Blanchard (1907) observed that “Diverging and intergrading forms are abundant in the genus” (page 102), and Fernald (1946a, p. 125) remarked that “no genus in North America, except of course Rubus and Crataegus, has offered so much perplexity and has such contradictory treatment as Amelanchier”. This genus has been taxonomically difficult because there are relatively few taxonomically useful morphological characters, and some of these characters, such as the leaves, are quite variable (Campbell et al. 2008). Moreover, many of the taxa have not diverged much from one another (Campbell et al. 1997a), many of them hybridize with one another, and several Amelanchier are apomictic and can produce seed asexually (Campbell et al. 1985, 1987, Weber and Campbell 1989, Dibble et al. 1998, Campbell 1999). Relatives of Amelanchier – apples (Malus), cotoneasters (Cotoneaster), hawthorns (Crataegus), and mountain ashes (Sorbus) – also contain apomicts and are taxonomically challenging (Campbell and Dickinson 1991).
      Apomixis is usually associated with polyploidy, hybridization, and the occurrence of groups of individuals or populations that are minimally differentiated from one another, that are sometimes called agamospecies or microspecies, and that often confound efforts to develop a useable taxonomy (Campbell and Wright 1996).
      Apomixis was first documented in Amelanchier in 1985 (Campbell et al. 1985). A brief overview of the primary evidence for apomixis in Amelanchier is presented on the Evolution page. Current taxonomic understanding of Amelanchier developed before this year and without consideration for the impact of apomixis and hybridization on diversification. Improvements in the taxonomy of the genus require a full appreciation of the roles of apomixis and hybridization and a species concept that is appropriate to the nature of evolution in this genus. Diversification in agamic complexes, such as Amelanchier, is explained in greater detail on the Evolution page.
      Agamic complexes contain the "most intricate variation patterns known in plants" (Grant 1981, p. 434). Hybridization between apomicts or between apomicts and sexuals can spawn new variants, or microspecies, that may persist through apomixis and polyploidy. Morphologically more or less uniform and minimally differentiated from other taxa, microspecies often undermine efforts to develop a useable taxonomy (Gustaffson 1946-1947, Stebbins 1950, Grant 1981, Nogler 1984, Richards 1986, Dickinson 1999, Bayer and Chandler 2007, Fehrer et al. 2007, Noyes 2007). Microspecies can occur in great numbers, with up to 100 observed in areas as small as a hectare in Crepis (Stebbins and Babcock 1939) and Taraxacum (Kirschner and Stepanek 1994).
      Species are difficult to conceptualize and delimit in groups with both sexual and asexual reproduction (Stebbins 1950, Grant 1981, Levin 2000, Coyne and Orr 2004, Rieseberg et al. 2006). Application of the biological species concept (Mayr 1942, 2000) in agamic complexes leads to taxonomic under-differentiation. Interbreeding is frequent among taxa within agamic complexes, which could therefore be considered giant hybrid swarms and all diversity lumped into one species (Kellogg 1990). Sax (1931) suggested this for some genera of subtribe Pyrinae, which includes Amelanchier.
      Use of a morphological or taxonomic concept by 19th and early 20th Century workers lead to taxonomic over-differentiation and the proliferation of formally named microspecies. Later workers in many agamic complexes eschewed microspecies because morphology was complemented with data about reproductive mode (sexual or apomictic), ploidy level, geographic distribution, and DNA sequences. Hence, for example, European Rubus subgen. Rubus went from over 2000 taxa to about 300 species (Weber (1996), North American Crataegus from perhaps 1200 species to 140-200 (Palmer 1925, Phipps et al. 2003), Antennaria from about 350 taxa to 40 world-wide (Bayer 1989), and European Sorbus s. l. from about 91 taxa to just 12 species (Aldasoro et al. 1998).
      The phenetic species concept (Sokal and Crovello 1970) has been the concept of choice in recent studies of agamic complexes (e.g., Phipps and Muniyamma 1980; Bayer and Stebbins 1982; Sinnott and Phipps 1983; Dickinson and Phipps 1985; Dickinson 1986; Bayer 1987b, 1990; Smith and Phipps 1987, 1988; Soreng 1991; Aldasoro et al. 1998; Dibble et al. 1998; Phipps 1998, McAllister 2005, Nybom and Bartish 2007). The phenetic species concept is often dismissed because the size of the gap between species is arbitrary (e.g., Meier and Willmann 2000). If, however, one accepts some gene flow between species (Rieseberg et al. 2003; Coyne and Orr 2004), then application of the biological species concept requires choosing some level of gene flow. Arbitrariness also pertains to the genealogical species concept (Baum and Shaw 1995). This concept delimits species based on the genealogical history of their genes, and, in its original formulation, it required that all genes within a species coalesce. Coalescence may require a long time to occur (Hudson and Coyne 2002, Syring et al. 2007), or it may never occur within a lineage in some genes if natural selection maintains polymorphism (Coyne and Orr 2004). It is unrealistic to require that all genes coalesce within a species, but what proportion of genes should be required to achieve coalescence? As noted by Coyne and Orr (2004. p. 34): “all species concepts require some subjective judgment.”
      The species concept we favor in Amelanchier is the general lineage species concept, which defines species as “segments of population level evolutionary lineages” (de Queiroz 1998, p. 60). De Queiroz (1998) contended that all species concepts, explicitly or implicitly, define lineage segments and that differences between concepts mostly reflect different criteria for species delimitation. Speciation, which is often complex and temporally extended, could be judged to be complete at different times using different species criteria. A set of populations could, for example, be reproductively isolated from all other sets of populations in a genus, and therefore qualify as a species according to the biological species concept before or after morphological distinctness, gene coalescence, or other species criteria were satisfied. “Because every species criterion will fail to identify separate lineages under certain conditions, the best inferences about lineage separation will be based on lines of evidence described by several different species criteria” (de Queiroz 1998, p. 70). While there is considerable disagreement about species concepts, there is broad support for use of multiple data sources in species delimitation (e.g., Bayer 1997, Mayr 2000, Meier and Willmann 2000, Sites and Marshall 2004, Soltis et al. 2007). Antennaria, one of the most thoroughly studied genera with apomicts, is a model of how understanding evolution and diversification is advanced by eclectic data (Bayer and Stebbins 1982, Bayer 1987a, 1997, Bayer and Chandler 2007).