This LaTeX document is available as postscript or asAdobe PDF.

Dairy Cattle Test Day Models: A Case Study
L. R. Schaeffer

This case study is an example of a patent that has met with considerable resistance amongst the animal breeding research community and its validity has been challenged in Canada and the United States. Dr. Robert W. Everett of Cornell University and the Cornell Research Foundation, Inc. filed for a patent entitled "Method of Bovine Herd Management" on February 25, 1993 in the United States, with the Canadian application filed on February 15, 1994. The patents were granted on October 4, 1994 and July 14, 1998 for the United States and Canada, respectively. The Cornell Research Foundation is on the verge of obtaining a successful application in the European Union. The details of the patent will be discussed. Background information on the dairy cattle industry and milk recording programs is needed to understand the claims in the patents and to judge its validity.

Background

The Beginning of Milk Testing

In dairy cattle, herds enrolled on a milk recording program are visited approximately once per month by a field supervisor. These visits have been called test days (TD). The supervisor weighs the milk for each cow at each milking, takes a sample for fat and protein analyses, and records other events such as calvings, dry off dates, and culling information. Test day information is processed, validated and stored by the milk recording centers to produce a summary of the amount of milk given by each cow to date, and a projected 305-day lactation total yield per cow and per herd. Herd owners utilize these statistics for management purposes to adjust feeding programs, to plan milk income, and to cull poor producing cows from the herd. Dairy producers pay milk recording organizations for these services. This routine has been commonplace in milk recording programs around the world since the early 1900's. Evidence of this early history comes from the National Cooperative Dairy Herd Improvement Program Hand Book which contains a history of dairy record keeping written by D. E. Voelker. The following is a quote from that section (kindly provdided by A. E. Freeman, July 14, 2000):

"In September 1905, a Danish immigrant, Helmer Rabild, working as an inspector for the Michigan Dairy and Food Commission, called a meeting and six to eight dairy farmers from Newaygo County, Michigan attended. As the result of Rabild's knowledge and enthusiasm, the first cow testing association was formed. In 1906, the association's members represented 31 herds and 239 cows. In 1908, when Rabild was working for the dairy division of the U. S. Department of Agriculture (USDA), similar cow testing associations were started in Maine and New York."

There is no mention of test day records, but Voelker stated that the standard lactation length of 305-days was established in 1935. Most likely, Denmark or other European countries must have had cow testing associations prior to 1905.

Genetic Evaluation Methods Using 305-d Yields

Geneticists began to utilize the 305-day lactation yields of cows obtained from milk recording organizations in order to evaluate the genetic merit of dairy bulls almost as soon as the data started to accumulate (USDA 1935). The genetic evaluations were important tools for the artificial insemination (AI) industry to select bulls with superior genetics for production traits. The methodology to analyze the 305-day lactation records has evolved over time to become more and more sophisticated. The application of new genetic evaluation methodology has closely followed the evolution of computing hardware. New technological advances have allowed the application of new theories of genetic evaluation that were previously considered impractical or even impossible. The key issue here is that ideas about statistical models for genetic evaluation purposes generally preceded the applications of those models to real data due to lack of adequate computing hardware to do the job efficiently. Initial genetic evaluations were daughter-dam comparisons based on simple averages of differences in daughter and dams records (Lush, 1931, 1933). Eventually, herd management was found to be a major influence on daughter-dam comparisons. The dam's records were made several years ahead of those of their daughters. The use of a contemporary herd average was proposed by Henderson et al. (1954). The Herdmate Comparison procedure was adopted in the United States in 1962 and compared cows within herd-year-season of calving groups, and combined these deviations across herds for bulls' daughters. The Herdmate Comparison did not account for the merit of a bull's mates, the genetic level of competition within a herd-year-season, or for differential culling among daughters of bulls. The Modified Contemporary Comparison method followed in 1974 which was a patch-work of corrections to the Herdmate Comparison.

