Rethinking classic examples of recent speciation in plants


  • L. D. Gottlieb

    Corresponding author
    1. Section of Evolution & Ecology, University of California, Davis, CA 95616, USA
      Author for correspondence: L. D. Gottlieb Tel: +1 530 752 2864 Fax: +1 530 752 1449 Email:
    Search for more papers by this author

Author for correspondence: L. D. Gottlieb Tel: +1 530 752 2864 Fax: +1 530 752 1449 Email:


This paper reviews the evidence that three pairs of diploid annual plant species are related as progenitor and recent derivative. The species pairs are Layia glandulosa and Layia discoidea, Clarkia biloba and Clarkia lingulata, and Stephanomeria exigua ssp. coronaria and Stephanomeria malheurensis. The three cases are examples of Verne Grant's model of ‘Quantum Speciation’, in which a derived species is budded off and acquires new traits while the parental species continues more or less as before. The derived species differ from their progenitors in different ways and show different modes of reproductive isolation. However, the number of differences between each derivative and its progenitor appears to be few and relatively simple in genetic terms. Comparison of a recently evolved species with its progenitor can reveal what happens during speciation because overall divergence is minimal and the direction of evolution is clear. Evidence from DNA sequences may be particularly useful to recognize additional examples of progenitor and derivative relationship.


More than 30 yr after its original publication, Verne Grant's Plant Speciation (Grant 1971) is still the most lucid, well thought out and complete description and analysis of species divergence and species origins in plants. It developed one of the major themes of his previous treatise, The Origin of Adaptations (Grant, 1963), and also continued the program initiated by Stebbins (1950) that placed plant evolution and speciation in the larger context of genetics and population biology. Thus, I am pleased to be here to honor him at this symposium. My paper describes research by Jens Clausen and Harlan Lewis and some of my own studies that have to do with the discovery and significance of diploid plant species related as progenitor and derivative (P–D).

Grant (1981, pp. 153–169) distinguished several categories of ‘primary speciation’ (excluding hybridization and polyploidy). One is geographical speciation. This concept supposes that allopatric races are the precursors of species because their spatial isolation enables them to put together gene combinations that adapt them to their local habitat without disruption by gene flow from related populations, yielding reproductive isolation as an unselected byproduct. Species formed in this manner are true sister species in cladistic terminology, each diverged in some degree from their common ancestor. Geographical speciation can be bypassed when species are ‘budded off’ from ancestral species via small, locally isolated peripheral populations. Grant labeled this mode of origin ‘quantum speciation.’ He considered that it would be enhanced by inbreeding in the founder populations, and probably ‘rapid and radical in its genotypic or phenotypic effects or both.’ Most founders would have low adaptive value and become extinct, but a few might acquire adaptive traits and give rise to derivative species. Presently recognized examples of P–D relationships appear to be the outcome of quantum speciation.

The significance of speciation from peripheral or founder populations was originally discussed by Mayr (1942). Fifty years later (Mayr 1992, p. 227) he re-emphasized its importance and pointed out that the current cladistic points of view wrongly attempt to define species that have an identifiable derivative out of existence: ‘Consistent with cladistic dogma, a species that by budding has given rise to another species becomes paraphyletic and loses its position as a member of a classification, and its status as species. This principle is based on the assumption that speciation is a process by which the stem species disappears when it splits into two daughter species. However, it is now becoming quite clear that speciation by splitting is not the usual process of speciation. Actually, most species bud off founder populations at their periphery; some of which, through peripatric speciation, become new species. But, contrary to cladistic dogma, the mother species, although now having become “paraphyletic”, does not disappear but continues as before’. Rieseberg & Brouillet (1994) have also discussed this issue as it applies to plants.

Whatever their particular mode of origin, most species arose long enough ago that by the time we study them they are often divergent in a number of traits, reflecting continuing evolutionary processes. Grant (1981, p. 92) stated: ‘Biological species represent a stage in divergence, and other stages of incompleted speciation and secondary refusion of species also exist. Indeed, the correct expectation is that all populations will not be grouped into discrete biological species at any moment in history’. Many years previously, Clausen (1951, p. 90) made a similar point: ‘We find species in all stages of differentiation. The most distinct ones, those that are the end products of the processes of speciation, differ in morphological characters, in ecological fitness to the environment, in the genetic systems by which they control this fitness, and in the homology of the chromosomes … The most normal pattern of speciation, however, is a more or less simultaneous and gradual separation in morphologic, ecologic, genetic, and cytologic characteristics’. Of course, Clausen was referring to cases of primary species, not those that result from hybridization or polyploidy.

