Department of Plant Sciences, University of California-Davis, 1 Shields Avenue, Davis, CA 95616, USA; e-mail: tmgradziel@ucdavis.edu

Historically, the term ‘breed’ referred to a type of domesticated animal such as the Clydesdale horse that has been selected for specifi c phenotypes or welldefi ned traits. The term ‘breeding’, in turn, refers to the selection of parent combination to achieve the desired phenotype in subsequent offspring. Just as natural selection can result in the gradual evolution of individuals and populations towards greater fi tness within the selecting environment, human selection of parental combinations and resultant progeny can result in pronounced phenotype changes in individuals and populations, which can occur relatively rapidly depending upon the intensity of selection. A primary objective of most breeding approaches is to maximize the desired response to selection. In plant systems, the goal of breeding is also the development of an improved phenotype which is often referred to as a ‘variety’, or more specifi cally a ‘cultivar’ (derived from ‘cultivated variety’) to distinguish it from the more broadly defi ned ‘botanical variety’. Because most perennial, woody plants, including all stone fruits, can be asexually propagated, a typical cultivar is usually a single genotype which may be the result of selection over a very large number of years and/or from a very large population of progeny. For example, virtually all commercial sweet orange (Citrus sinensis) plantings are essentially asexual propagations of a single ancestral genotype (Soost and Roose 1996). Chance mutations leading to improved phenotypes (improved fl avor, sweetness, color, later maturity, etc.) among the millions of otherwise clonally identical trees cultivated over the past several hundred years have been discovered and, if found to be trueto-type following asexual propagation, are often propagated and distributed as new orange cultivars such as the Washington Navel and Valencia cultivars (Nicolosi et al. 2000). Recent evidence indicates that fi g (Ficus carica) has been cultivated for over 11,000 years (Kislev et al. 2006) supporting a very early domestication of fruit crops. In addition, many modern fruit cultivars have been cultivated continuously for hundreds to thousands of years since their initial selection (Janick 2002, 2006), presumably derived from the leading cultivars of their day. The capacity of asexual propagation to essentially capture these rare, horticulturally elite genotypes and, in addition, allow their continued improvement through the accumulation of desirable sports or mutations, offered considerable advantages over

early breeding efforts with cereals and other seed-propagated crops. This is because propagation by seed inevitably results in a risk reshuffl ing of genes resulting in genetically and so phenotypically variable progeny. In contrast, modern cultivar breeding methods including marker-assisted selection (MAS), and marker-assisted breeding (MAB), have proven more effective in crops which are seed propagated as compared to those which are vegetatively propagated. This is due in part to the long seed-to-seed breeding cycle time as well as the relatively low population size restrictions of vegetatively propagated tree crops. However, both molecular-based and traditional analysis of genetic variability are demonstrating that clonal propagation combined with traditional clonal selection strategies is among the most effective methods for capturing the full range of genetic potential, including additive, dominance, and epistatic interactions, for crop improvement (Luby and Shaw 2001; Janick 2006; Ortiz et al. 2006). These and ongoing studies are also shedding light on fundamental reasons why MAS and associated molecular strategies as currently applied, are less effective in tree crop cultivar development.