A few years earlier (1971), Dr. C. R. Henderson of Cornell University, working with Eastern AI Cooperative in Ithaca, New York, developed a procedure based on best linear unbiased prediction or BLUP methodology (Henderson, 1973). BLUP methods were actually first known to econometrists in 1962 (Goldberger), but Henderson is generally credited with promoting their use in animal breeding since 1963. The advantage of BLUP was that it forced the user to articulate a statistical model that described how the 305-day lactation records were sampled and the factors that could influence yields. The model in 1971 was called a Sire Model. Sires were assumed to be unrelated to each other, and to be randomly mated to dams. Each dam was assumed to have only one progeny. The results were published and called the Northeast AI Sire Comparison by the Northeast Dairy Records Processing Laboratory (DRPL) located at Cornell University. Henderson (1975, 1976) discovered a computational shortcut that allowed genetic relationships among animals to be included in BLUP methodology, and therefore, relationships among sires could be part of the genetic analyses. This discovery was implemented into the Northeast AI Sire Comparison immediately. The success of these models on the genetic progress in the northeastern United States was quantified by Everett and Keown (1984).

Although Henderson first described an Animal Model around 1976, it was not until 1985 that Animal Models were applied to national dairy milk records in Australia, and later in 1989 in Canada and the United States due to the need for computer hardware to catch up to statistical and genetic theory. Animal Models seemingly solved many problems of bias caused by non-random mating. Interbull Bulletin No. 5 (1992) summarized the genetic evaluation procedures used by over 28 countries in the world, and 14 of these utilized an Animal Model of one variation or another with BLUP methodology. The latest Interbull Bulletin No. 24 (2000) shows that nearly all countries use an Animal Model today, except for a few which have progressed to using test day records.

From 1935 to 1999, genetic evaluations in dairy cattle have taken 305-day lactation yields as the standard observation for analysis. A 305-day lactation yield of a cow, however, is estimated from the monthly test day information collected during her lactation. There could be from 7 to 10 or more test day records per cow which are combined to provide an estimate of the total amount of milk given in 305 days. At one time in Canada there was a recording program which collected daily milk yields of cows in particular herds. The 305-day yields were simply the sum of those daily yields. Because the collection of daily yields was very expensive, Canada and the majority of countries collect just monthly test day records.

Making Use of Test Day Records

One of the major contributors to milk recording programs in Canada was Dr. John Moxley of McGill University in Montreal. In 1968 he presented a paper entitled "Within herd evaluation in the Dairy Herd Analysis Service Program" to the Canadian Society of Animal Production Meeting. The abstract of that paper was as follows:

"The Dairy Herd Analysis Service provides a cow rating for each cow in the herd on the monthly test day report. This cow rating is based on 4% fat corrected milk and the rating takes into consideration age, stage of lactation and seasonal effects peculiar to the herd. The rating for each cow is first reported after the cow has had normal tests and is updated on each subsequent test day. Genetic changes in the herd are automatically adjusted for on each test day by using 100 as herd average on each test day. The DHAS cow rating provides a basis for culling while the lactation is in progress. The monthly adjusted herd average test day yield is useful in evaluating seasonal changes in herd feeding and management. Complete and part lactations can be appropriately weighted to provide efficient ratings of AI sires."

Clearly, Moxley was promoting the use of test day information, and adjusting those yields for age and stage of lactation. In meetings of the Canadian Dairy Technical Committee (which reported to Agriculture Canada at that time), Moxley suggested that AI sires could be evaluated using test day records directly (mostly to avoid the confusion about extending part lactation records to a 305-d basis). However, computing technology in 1968, was not capable of handling the very large number of test day records, roughly 10 times more than the number of 305-day records. Also, this was before Henderson's BLUP methodology was widely known and before appropriate methods for analyzing test day records were known. The suggestion was not feasible at that time.

P. D. P. Wood (1967, 1968) began publishing his work with lactation curves making use of TD records of dairy cows. The idea was to find a way of predicting the total 305-d yield from a few TD records. Differences in the shapes of curves between cows were noted, and the possibility of selecting for the shape of the lactation curve was put in front of researchers. Wood also noted that the diet of the cow could affect the shape of the curve. All of this work was done on a phenotypic basis rather than genetic, and either on individual cows or groups of cows. Many other functions have been applied to this problem since 1967 (Ali and Schaeffer, 1987; Schaeffer and Burnside, 1976; Schaeffer et al. 1977).

When did TD records start to be used for herd management or genetic evaluation? The following message was sent to the Animal Geneticists Discussion Group on July 11, 2000 from Les Jones of the Australian Dairy Herd Improvement Service.