Although it is generally agreed that evolutionary divergence is ongoing, the obvious inference has not been widely drawn: we rarely come upon a species soon after its origin. Consequently, it has been difficult to know what any particular species (other than neopolyploids) was like when it originated. There are many rich descriptions of plant species and how they are reproductively isolated by various mechanisms, but there is little specific evidence about the course of their divergence. Thus, it remains critical to examine particular cases of speciation, and to find out whether the general models of the processes are consistent with the facts.

Layia discoidea/Layia glandulosa

Perhaps the first P–D relationship identified in diploid plants was that between Layia glandulosa, a very widespread desert annual of sandy habitats from Baja California, Mexico to eastern Washington, USA, and Layia discoidea, a small, spring-flowering endemic of serpentine soils in a remote region of the inner south coast ranges of California (Clausen et al., 1947; Clausen, 1951, pp. 80–82). Discovered in 1940, L. discoidea was initially thought to belong to the Helenieae tribe of the Compositae because it lacked ray florets, but no known genus seemed appropriate. However, Clausen and his group thought its overall appearance suggested a relationship to Layia, a genus of Madiinae they were studying at the time. The species was found to have eight pairs of chromosomes, similar to many Layias, and it produced vigorous and fully fertile hybrids when crossed to L. glandulosa. The result was surprising because L. glandulosa has large, showy, white (ssp. glandulosa) or yellow (ssp. lutea) ray florets and bracts that enfold the seeds and a pappus of white, flattened, linear-attenuate bristles on the disk florets. Layia discoidea has no ray florets and no enfolding bracts and its disk floret pappus consists of short, stubby, brownish scales. Its lack of ray florets suggests a loss because it is the only one among the more than 100 species of Madiinae without them. In the region where L. discoidea is endemic, populations of ssp. lutea can be found growing in grassy openings and other nonserpentine sites. Both species are obligate outcrossers.

Clausen et al. (1947) crossed L. discoidea with a white-rayed plant of L. glandulosa and grew out an F2 of more than 1200 plants. The morphological differences between them segregated widely, with some plants having no rays and others up to eight rays, each associated with an enfolding bract. Differences in the pappus also segregated. The absence of ray florets/bracts appeared to be governed by two loci and the differences in pappus by at least five. The result was noted widely at the time because ray floret presence/absence was a character generally assumed to have generic significance, yet in this example it behaved as a simple Mendelian trait (Stebbins, 1950, p. 191). Although L. discoidea has no rays, the F2 provided evidence that it carried genes that shortened rays and others that modified their color. Thus, pale yellow rays were found on 11% of the progeny, suggesting that L. discoidea contained genes from ssp. lutea. The vigor and full fertility of the F2 progeny, and the ability of the parental genes to recombine freely with no evident developmental compatibility, indicated to Clausen that the two sorts of plants clearly belonged to a single species, regardless of their morphological distinctness. Layia discoidea was initially treated informally as a subspecies of L. glandulosa, but Keck (1958) eventually assigned it species status, perhaps reflecting his more taxonomic outlook and acceptance of additional criteria for the recognition of species in plants.

We carried out an electrophoretic study of 12 enzymes encoded by 21 genes in the two species to estimate their overall genetic divergence. They proved to have a very high genetic identity with no fixed differences and the same allele in highest frequency at each locus (Gottlieb et al., 1985). Recently, Baldwin (2003, p. 210) analysed the sequences of their nuclear rDNA internal transcribed spacer region. He found that L. discoidea and ssp. lutea comprise a robustly supported sublineage within L. glandulosa. He also found that L. discoidea diverged relatively recently because branches leading to the different species within the genus all showed a constant rate of divergence and the branch leading to L. discoidea was the shortest of them. This evidence, in conjunction with the larger geographical distribution of ssp. lutea, the specialized serpentine habitat of L. discoidea, the developmental compatibility revealed in the F2, and the complete reproductive compatibility between them, suggests that they have a P–D relationship.

