with average performance in many environments is an ongoing debate among plant
breeders.
3.2.7 Plant Breeding Is a Numbers Game
Before discussing breeding strategies in more detail, it is important to put in context the
scale on which a plant breeder must work. As illustrated above, a breeder may be
working with multiple objectives that require the selection of many genes. The breeder
will seldom know exactly how many important genes are segregating in a population,
but there may be information about some of the genes. An example might be a population
in which the breeder knows that there are a few specific genes affecting disease resistance,
height, and flowering time. Suppose that there are only two genes affecting each of these
traits, for a total of six genes, all segregating in a population derived from a biparental
cross as shown in Figure 3.1. In the F 2 generation, the probability of a specific homozygote
at each locus is^14. If all six genes assort independently, then the probability of a specific
genotype that is homozygous at all six loci is one in 4^6 , or 1 in 4096. A breeder who
wishes to be reasonably certain of recovering this genotype in the F 2 generation would
need to grow many thousands of progeny. Given the fact that many other unknown (or
unpredicted) genes will segregate, and that the true genotype is often obscured by the
environment, it is not unusual for a breeder to evaluate hundreds or thousands of
progeny from a given cross, and to work with many crosses simultaneously. Breeders
remark that finding the perfect variety is like winning the lottery. The fact that they often
“win something” is a result of “buying many tickets,” but the elusive jackpot may never
be won.
3.2.8 Plant Breeding Is an Iterative and Collaborative Process
A common depiction of plant breeding is that it is an ongoing process of gradual improve-
ment, often represented by a gradually upward-sloping graph of yield versus time (e.g., see
Fig. 3.5).^3 The sloping line represents the average of many plant varieties released in a
given year. The measured performance may be historical, in which case it will reflect chan-
ging cultural practice and fluctuation due to “good or bad years,” or it may be based on a
modern experiment in which the performance of older “retired” varieties are tested together
with new varieties in the same environment. While the typical graph represents yield,
many other objectives are selected simultaneously (Section 3.2.5). Therefore, the one-
dimensional progress shown by the graph in Figure 3.5 does not accurately represent
what has been achieved, nor does it account for the fact that objectives and cultural practices
change over the years, such that perfection is a moving target.
One might ask “Why not make the perfect cross and select the perfect pure line and be
done with it?” The first answer is that the perfect cross cannot possibly contain all the best
alleles. Disease resistance may come from one parental source, high protein from another,
stem strength from another, and so on. In fact, the perfect parents have probably not been
(^3) An interesting thing about this study, reported by Duvick and Cassman (1999), was that corn varieties showed no
improvement if they were grown using the same cultural practices (wide rows) that were used in 1930 to accom-
modate the driving of horses between rows. Yet, old varieties performed poorly under the modern practice of
narrow rows. Thus, the genetic gain that was achieved was accompanied by a trend toward narrow rows—a
good example of genotype–environment interaction.
3.2. CENTRAL CONCEPTS IN PLANT BREEDING 57