"In Australia, farmers have been using production indexes at least since the early 1980s. In this index, the test day yield of a cow on a particular day is corrected for age and stage of lactation and then compared with the average corrected yield on the particular day. The deviations are then combined into an overall measure of merit called a Production Index. Farmers make use of this information in making culling decisions. This approach has had considerable benefit under Australian grazing conditions, where average yields between test days can differ markedly.

My colleague, Kevin Beard, has shown me a covering letter dated August 3, 1981 to a herd recording organization providing information on the Production Index. Kevin examined the appropriate correction factors in his Master's thesis at Melbourne University, submitted in 1983.

The Australian Dairy Herd Improvement Scheme has been using test days in a similar way for the calculation of breeding values since 1985, a year after we started using the Animal Model. (Proceedings of 5th Conference of Australian Association of Animal Breeding and Genetics, 1985. Jones, L. P. Australian breeding values for production characters. Pages 242-247).

Credit for the idea should go to my predecessor, Geoff Robinson. He had planned its use for this purpose before I took over the position in 1982. Clearly this approach makes less use of the test day records than the more recent test day models. It enables us to use a longer herd-year-season than previously, but does not compute estimates of persistency."

See also Jones and Goddard (1990).

Test Day Models

After 1989, computer hardware rapidly changed. Processing speeds increased, computer random access memory changed from kilobytes to megabytes to gigabytes, and disk capacity was skyrocketing. Computing moved from mainframes in big air conditioned rooms to workstations on researchers' desks. By 1991, Moxley's 1968 proposal to directly use TD records in genetic evaluation of AI sires started to fall within the realm of possibility. For anyone deeply involved in milk recording, the idea of analyzing TD records would have been a natural and obvious progression. Also, dairymen started to think more about the costs associated with monthly testing and were asking if supervised tests were essential for accurate estimation of 305-day yields. Could recording schemes be modified to have a combination of supervised and unsupervised recording? Could the interval between test days be greater than 30 days and could the number of TD records per lactation be decreased? Do fat and protein content have to be determined at each test? Kachman and Everett (1989) showed how to correct test day records per cow within herds, followed by work on lactation curves and heritabilities of TD yields (Stanton et al. 1992), and coined the term Test Day Models (TDM). Ptak and Schaeffer (1993) presented a TDM for Canadian dairy cattle based on work started in 1991. The shift to TDM analyses based on the recognized growth in computer hardware developments, and based on a need to reduce the costs of milk recording for dairy producers was an inevitable and obvious occurrence as evidenced by early studies in Canada, Australia, and the United States.

Connectedness of Researchers

An important question was whether the developments in Canada, Australia, and the United States were independent of each other. The actual details of how each country proceeded to TD record analyses were very different in terms of models and procedures, but the idea to work with TD records may not have been independent. The interrelationships between scientists involved in this research were very interwoven. Moxley and Schaeffer were students of Henderson at Cornell, although at different periods, and Everett was a faculty colleague at Cornell. As a graduate student, Schaeffer worked with Everett on TD records for studies on days open. In Canada, Schaeffer and Moxley participated together on the Canadian Dairy Technical Committee in the mid 1970's. Discussion of Moxley's proposal to use TD records arose more than once during those semi-annual meetings. Everett was deeply involved in the DRPL lab and milk recording issues in the Northeastern United States. Interestingly, Everett took two sabbatical leaves in Australia first in 1980-81 and later in 1987-1988, and had to be aware of the Australian use of TD records and their Production Index during those visits.

The University of Guelph and Cornell University held annual colloquia through the 1980's and 1990's. Thus, the research efforts of both groups were known to each other. Whether or not Moxley had any influence on Everett or the Australians is not known, but being deeply involved in milk recording programs in their respective institutions they were probably aware of each others' programs for dairy producers. In my mind, the credit for using TD records in genetic evaluation should ultimately be given to Moxley, but exactly who was first to think about this is not absolutely clear.