Clausen's F2 progeny segregated 13 radiate to three discoid plants and this ratio was his evidence that absence of rays/bracts was governed by two genes. We (Gottlieb & Ford, 1987, Ford & Gottlieb, 1990) re-examined the model because Clausen's pioneering study was incomplete. It did not include replicate crosses or progeny tests to confirm the assignment of genotypes to F2 individuals, neither had it attempted to distinguish the individual effects of the two postulated genes. The results of our genetic crosses confirmed the two-locus model and their independent assortment, as well as the finding that L. discoidea has genes that reduce ray number and size and modify their color. The presence of rays/bracts is dominant and is conferred by the genotype R– G–. Layia glandulosa is monomorphic for dominant alleles at both loci whereas L. discoidea is fixed for the recessive r allele and polymorphic at the second locus (thus, rr Gg and rr gg).

Backcross generations revealed a wealth of variant, intermediate and recombinant forms. Perhaps the most interesting is a novel floret morphology designated ‘gibbous florets.’ Individuals with strong expression of gibbous florets were RR gg and those with weak expression (or none) were Rr gg. Gibbous florets are larger than disk florets but smaller than rays and have a pale yellow color. They have a bilaterally symmetrical tubular corolla with three large outer lobes and two small inner lobes. The outer side of the corolla tube is longer than the inner side, bulging out at the base, hence gibbous, and curving in at the top. They have dense pubescence on the ovary and pappus. They set seed when pollinated. In summary, gibbous florets have aspects of rays (size, color, bilateral symmetry) and disk florets (five corolla lobes, no ligule, ovary pubescence, pappus, anthers) as well as several unique traits (tissue bulges and folds).

The gibbous florets demonstrate that novel combinations of genes can be put together in hybrids between these species to produce an unusual but developmentally integrated structure without evident adverse effects. Such novel phenotypes might be favored in some habitats. In this regard, plants with gibbous florets have been found in at least two populations of L. discoidea. Some populations also contain a few individuals (less then 1% in the populations examined) that have tiny ray florets, 2–4 mm long. Their genetic basis is not known; they may reflect the retention at a very low frequency of certain alleles received ancestrally from ssp. lutea or infrequent ongoing hybridization with local populations of ssp. lutea that grow on nonserpentine sites in the region.

In addition to the difference in presence or absence of ray florets or bracts and pappus type, the two species differ in a number of quantitative traits. Layia discoidea is a smaller plant than L. glandulosa with smaller and fewer rosette leaves, shorter stems with fewer nodes and shorter internodes, and flowers sooner from smaller heads (Gottlieb et al., 1985). Ford & Gottlieb (1989, 1990) also looked into the question of whether expression of these quantitative differences was correlated with ray presence/absence. Two F2 progenies were studied and most of the possible recombinant phenotypes were recovered in both. The segregations revealed that the quantitative differences were largely independent of the presence or absence of the rays or bracts. For example, L. glandulosa has much larger heads than L. discoidea, but a number of individuals in the F2 had large heads without rays and other individuals had small heads with rays. Thus, the loss of ray florets or bracts did not require correlated changes of other attributes of the head.

Clausen came to think of L. discoidea as a possible relict from ancient times and was probably surprised that it was still able to exchange genes with L. glandulosa. Whether it is relict or not, and the evidence from electrophoretic comparisons and Baldwin's studies (Baldwin, 2003) says it is not, the genetic compatibility between the two species permits many experimental studies. One of the most interesting would be an analysis of the ability of L. discoidea to grow on serpentine because it represents a critical adaptation that evolved after the species became separated as a founder population. In principle, the adaptation could be studied by selecting serpentine-adapted individuals in a genetic backcross program that introgressed genes from L. discoidea into L. glandulosa. If such lines could be established, and it might not require more than a few generations of backcrossing, the responsible genes might eventually be identified by positional cloning or quantitative trait loci (QTL) analysis. If they could be identified, comparison of their biochemical functions with the functions of homologues in L. glandulosa might reveal how serpentine tolerance evolved. The purpose of describing such a research program is to illustrate how a P–D pair of species can be studied to learn how the characteristics of a newly evolved species were derived from the genetic legacy of its progenitor. The absence of ray florets in L. discoidea suggests additional worthwhile research. Ray florets are generally thought to attract pollinators. Yet, L. discoidea has retained genetic self-incompatibility even though it lacks ray florets. The species may use different attractants and have different pollinators so study of its pollination biology is appropriate.

