The findings show that cattle exhibit significant levels of population subdivision. Subpopulation error is larger for strongly subdivided populations than for more homogenous ones. Balding and Nichols' correction, which was applied as if the population was divided into an unknown number of subpopulations, favored the defense by making a DNA match less likely or more error-prone. On the other hand, DNA match probabilities calculated with the simpler product estimator produced results more favorable to the prosecution in increasing the likelihood of a match. The authors suggest the use of an alternative procedure that involves the calculation of a match probability based on a breed's database and the selection of the higher value. Although this method is more complex than the others and requires the knowledge of local cattle populations' allele frequencies, it is an intermediate and more accurate estimation of a match probability. Still, the choice of the populations with the higher value of match probability was slightly more conservative, which favors the defense. The samples included in this study consisted of 15 cow thefts in which DNA profiling was used as evidence to show that the cattle alleged to have been stolen came from the specific stock of the alleged victim. In these cases, the remains of the slaughtered stolen animals were left by the thieves on the owner's farm. Several samples of the remains were preserved for use as reference material for DNA comparisons with the evidence collected from the butchery. 1 table, 1 figure, and 43 references