The Canadian Test Day Model

Canada introduced a Multiple Trait, Random Regression TDM in February 1999 after eight years of research and trial runs. This system utilizes TD records from the first three lactations between 5 and 305-d in milk. Milk, fat, and protein yields and somatic cell scores are analyzed as separate traits within each lactation. Thus, there are twelve traits in total, assuming that the shapes of their lactation curves were different. A three parameter linear function is used to define the general shape of the curves within time periods, regions of Canada, parities, seasons of calving, and age at calving. Random regressions are in the model to account for the genetic and environmental cow to cow variation around the standard curves. Research on this model is continuing and many improvements can be made. Details of Canada's experience with this model can be found in Schaeffer et al.(2000). The Canadian TDM fulfilled some of the concerns of dairy producers in that cows do not need to have monthly TD records per lactation, and not all TD need to measure fat and protein content. Four TD records in the TDM seem to supply as much information as 10 TD records in the previous 305-d lactation Animal Model system. Several countries have already adopted a TDM or are in the research stages towards a TDM.

Other countries (the United States, Germany, Finland, the Netherlands, Italy, and the United Kingdom) have also been developing TDM during the same time, but have not yet officially adopted a TDM. In the United States implementation of a TDM has been stalled by the Everett-Cornell patent. Outside of the United States, the attitude of researchers and organizations that compute genetic evaluations of dairy cattle has been defiant in the belief that the patent was not valid and that eventually this belief would be proven true.

Summary of Background

TD records have been collected by milk recording agencies around the world for at least 100 years. Since 1935 TD records have been combined into standard 305-d lactation yields that have been used to evaluate dairy bulls genetically so that dairymen could make informed mating decisions about their cows. The manner in which genetic evaluations have been computed has evolved over the years as the result of mutual exchanges and publication of research information that has occurred at a number of institutions around the world, and as a result of technological advances in computer hardware and software. TD records may have first been utilized in Australia in the 1980s for herd management and in 1985 for genetic evaluations. The latest genetic evaluation procedures, based on Moxley's 1968 proposal, have been to directly analyze TD records in what are generally termed Test Day Models (TDM), of which the Canadian TDM is just one example.

The Everett-Cornell Patent

Given the previous background information about milk recording and genetic evaluation systems, one might wonder what aspects could be patentable, who would want to patent anything related to a system that has been viable for almost 100 years, and why? When one thinks about a patent, there is the notion that there must be something new that has been invented which did not exist previously. An invention is usually thought to be a physical entity like a light bulb or mouse trap. The Everett- Cornell Patent does not seem to fit the conventional notions about patents. The invention in this case is a "Method of Bovine Herd Management" which includes the gathering of data, the mathematical treatment of that data, and the subsequent use of the modified data by dairy producers. Thus, nothing physical has been created or built, but rather the invention is a system or process. A process is patentable if it is novel and not obvious. Are those criteria met in this case?

Gathering of Data

The Method of Bovine Herd Management includes the gathering of TD data on a routine basis. This includes all quantitative and qualitative milk production data for each member of a herd. Qualitative milk production data refers to information about number of times the cow was milked on that day, her pregnancy status, health status, temperament, milking speed, birthdate, sire and dam identification, breed composition, and any other information that may become important in the future. Quantitative data refers to the amount of milk, the percentages of fat, protein, lactose and other constituents of milk, as well as somatic cell counts. Milk recording organizations have been gathering data in this manner since the early 1900's.

The collection of data requires the establishment and continuous updating of a database for each cow within a herd. Databases from several herds could be linked or combined for the purpose of computing genetic evaluations. Conceptually, the database could be national in scope as well as the genetic evaluation system. These elements are also included in the patent. Note that Moxley described this exactly in his 1968 presentation to the Canadian Society of Animal Production. Thus, the gathering of data can not be considered novel.

Modifying TD Data

The Method of Bovine Herd Management includes the use of a mathematical herd management model to modify TD data to determine the actual productivity of each cow in the herd. The patent claims that up to 40% of the environmental variation in TD yields can be removed. The model, as given in the patent for the purpose of disclosure only, is

y_ijklmn = H_i + A_j + D_k + P_l+F_m + e_ijklmn,

where y_ijklmn is the test day observation on a production trait (milk, fat, or protein yields, or somatic cell scores), H_i is a test day effect within herd (identifies all of the contemporaries in the herd on test day as well as the environment), A_j is the effect of age (in months) on the day observed, D_k is an effect of days in milk when observed, P_l is an effect of number of days pregnant when observed, F_m is an effect of month of freshening, and e_ijklmn is a random residual effect.

