrier, and Staffordshire Bull Terrier, which all descend from a common stock without additional genetic input. 18. There is no other scientifically reliable technique to assign with certainty an individual dog to a breed. With most pure-bred dogs the breed is obvious from a
Statement of Dr Stephen Collier Page 5 cluster of physical characteristics that typify the breed. However, with non-show breeds, such as the American Pit Bull Terrier, greater physical variability is the norm, and some individuals may not obviously be members of their breed. Additionally, because there are several closely related breeds, uncertainty can arise as to which breed an individual most closely resembles. When individual dogs have recent ancestors from more than one breed, the difficulties of identification become insuperable when the mixes include the bulldog breeds. Transmission of physical features 19. My understanding of the identification process used by the Council in this case is that it relies heavily, if not solely, on correlating observed physical characteristics of a given dog of unknown breeding with a checklist of physical features derived from a breed standard for the American Pit Bull Terrier published by the United Kennel Club. I attach a copy of an assessment which I am told was carried out on the dog Rusty to which this case relates, and which appears to me to reflect that type of process. 20. Written breed standards are not definitive descriptions of breeds sufficient to provide a template for identification. Rather they are guidelines for the assessment of traits in show competition, and they rely on a consensus of interpretation. They are written to be applied to dogs that are pedigreed examples of the relevant breed. They never were designed to distinguish one breed from another. 21. Physical traits of an animal are a result of its genes in interaction with its developmental environment. Half of an animal s genes come from its mother, and half from its father. As genes are on chromosomes which are paired, one inherited from each parent, genes are paired at each locus, with one inherited from each parent. 22. All dogs have the same compliment of genes, that is, the same number of genes, with the genes performing the same tasks in each individual. However, genes have various versions called alleles which can act in slightly different ways, so it is the total complement of alleles that creates the individual physical and other differences seen between individuals.
Statement of Dr Stephen Collier Page 6 23. For some genes, one allele may dominate the expression of the other allele, which is said to be recessive. In cases of Mendelian inheritance, where a single gene locus determines a trait, the dominant allele (from the father or mother) will be what is expressed, while the recessive allele will have no or little influence. 24. Most traits, however, are influenced by many genes and their alleles, so their inheritance is complex, and a great variety of forms may result where these genes have multiple alleles. Variation in human skin colour is a good example. For this reason, for a trait influenced by many genes, an individual may not closely resemble either parent. 25. A recessive allele nevertheless exists as part of the offspring's own genetics, so as to be capable of being passed down as part of a gene pair to future generations, where the recessed physical feature may reappear. 26. With simple Mendelian inheritance, where a trait is influenced by a single gene locus, recessive alleles, while not expressed, are passed to offspring and maintained in the family gene pool. The recessive trait can be expressed in any generation if an individual inherits two copies of the recessive allele. This is well known in dogs, even when the recessive trait is subject to negative selection by breeders. An example is the white German Shepherd. 27. In this way, a cross-bred animal, particularly where it is the result of several generations of crossbreeding, may exhibit a collection of physical features which reflect the genes of its parents, grandparents and more distant ancestors. 28. Purebred dogs have fairly high frequencies of homozygosity at gene loci. That is, the gene locus has paired alleles of the same type. This is because pedigree dogs with breed registers are effectively closed gene pools with all breeding within the pool. Because of strong breeding selection for particular forms of traits, such as coat colour or head shape, this uniformity is achieved by having high rates of homozygous alleles at the loci influencing those traits. Of course, not all gene loci are homozygous, and there is considerable allele variability within any breed. Allele combinations are resorted in every generation, and a dog will have alleles originating from ancestors many generations back.
Statement of Dr Stephen Collier Page 7 29. For cross breed dogs, the instances of homozygosity and heterozygosity (paired alleles of a different type, one of which may be recessive, and not show up in the physical features of that particular generation) are much more random. 30. To give a practical example of that phenomenon relevant to this particular case, a dog which is the product of one or more generations of breeding from ancestors, none of which contained any element of American pit bull terrier, but some or all of which contained some physical features similar to those of an American pit bull terrier, may be born with a collection of features drawn from those ancestors which give it very substantial similarity to an American pit bull terrier in physical appearance, without in fact having any element of American pit bull terrier in its breeding. The following paragraphs provide further explanation as to why this is so. 31. Because dog breeds are essentially patterns of allele genotype which individuals of the breed match fairly closely, related breeds share many of the patterns, and indeed, all breeds share many. Related breeds have common ancestors, from which they inherit alleles and patterns of their variation, so they have many traits in common, or with marked similarity of expression. There are many bulldog breeds, which have differentiated in the last 100 to 200 years, though the exact history of breeds is often uncertain, and even major aspects of a breed s development can be controversial and not susceptible to ultimate clarification. 32. For example, one school of thought has it that the American Pit Bull Terrier is the original baiting bulldog, while another school believes it resulted from a cross of that bulldog and various terriers. What is clear is that all the bulldog breeds have a strong element of the old bulldog, and it is probable that different strains of what we call the American Pit Bull Terrier have different components, some having terrier infusions and some not. The bull breeds common in Australia include the British Bulldog, French Bulldog, American Bulldog, Boston Terrier, Boxer, Bullmastiff, Bordeaux Mastiff, American Pit Bull Terrier, American Staffordshire Terrier, Staffordshire Bull Terrier and Bull Terrier. In addition, there are breeds in creation like the Australian Bulldog, which is a blend of several bull breeds, and there are the numerous crosses and blends, with heavy bull breeds components, that make up the large population of working pig-dogs. All of these breeds and types have close genetic relationships and share patterns of alleles.