Clarkia lingulata/Clarkia biloba

The concept of P–D species was first proposed explicitly by Harlan Lewis during the 1950s and 1960s to explain the results of studies and experiments that he carried out on species of Clarkia. He also deserves credit for being the first to propose that speciation in plants can occur fortuitously without evolution of distinctive adaptations. Although his ideas and experimental results were widely known, they had little initial impact. They were thought to be relevant only to certain kinds of annual plants until Hampton Carson (1971) proposed a similar separation of reproductive isolation and adaptation to account for some cases of speciation in the Hawaiian Drosophila. Carson's proposals were greatly bolstered by Dobzhansky's (1972) favorable review although, some years later, they were found wanting by population geneticists (Barton, 1989).

Clarkia is a genus of annual species that are nearly all native to California, where they generally occur in discrete colonies usually in oak woodlands and adjacent habitats (Lewis, 1953; Lewis & Lewis, 1955). Forty-two species are now recognized, including 32 diploid, seven tetraploid and three hexaploid species. Their phylogenetic relationships are very well understood from the studies of Lewis and colleagues and from recent studies of the sequences of PgiC genes (Gottlieb & Ford, 1996, Ford & Gottlieb, 1999, 2003). The flowers of Clarkia are characteristically pollinated by insects and generally outcrossed, although all individuals are self-compatible and may frequently self-pollinate when insects visit multiple flowers on the same individual; some species are normally self-pollinated. The extent of morphological divergence varies from a difference in a single character between some pairs of species to differences in entire suites of traits that might serve as evidence of generic distinction in other plant groups. Closely related diploid species generally differ by substantial amounts of chromosomal repatterning, primarily involving reciprocal translocations and aneuploidy (n = 5 to n = 9). The chromosomal differences function as strong barriers to gene exchange because they upset meiotic pairing and disjunction in interspecific hybrids and cause very high or complete sterility. Some pairs of diploid species also differ by factors that prevent the formation of hybrids.

Lewis (1973) assembled a list of pairs of diploid species in the genus that he believed had P–D relationships. In general, the derived species was found in a limited area often adjacent to that of its putative parent and its habitat was frequently more xeric. The two species were very similar morphologically and, frequently, the derivative had not been recognized as distinct until experimental hybridizations were carried out. If his argument was correct and the derived species had arisen recently, it was expected to be closely similar genetically to its parent. Lewis’ proposals warranted testing and when I came to California, I set out to examine the extent of genetic divergence between the members of two putative P–D pairs by assessing electrophoretic variability of sets of their enzymes, a technique that had just become available.

The initial analysis involved the self-pollinating Clarkia franciscana, known then from a single population on a serpentine outcrop in San Francisco, CA, USA, and the morphologically similar, outcrossing Clarkia rubicunda (Lewis & Raven, 1958). The electrophoretic study suggested that the proposed relationship between them was not a recent one because C. franciscana was fixed for alleles at six of the eight genes tested that were not found in C. rubicunda (Gottlieb, 1973a). Because C. franciscana is predominantly self-pollinating, it was possible that its divergence had increased because of stochastic factors associated with its inbreeding and likely history of sharp fluctuations in population size. In any event, its large genetic difference from C. rubicunda made its mode of origin uncertain. Clarkia franciscana has since been discovered on the east side of San Francisco Bay. An electrophoretic study showed the second population differed from the one in San Francisco at a number of genes, indicating the species was once more widespread (Gottlieb & Edwards, 1992).

A second electrophoretic test of P–D relationship was carried out on Clarkia lingulata and its putative progenitor Clarkia biloba, a pair of outcrossing species (Lewis & Roberts, 1956, Lewis, 1962). Their genetic identity proved to be high and C. lingulata had fewer alleles and lower mean heterozygosity than C. biloba, consistent with the proposed relationship (Gottlieb, 1974). Clarkia lingulata, with several-thousand individuals in an ordinary growing season, is known only from two sites in the Merced River Canyon of the Sierra Nevada in central California. The sites are at the southern periphery of the much larger distribution of C. biloba. The species are distinguished morphologically only by a difference in petal shape; petals of C. biloba are bilobed and those of C. lingulata are not. However, their chromosome complements differ by a reciprocal translocation and two paracentric inversions and C. lingulata has nine pairs whereas C. biloba has eight. The additional chromosome of C. lingulata is homologous to parts of two other chromosomes in the complement. Hybrids between the species are readily made, but the chromosomal differences result in near-complete sterility.