Although not stated in the patent explicitly, only first lactation cows are intended to be included in these analyses. Otherwise, the model would need to have lactation effects and interactions of these with age at calving and days in milk because cows in different lactations, calving at different ages would have significantly different shapes of lactation curves. The residual effects are correlated within a cow, and the patent implies that an autocorrelation structure is assumed between test days on the same cow.

After estimating the differences in effects in the model by generalized least squares, then residuals, $\hat{e}_{ijklmn}$, are estimated as

$$\hat{e}_{ijklmn} = (y_{ijklmn} - H_{i} - A_{j} - D_{k} - P_{l}-F_{m}).$$

Finally, the residuals are combined into a total lactation value, $\hat{t}_{n}$, based on n TD records, as

$$\hat{t}_{n} = {\bf 1}'{\bf RM}({\bf M}'{\bf RM})^{-1}\hat{\bf e},$$

where $\hat{\bf e}$ is a vector of residuals for a cow, ${\bf R}$ is the autocorrelation structure of the residuals, and ${\bf M}$ is described as a mapping incidence matrix, but is not explicitly given in the example. ${\bf M}$ must contain covariances between the observed TD records and total 305-d yields. The modified total yields, $\hat{t}_{n}$, are also stored in the database along with the original TD yields.

The patent discusses primarily the application to milk yields, but the same procedure is assumed to be applied separately to fat and protein yields, and possibly other traits. A further complexity to this approach would be a multiple trait application to all production traits simultaneously. Also, the patent discusses only the use on dairy cows, but TD models have been applied to dairy goats and dairy sheep too. Whether or not these species are covered by the patent is not clear. Finally, could this patent also cover any longitudinal data collected on cows, such as body weights, condition scores, feed intakes, milk urea nitrogen levels, and so on?

The specific within herd model, as described in the patent, was novel at the time. All other developments concerning TD models that have been reported in the literature before and after the patent would not be in conflict with the patent because these models are quite different from that in the patent description. However, there is an 'include all' statement in the patent that reads as follows:

"Since other modifications and changes varied to fit particular operating requirements and environments will be apparent to those skilled in the art, the invention is not considered limited to the example chosen for purposes of disclosure, and covers all changes and modifications which do not constitute departures from the true spirit and scope of this invention."

Therefore, any kind of TDM applied to TD data on dairy cows would be an infringement of the patent rights. This is analogous to someone inventing a mouse trap of one particular design, but the patent also covers any other future, possibly different, design of mouse traps whose true spirit is to catch mice. How can a patent claim rights to all future, obvious developments to improve upon the model that was disclosed?

The word 'art' should be replaced by 'science'. Through science models are developed based on statistical tests and comparisons for the better fit to the data. Hopefully new models explain the biology of the situation too. Should it be permitted to patent a statistical model? Statistical models always have a degree of error in explaining the variation in a set of data. If a patent is granted for a particular model, then should that patent include claims for any future enhanced models for analyzing the same kind of data?

Use of Modified TD Records

The patent describes the use of modified TD records by herd owners to make decisions that could improve the milk productivity of the herd. The decisions that could be made would be

1.

To cull the less productive members from the herd,

2.

To breed the more productive members to the genetically superior AI bulls,

3.

To change the feed formulation and management practices of the herd,

4.

To change the environmental conditions in which an animal or group of animals are housed, and

5.

To mate AI bulls with cows from different herds based on genetic evaluations from the TD data.

Dairy producers have been striving to maximize milk production for over a century through the use of information obtained from a milk recording organization and from national genetic evaluation systems that utilize data from all milk recording organizations in that country. Obviously, genetic evaluations from TD models will be more accurate and will result in better informed decisions by dairy producers, but this is just a part of the evolution of genetic evaluation models that began back in 1931. Everett did not 'invent' anything new, but has merely been one participant in the general evolution of statistical models and procedures to the analysis of TD records. Everett's model is drastically different from the Canadian TDM of February 1999, for example, but in true spirit the Canadian TDM is aimed at the accurate evaluation of genetic merit of bulls and cows using TD records based on an idea from 1968.