Statement of Dr Stephen Collier Page 8 33. While typically dogs in purebred form are distinguishable from each other by appearance, this is not straight-forward for the most closely related breeds like the American Pit Bull Terrier, the American Staffordshire Terrier and sometimes the Staffordshire Bull Terrier. 34. The major difficulty comes in attempting to identify dogs that are not purebred. For example, a pig-dog may be a Cattle Dog crossed with a bull breed dog, and while it is possible to look at the dog and see both Cattle Dog and bulldog elements in its anatomy, it usually will not be possible to say which bull breed was involved. Many pig-dogs are not first crosses of two pure breeds, but often have 6-8 different breeds in their last 3 generations, 3 or 4 of which may be bull breeds. An example is a litter of pups currently being advertised in The Land. The pups are Boxer X Bullmastiff X American Bulldog. These pups essentially are pure bulldog, by being comprised of 3 bulldog breeds, and they will have inherited alleles that all 3 breeds have in common. I would expect these pups to closely resemble American Pit Bull Terriers, though probably being bigger than is typical. 35. The American Pit Bull Terrier is a bull breed that is generalised in anatomy and not exaggerated in any features. It has an average type of dog conformation, and is remarkably variable in size and in various physical traits such as ear form, colour, eye form, robusticity etc. Some of the other bull breeds do have exaggerated physical traits, but when they are crossed with other breeds the alleles for those exaggerations are resorted and mixed with different alleles, with the result that the trait changes in form towards the more generalised expression. 36. For this reason, a blending of bull breeds, with or without non-bull breed mixture, well may produce conformation which closely resembles that of the American Pit Bull Terrier, even if it has no members of that breed in its ancestry. A dog that is part American Pit Bull Terrier will not be distinguishable from a dog that is part one of the other bull breeds. General principles for the testing of scientific methodology 37. If you were to develop a scientifically valid method to identify a dog s breed, on the basis of its conformation, you would have to:
Statement of Dr Stephen Collier Page 9 (a) Make systematic observations of dogs known to be of the target breed, of dogs of related breeds, and of other dogs. (b) (c) (d) From those observations, select physical traits or combinations of traits that you think are restricted to the target breed. Develop an objective system to score or quantify the presence of the discriminating traits in an individual dog. Make a reasoned judgement as to what sum of trait scores indicates with a high degree of confidence that an individual is of the target breed. 38. These steps allow you to form a hypothesis that your identification method is valid and works to correctly identify individual dogs to the target breed. 39. You then must test the hypothesis, which procedure is at the heart of the scientific method. To test the hypothesis you must: (a) (b) (c) (d) Select a sample of people, say ten, to apply the test to samples of dogs, at least five of each of dogs known to be of the target breed, of dogs of related breeds, and of other dogs. The people performing the test must not know the reason for the test. They should have a similar familiarity with dogs as those who will perform the test in practice (council ACOs), but not be experts in dog judging, anatomy or behaviour. People like veterinary nurses, dog groomers, dog obedience class instructors etc would be suitable. Level of observer error would have to be tested in two ways. A sample, say two, of the testers would have to test five dogs five times each, with an appropriate interval between each test. This assesses the level of consistency of results, and there is a mathematical formula to calculate a reliability score for this. Secondly, as each tester has tested all the same dogs, consistency of results can be assessed. The results of point A must be analysed statistically to get an estimate of reliability of the test. If the above hypothesis testing finds the hypothesis is valid, that is, the test works, a level of reliability must be chosen for its application. If the test results
Statement of Dr Stephen Collier Page 10 indicate that an individual dog is correctly assigned to the target breed with 95% accuracy, it may be judged that the test can justly be applied in practice, but if the accuracy is 75% the test may be abandoned. It is equally necessary to consider false positives. If there is a 5% incidence of incorrectly assigning other breeds of dog to the target breed, the test may be held as valid; but if the figure is 35% it clearly is invalid. The cut-off points for acceptance of the test are matters of judgement, but given the serious consequences of seizing and destroying a dog based solely on its asserted breed, a required accuracy of 99% might be applied. Consistency of the current points system with scientific method 40. It is my understanding that the points assessment system has never been tested in any of the ways above. Most of the identification points relate to many breeds of dog, and some relate to all dogs. 41. It clearly does not involve any unique identifiers of an APBT. For example, Point 7, eyes round or almond relates equally to all dogs. Dog eyes only come round to almond, with nearly all being almond. Point 5 also relates to all but a few short-faced breeds. 42. Many other points require subjective judgement to a degree that makes them meaningless from any perspective of scientific or factual reliability to identify a specific breed, for example, Point 4, Muzzle well pronounced jaws, displaying strength. Dogs are predators that catch prey with their jaws. They must be pronounced and strong. All this really means objectively is that the dog has jaws and the muscles to work them. 43. The points that might relate mainly to APBTs and similar breeds are either absent or not weighted. As all 22 points are of equal weight, a dog can qualify as an APBT Type without having any specific APBT features. 44. In current practice, the real test of whether a dog is of APBT Type is the decision to apply the test, that is, the suspicion that a dog might be of the proscribed type. If the test is applied, almost any dog will score 46 points or more. This is true for Golden Retrievers, Kelpies etc.
Statement of Dr Stephen Collier Page 11 45. Of course it is true that ACOs do not apply the test to Golden Retrievers but only to dogs they suspect may be APBT Types. However, that simply reinforces the unreliability of the process - the suspicion that it is an APBT, coupled with the fact that the checklist is capable of identifying any dog with any level of physical resemblance to an APBT as meeting the required score, means that the suspicion effectively dooms the dog. Dated: July 2005 Signed: (Dr Stephen Collier)