The unified species concept and a criterion of limited homogenizing gene flow as evidenced by genetic and morphological markers were applied to species delimitation within Navarretia sinistra. Concordant patterns of variation diagnose two morphologically cryptic species. As a consequence, the basionym Gilia linearifolia is here lectotypified and re-established for this long neglected epithet. Navarretia linearifolia shows strong differentiation from N. sinistra in allozyme data and DNA sequences from chloroplast regions, nrDNA, and introns of the low copy nuclear genes idhA, idhB, and g3pdh. In macroscopic features, N. linearifolia differs from N. sinistra primarily in tendencies, rather than absolute differences. Two finer-scale features are diagnostic: pollen sexine sculpturing and mature seed color. The combination Navarretia linearifolia subsp. pinnatisecta is made for the large flowered populations of this species geographically restricted to the NW region of the California floristic province. The smaller flowered N. linearifolia subsp. linearifolia extends from California to Washington, with a more westwardly distribution compared to N. sinistra, which ranges east into Idaho, Utah, and Colorado.
Estimates of taxon diversity within genera may depart from biological reality for several reasons. Taxon concepts at both generic and species levels may vary by taxonomist. Some species that can be identified readily may have escaped recognition or acceptance by workers because diagnostic features were overlooked. And, conversely, some named species may be based on the inflation of a novel feature that lacks taxonomic relevance. Additionally, some species may be truly cryptic, with substantial genetic change masked behind indistinguishable or nearly identical external morphologies. These scenarios are not necessarily discrete or mutually exclusive. Differences in taxon concepts often stem from differences in opinion regarding diagnostic features, and many cryptic species complexes dissolve upon discovery of diagnostic characters for distinguishing each species. Regardless of the reason, species circumscriptions that reflect biological reality are important not only for accurate assessments of taxon diversity, but also for conservation planning and studies of character evolution, intraspecific phylogeography, and so forth.
As hypotheses, species circumscriptions are testable and a variety of means for assessing and delimiting species boundaries exist (e.g. Sites and Marshall 2003, 2004; Wiens and Penkrot 2002). Criteria that directly or indirectly assess homogenizing geneflow are particularly useful in distinguishing biological breaks between sibling or cryptic species; that is, among evolutionarily separate lineages not easily distinguished morphologically because their diagnosis requires methods beyond, as Cronquist (1988) stated when discussing criteria for species recognition, “ordinary means.” Certain taxa now in Navarretia but circumscribed as Gilia section Kelloggia by Day (1993a) have been confused in a manner consistent with their designation as a cryptic species complex. Grant and Grant (1954) constructed a taxon concept for Gilia capillaris Kellogg that recognized this species as variable in several respects. Day (1993a) noted vegetative, calyx, and pollen differences in G. capillaris as thus circumscribed and segregated the previously established Gilia sinistra M. E. Jones as a distinct species. Day's observations untangled much of the long history of confusion regarding these two species whose morphological similarities are most striking when considered in light of DNA-based phylogenies that indicate they do not form a monophyletic group exclusive of species of the vegetatively distinct genus Navarretia (Johnson et al. 1994; Porter and Johnson 2000). An iterative series of investigations within these taxa has revealed an even more striking instance of cryptic diversity. This work, in conjunction with a thorough review of nomenclature indicates that, like Gilia sinistra hidden by synonymy in G. capillaris, the long neglected name G. linearifolia Howell should be reestablished (as Navarretia linearifolia (Howell) L. A. Johnson) for material recently considered conspecific with Gilia sinistra/ Navarretia sinistra (M. E. Jones) L. A. Johnson (Day 1993a, b; Grant and Day 1998; Porter and Johnson 2000).
Here, a variety of data used to address species limits in this cryptic complex of nonspiny Navarrretia are presented. To facilitate communication, the nomenclature proposed herein (Table 1) is used hereafter, except when discussing names in their historical context. A key to the species of nonspiny navarretias is also provided.
MATERIALS AND METHODS
The unified species concept (de Queiroz 2007), which equates species to segments of separately evolving metapopulation lineages, was used as the basis of species delimitation. Indirect inferences of gene flow were used as a criterion for recognizing such lineages under the premise that, for sexually reproducing species, gene flow will homogenize populations within a metapopulation lineage, whereas the absence of gene flow ultimately will lead to divergence between distinct metapopulation lineages in molecular characters, morphological characters, or both. Allozymes, DNA sequence variation, and morphology were used to assess divergence between N. linearifolia and N. sinistra as circumscribed here against the alternative hypothesis that these two entities compose a single species as treated prior to this study.
Allozyme Data —Allozyme variation was surveyed from 30 individuals per population from 10 populations of N. linearifolia, two populations of N. sinistra, and four populations of N. capillaris (included for comparison; Appendix 1). Uneven numbers of populations for each of the two putative species, Navarretia linearifolia and Navarretia sinistra, were surveyed because populations were sampled before cryptic diversity was suspected. Seven enzymes (10 putative loci) were scored reliably from assays conducted on 11% starch gels using buffer systems as follows. Buffer 6: pgi-1 and pgi-2; buffer 8-: aat-1, aat-2, tpi-1, and tpi-2; buffer 11: idh; buffer 11+: g3pd; buffer M: 6pgd-1, and 6pgd-2 (buffers 6, 11, and M from Soltis et al. 1983; 8-