Dairy producers utilize more than just TD records and genetic evaluations in their decision making process. Producers want to maximize profitability, and factors they include besides productivity are conformation, health, reproduction, temperament, milking speed, labour, and feed intake. The amount of emphasis placed on milk production would undoubtedly vary from one herd to the next. Ironically, even herd owners that do not participate in milk recording activities can benefit from the genetic evaluations of bulls based on the TD data. The patent can not really dictate how TD information would be used, but rather the patent wants to limit any organization from collecting, analyzing, and dispersing the results without permission from or re-imbursement of a fee to the holders of the patent.

Concluding Remarks

There is an e-mail newsletter produced by Gregory Aharonian that questions the quality of patents that are granted in the United States. He makes a living by searching the literature on the originality of patent applications. A discussion with Aharonian appears in the December 2001 issue of Scientific American (page 33). Aharonian claims that patents are issued without looking at the literature for prior art, and he is publicizing his results. The problems seem to be the growth rate in patent applications, the difficulty in hiring quality examiners, and the lack of funding going to the Patent and Trademark Office. Examiners do not have the time and resources to seek prior art, and applicants are refusing to do much searching on their own. Consequently, a patent is granted that covers all test day model methods and use of test day records instead of a patent that is more narrowly focused on one particular test day model method.

The novelty and obviousness of the Everett-Cornell patent has been seriously questioned. Clearly, the practices of gathering TD data, manipulating it into various forms for use by dairy producers, and the dispersal of TD information to producers have existed for nearly 100 years. The patent claims rights to a practice that has been public knowledge for a long time. The only novel idea within the patent was the specific mathematical model and procedures that Everett and co-workers developed for the analysis of TD yields. Everett was not the first researcher to apply a model to TD records. The model, as described in the patent, is not necessarily the best model that could be applied. Details have been omitted or simplified in the patent description. If the Everett model had explained 100% of the variation in TD records, then a patent would be justified, but unfortunately, every statistical model is just an approximation to reality, and has a degree of error associated with it. Most scientists publish their models in order to stimulate other researchers to find fault or to improve upon their model.

In March 2001, the Canadian Dairy Network submitted a formal request to the Canadian Intellectual Property Office for a re-examination of the validity of the Everett-Cornell patent. In accordance with the Patent Act, the Canadian Intellectual Property Office appointed a Re-examination Board in June 2001, and on October 31, 2001 this board concluded the following:

"In accordance with Section 48.2(2) of the Patent Act, the Re-examination Board has reviewed the request for re-examination of patent 2,130,471 and has determined that it raises a substantial new question of patentability regarding claims 1 through 19."

The Cornell Research Foundation is allowed to respond with its own submission and arguments prior to January 31, 2002. The final decision is expected by March 31, 2002.

Similarly, the USDA asked for a review of the validity of the patent a few years ago, but the judge upheld the patent. Clearly, legal viewpoints about the patent are different from the 'logical' viewpoints of animal breeders. From a scientific viewpoint, the claims in the patent are a copy of other people's ideas, except for the particular mathematical model (and even that may have been borrowed from the Australians).

Several countries in Europe have proceeded to implement a TDM for genetic evaluation of their cattle. Germany, Finland, the Netherlands, Italy and the United Kingdom are proceeding with their own TD models. Approval of the Everett-Cornell patent in Europe should be much more difficult than in North America. According to Article 52 of the European Patent Office (www.european-patent-office.org/online)

1.

European patents shall be granted for any inventions which are susceptible to industrial application, which are new and which involve an inventive step.

2.

The following in particular shall NOT be regarded as inventions within the meaning of paragraph 1.

(a)

Discoveries, scientific theories and mathematical models;

(b)

Aesthetic creations;

(c)

Schemes, rules and methods for performing mental acts, playing games or doing business, and programs for computers;

(d)

Presentation of information.

3.

The provision of paragraph 2 shall exclude patentability of the subject-matter or activities referred to in that provision only to the extent to which a European patent application or European patent relates to such subject-matter or activities as such.

However, the Everett-Cornell patent is close to being approved by the European Patent Office. Why? If the patent is granted in the European Union, then there could be many legal battles with each individual country in the years to come.