Two more recent analyses of their DNA divergence also confirmed a close phylogenetic relationship. The first, a study of restriction site variations in their chloroplast DNA, concluded that they were sister species and found that they differed by many fewer mutations than any other pair of species in their taxonomic section (Sytsma & Gottlieb, 1986). A subsequent analysis of PgiC sequences also strongly supported their sister relationship (Ford & Gottlieb, 2003).

To find out whether C. lingulata had different adaptations from C. biloba, Lewis undertook an intensive series of studies. He established experimental sympatric populations at various sites south of their present distributions by sowing a mixture of their seeds, making it possible to examine relative performance under the same conditions of environmental stress (Lewis, 1961). Each site was started with 300 000 seeds in a proportion favoring C. biloba five to two. At the driest sites, the sowings died out rapidly but at several sites populations persisted for at least five generations. At these localities, C. lingulata consistently flowered about 2 wk earlier than C. biloba. But Lewis learned that early flowering in the experimental sites was not sufficient to counter the effects of pollinators that did not distinguish C. lingulata from C. biloba. Because C. lingulata was less frequent, it had a greater probability of receiving pollen from C. biloba than from conspecific individuals. The consequence was that C. lingulata was effectively ‘hybridized out’ of the populations because it lost a much higher proportion of gametes to the formation of sterile hybrids than C. biloba. The result suggested that when C. lingulata became established as a species, it must have been spatially isolated.

In other experimental comparisons, Lewis (1969, 1973) found that when water, nutrients and other factors were not limiting, C. biloba grew larger than C. lingulata, on average, and produced significantly more seeds per plant, and the same was true under various conditions of stress and competition. When water was withheld at the time plants were in bud, more individuals of C. lingulata than C. biloba survived to produce seeds, but the surviving C. biloba plants produced more capsules and more seeds so there was no overall difference in their total seed output. These experiments, and others involving reciprocal soil transplants, suggested to Lewis that C. lingulata was not better adapted than C. biloba to the sites it occupies and, he argued, the earlier flowering of C. lingulata was not necessarily advantageous. It was not an advantage during normal growing seasons because plants that flowered earlier were smaller and had a lower seed set. It was advantageous only when drought shortened the growing season. In any event, early-flowering individuals of C. biloba were also found in the Merced Canyon and Lewis found that earliness was readily selected, so it could not be regarded as a unique adaptation in C. lingulata.

The studies on the two species and others led Lewis to propose that diploid speciation in the genus followed a common pathway in which closely related pairs of species bear a relationship of parent to offspring and not one of siblings (Lewis, 1962, 1973). Recent derivatives originated at the margin of distribution of the parental species, and the new species and its parent were unable to form stable mixed populations because sterile hybrids were formed as readily as intraspecific offspring. Consequently, the derivative must have arisen in isolation. However, isolation, in these cases, resulted from ‘catastrophic selection’ whereby nearly an entire population (except for one or a few individuals) was suddenly eliminated by an extreme and abrupt change in the local environment.

The novel aspect of his model was that genetic isolation did not originate through colonization of a site geographically separate from the parental area. Rather, it resulted from near elimination of parental plants at a local site and, highly important, the survivors and their descendents became isolated in a habitat to which they were already adapted. Although Lewis envisioned this situation arising from a catastrophe, it is likely that periods of adverse climatic conditions often cause temporary reductions in population range and numbers of individuals. Subsequent selection for full fertility and chromosomal structural homozygosity could occur without competition from the parent. Once a fertile derivative had evolved, the parent would be excluded because any migrant would produce only sterile hybrids. Similarly, the derivative could not invade populations of the parent until and unless barriers restricting hybridization evolved. Lewis presumed most survivor populations would be ephemeral but occasionally one of them, and the best case was C. lingulata, might evolve into a new species. Speciation was viewed as a byproduct of the accumulation of chromosomal structural rearrangements and was independent of the formation of ecogeographical races. Such races also evolved in several species of Clarkia in response to continuous selective pressures associated with edaphic or climatic differences but were not thought to contribute to speciation in the genus. The speciation process was rapid, abrupt, independent of the evolution of new adaptations and ‘largely fortuitous’ (Lewis, 1973, p. 169).

