Breeding work for drought-prone environments has been largely empirical to date, with grain yield being the primary trait for selection in wheat breeding programmes. However, most breeders select strongly for traits other than yield in the early segregating generations and do yield testing only at later stages, when a certain level of homozygosity has been achieved and large enough seed quantities are available. The decision to advance or reject a genotype is often complex and, in practical terms, breeders most often use a system of multiple cut-offs. In early generations, they select genotypes that, presumably, achieve the levels required for the primary traits evaluated in segregating populations (plant type, plant height, growth cycle, spike fertility, etc.).
When a breeding programme for drought adaptation is assisted by analytical selection, the conceptual model used considers yield under drought to be a function of: (1) yield potential; (2) flowering date (which indicates whether the crop will avoid drought stress); and (3) secondary traits that provide drought resistance. Physiological secondary traits can be used for the selection of parents to be included in the crossing block, as direct selection criteria for screening among a large number of genotypes (i.e., segregating populations) and/or when the amount of seed available is too small to carry out field trials with replications. Whereas intensive work is continuously being carried out by physiologists in the area of drought adaptation, few breeders routinely use physiological criteria in their mainstream breeding programmes. In the first place, the evaluation of some of the traits proposed by plant physiologists is time-consuming or expensive. This is not practical for application to the thousands of entries that comprise the segregating generations of breeding programmes. Then, the real value of a given trait may only be assessed by determining the genetic gain in segregating populations following selection, while many traits are not available in well adapted genotypes and their validation frequently requires the development of appropriate breeding material, which is again costly and time-consuming (Royo et al., 2005). Finally, selection in segregating populations requires screening at the plant level or between very small plots, thus hindering the use of traits that require large field plots for their assessment.