If the patent was valid in every country, then there could be serious implications for genetic evaluation programs. For Canada, royalties would have to be paid to the Cornell Research Foundation and maybe some penalty fees which Canada can ill afford. The Genetic Evaluation Board in Canada would have to decide whether to continue with a test day model and pay a license fee to Cornell or revert back to a less accurate animal model using 305-day yields. In other countries, research on test day models may cease altogether as no one will be able to or want to pay royalties. This would be a serious setback to the overall genetic improvement of dairy cattle around the world, except for animals evaluated by Cornell University.

The best-case scenario is that the Everett-Cornell patent will be found invalid due to the existence of published documents and scientific papers as evidence of prior art. Hopefully, if the Canadian Dairy Network is successful in its re-examination of the patent, then they will share their legal documentation with other countries. The Everett-Cornell patent is an example of a case where the examiners apparently failed to look up the literature on prior art, or if they did, then they did not understand this area of expertise or the background history behind milk recording in dairy cattle. Either way, a patent was granted that should have been thrown out upon the first reading of it.

References

Ali, T. E. and Schaeffer, L. R. (1987) Accounting for covariances among test day milk yields in dairy cows. Canadian Journal of Animal Science 67, 637-644.

Everett, R. W. and Keown, J. F. (1984) Mixed model sire evaluation with dairy cattle - experience and genetic gain. Journal of Animal Science 59, 529-541.

Goldberger, A. S. (1962) Best linear unbiased prediction in the generalized linear regression model. J. Am. Stat. Assoc. 57:369.

Henderson, C. R. (1973) Sire evaluation and genetic trends. Proceedings of the Animal Breeding and Genetics Symposium in Honor of Dr. Jay L. Lush, American Society of Animal Science and American Dairy Science Association, Champaign, IL, pp. 10-41.

Henderson, C. R. (1975) Use of relationship among sires to increase accuracy of sire evaluation. Journal of Dairy Science 58, 1731-1738.

Henderson, C. R. (1976) A simple method for computing the inverse of a numerator relationship matrix used in prediction of breeding values. Biometrics 32, 69-83.

Henderson, C. R., Carter, H. W., and J. T. Godfrey. (1954) Use of contemporary herd average in appraising progeny tests of dairy bulls. Journal of Animal Science 13, 949.

Interbull. (1992) Sire evaluation procedures for dairy production traits practiced in various countries. Bulletin No. 5. 84pp.

Interbull. (2000) National genetic evaluation programmes for dairy production in Interbull member countries 1999-2000. Bulletin No. 24. 111pp.

Jones, L. P. and Goddard, M. E. (1990) Five years experience with the animal model for dairy evaluations in Australia. Proc. 4^th World Congr. Genet. Appl. Livest. Prod., Edinburgh, Scotland, XIII:382-385.

Kachman, S. D. and Everett, R. W. (1989) Test day model with individual herd correction factors. Proceedings of Annual Meeting of the American Dairy Science Association, Journal of Dairy Science pp. 60 (Abstract).

Lush, J. L. (1931) The number of daughters necessary to prove a sire. Journal of Dairy Science 14, 209-220.

Lush, J. L. (1933) The bull index problem in the light of modern genetics. Journal of Dairy Science 16, 501-522.

Ptak, E. and Schaeffer, L. R. (1993) Use of test day yields for genetic evaluation of dairy sires and cows. Livestock Production Science 34, 23-34.

Schaeffer, L. R. and Burnside, E. B. (1976) Estimating the shape of the lactation curve. Canadian Journal of Animal Science 56, 157-170.

Schaeffer, L. R., Minder, C. E., McMillan, I., and Burnside, E. B. (1977) Non-linear techniques for predicting 305-day lactation production for Holsteins and Jerseys. Journal of Dairy Science 60, 1636-1644.

Schaeffer, L. R., Jamrozik, J., Kistemaker, G. J., and Van Doormaal, B. J. (2000) Experience with a test-day model. Journal of Dairy Science 83, 1135-1144.

Stanton, T. L., Jones, L. R., Everett, R. W., and Kachman, S. D. (1992) Estimating milk, fat, and protein lactation curves with a test day model. Journal of Dairy Science 75, 1691-1700.

Wood, P. D. P. (1967) Algebraic model of the lactation curve in cattle. Nature 216, 164-165.

Wood, P. D. P. (1968) Factors affecting the shape of the lactation curve in cattle. Animal Production 11, 307-316.

This LaTeX document is available as postscript or asAdobe PDF.