Lewis’ explanation of how C. lingulata originated was a direct challenge to most models of species origins. The challenge was not to the notion that species were reproductively isolated, after all C. lingulata was very well isolated from C. biloba and clearly passed ‘the test of sympatry.’ The challenge was Lewis’ conclusion that the origin was a chance event and that C. lingulata was not better or differently adapted than C. biloba to its own habitat. The significance of this aspect of his argument appears not to have been understood at the time. For example, Stebbins (1970, p. 191) dealt only with the problem of the accumulation of multiple chromosomal rearrangements. He suggested that they arose gradually, one by one, when C. lingulata may have been more widely distributed and that its present distribution alongside C. biloba was a secondary sympatric occurrence. He assumed that the several translocations and the novel extra chromosome were selected because they served ‘as devices for keeping together adaptive combinations of linked genes.’ Verne Grant (1981, p. 159) also described Lewis’ work on the origin of the species and, for him, the case was an example of quantum speciation. He described approvingly Lewis’ cytogenetic analyses and conclusions but also had nothing to say about the apparent lack of distinctive adaptations in the new species.

A difficult problem in Lewis’ model is how to account for the reorganized genome of C. lingulata. Two possibilities were evident (Lewis 1962, pp. 263–265): they accumulated gradually over a long period of time through drift or some selective advantage of the homozygote (the view adopted by Stebbins), or all the differences resulted from a rapid process within a few generations. It proved easier to argue against the first alternative than to account for the second. The most important argument against gradual accumulation is the failure to find any chromosomal structural variability in the populations of C. biloba (ssp. australis) from which C. lingulata was derived. If the species had once been more widespread and had gradually accumulated new chromosomes, it is surprising that the electrophoretic studies did not reveal more than two low-frequency distinct allozymes. Lewis also pointed out that those species of Clarkia which have high frequencies of translocation heterozygotes have alternate chromosomal segregation and show no reduction in fertility. Lewis suggested that chromosomal reorganization occurred rapidly because of the temporary presence of a so-called mutator genotype, a situation that had been found in Drosophila melanogaster some years previously (now known to be a consequence of transposition). There is no evidence for or against this possibility in Clarkia. Active and inactive transposons in introns and 3′ untranslated regions of PgiC genes and in genes encoding enzymes of the anthocyanin pathway have been found in Clarkia (V. S. Ford & L. D. Gottlieb 1996, unpubl. data), but nothing is known of their role, if any, at the chromosomal level.

Stephanomeria malheurensis/Stephanomeria exigua ssp. coronaria

A P–D relationship also characterizes a population of the diploid, obligately outcrossing annual plant Stephanomeria exigua ssp. coronaria and its predominantly self-pollinating derivative Stephanomeria malheurensis (Gottlieb, 1973b, 1979). The two species grow side by side, sometimes with branches intermingled, on a volcanic tuff in the high desert of eastern Oregon, USA. The locality is the only one known for S. malheurensis, but is the most northern one for ssp. coronaria, which has a very large distribution extending to southern California. Their habitat in eastern Oregon is characterized by extreme climatic regimes, with low precipitation, much of it snow, and freezing temperatures between mid-October and mid-April. The harsh winter conditions have selected ssp. coronaria seeds that require freezing to break their dormancy (an adaptation that prevents them from germinating until temperature rises in the spring). The seeds of S. malheurensis appear to lack freezing requirements for germination and field observations indicated that at least some seeds germinate in the autumn; any seedlings would be killed during the severe winters. The number of plants of both species at the site is closely correlated with annual precipitation. Actual counts in the 1970s found a 40- to 50-fold difference in their population sizes between wet and dry years. However, wet years or dry, the number of plants of S. malheurensis never exceeded 2–3% of that of ssp. coronaria. Not only are there fewer individuals of S. malheurensis at the Oregon locality, but their average size is about half that of plants of ssp. coronaria and, for the same shoot dry weight, they have fewer heads. Thus, S. malheurensis responds quite differently from ssp. coronaria to their shared habitat and seems less fit by various measures.

The morphological resemblance of the two species makes it easy to confuse them. Useful morphological characters were not discovered until they were grown under uniform conditions. Then, it became apparent that the seeds of S. malheurensis are about one-third longer and two times heavier than those of ssp. coronaria and have more numerous pappus bristles. The seed size difference was useful in the field late in the growing season. Another useful difference was in the number of florets per head: five or six with narrow ligules in S. malheurensis, and four or five with wider ligules in ssp. coronaria.

