In the context of domestic animal diversity, the term "breed" is conventionally used for all recognized populations. There are in total about 5000 breeds, a small proportion of which are in planned programs for genetic improvement, while about 30% are at risk of extinction. The primary focus is on the conservation of breeds, including management for better utilization (breeding programs) and conserving those at risk, with the aim of minimizing the loss of among breed diversity. The majority of the world's breeds are in the developing countries, with production environments that are low to medium input and high stress, and each may be expected to have adapted to its specific environment. Empirical evidence strongly supports the expectation that the genetic basis of population differentiation for fitness traits will be nonadditive, with different adaptive gene complexes evolved in each breed. Genetic improvement programs therefore should start with an adapted population, with selection then for production traits. As not all breeds can be conserved, priorities must be determined. Suggested criteria (breed divergence, risk of breed extinction, breed "merit," and within-breed variation) are presented and evaluated. Similar approaches may be appropriate for the conservation of genetic diversity of forest trees.