Breed assignment procedures based on marker allele frequencies and individual genotypes can be used for conservation purposes, fraud detection, analysis of hybrid populations, forensic advice, and paternity tests. The technique ‑allocation of an individual to the breed maximising the probability of that individual belonging to it‑ is straightforward but rigorously speaking it has a number of weak points. We address one of them in this paper: the presence of outlier alleles. Outlier alleles are those having extreme frequencies in a certain breed, with breed specific alleles and missing alleles as particular cases. The approach followed to study this topic was to examine for each locus the behaviour of breed vs. number of alleles contingency tables. The significance of each table and of each cell in each table was estimated by three different methods ‑the usual chi‑square statistic and two simulated ‑MC and MCMC‑ exact tests. In the simulated cases significance was estimated as the proportion of realisations where the probability of the observed table exceeded that of the simulated one, proceeding analogously for individual cells.