Additional quantitative differences between the species became apparent when they were grown under uniform conditions. For example, the larger seeds of S. malheurensis germinate to produce seedlings with larger cotyledons, its leaf rosettes grow for a longer time before bolting, it has larger roots and a higher root : shoot ratio, and its stems are thicker and carry fewer branches that bear about half as many flowering heads. Not surprisingly, in view of its very much smaller population size and self-pollinating breeding system, the quantitative traits of S. malheurensis are significantly less variable than those of ssp. coronaria. Reduced variability was also found in an electrophoretic study of 12 enzyme systems encoded by 25 loci (Gottlieb, 1976). Stephanomeria malheurensis had fewer than half the number of alleles found in the highly polymorphic ssp. coronaria (19 compared with 45) with only a single unique one, and had nearly no heterozygous loci whereas the average locus of ssp. coronaria was about 15% heterozygous.

Reproductive isolation between the species is maintained by four factors (Gottlieb, 1973b): the restriction of pollen exchange resulting from the differences in breeding system (only three plants that appeared to be F1 hybrids were found at the Oregon site during many seasons of study); a crossability factor that reduces seed set from interspecific cross-pollinations compared with conspecific ones by about one-half; a reciprocal translocation; and a small number of genetic incompatibilities revealed by segregation of pollen viability in their F2 (Dobzhansky–Muller incompatibilities, as defined by Fishman & Willis, 2001).

Four F1 progenies were made between different pairs of individuals of the two species (Gottlieb, 1973b). They had an average of 24% viable pollen (estimated by acetocarmine staining); for comparison, experimental hybrids within either species were fully fertile. Pollen viability segregated broadly in an F2 of one of them, with a range from 0 to 95%, and a mean of 62%. The distribution was unimodal with a tail at low values. About 9% of the progeny showed various abnormalities of the cotyledons, including split cotyledons, single, large-lobed cotyledons, and tri- or tetra-cotyledony, suggesting differences between the species in embryogenesis.

Subspecies. coronaria has a sporophytic self-incompatibility system and is obligately outcrossing. However, a seed of the species collected at the Oregon site and grown in the greenhouse in Davis produced a plant that was self-pollinating (Brauner & Gottlieb, 1987). Seeds from the selfed progeny were grown out, but they were small and weak. To maintain the line, the seedlings were crossed to normal self-incompatible individuals of ssp. coronaria. Each generation, vigorous self-pollinating progeny were selected as parents for the next generation. By the F4, a true-breeding, self-pollinating and vigorous line was established. The original selfing plant and the descendent line closely resembled and were fully interfertile with ssp. coronaria. The plants that gave rise to the predominantly self-fertile and self-pollinating S. malheurensis may have arisen similarly by a fortuitous combination of the self-incompatibility allele we discovered with other genes facilitating self-pollination.

When both species are germinated on the same day, S. malheurensis bolts 10–14 d later. Mature plants of both species have the same total number of nodes but, in ssp. coronaria, about 30% are in the rosette and 70% in the stem, while only 55% are stem nodes in S. malheurensis, a difference apparently reflecting its later transition to bolting (Gottlieb, 1982). The differences in bolting time and node partitioning are important because they correlate well with various measures of differences in fitness. Consequently, we conducted an experiment to select early and late bolting lines from within the highly variable ssp. coronaria (Brauner & Gottlieb, 1989). After three generations of selection, the average number of days to bolting was 35.0 in the early line, 45.6 in the late line, and 41.4 in an unselected control line. The time to bolting was closely correlated with eventual flowering, reflecting the fact that both early and late-selected plants produced the same number of nodes on the stem and flowered after a similar number of days. Thus, the late-flowering plants did not display the distinctive pattern of node partitioning characteristic of S. malheurensis. As there is variability within the natural population of the subspecies for both number of days to bolting and node partitioning, the combination of traits in S. malheurensis might be selected from other starting points.

We now know that the presence of S. malheurensis at its single native site cannot be attributed to advantages of size or growth rates or various other attributes such as competitive ability (Gottlieb & Bennett, 1983). Yet, it is not difficult to speculate about an adaptive value for its larger seeds and distinctive plant body. For example, larger seed size may permit seeds to germinate in safe sites that are not suitable for the smaller seeds of ssp. coronaria. Larger seeds presumably require an increased supply of assimilates, which might be achieved by increasing the duration of rosette growth, the number of rosette leaves and the size of roots, and decreasing the number of branches, heads, and seeds. Alternatively, the larger roots of S. malheurensis might permit mining deeper levels of water. Thus, the changes in its morphology, allocation and duration of rosette growth could be interpreted as responses to selection. A different scenario suggests that the present appearance and capabilities of S. malheurensis might simply reflect the genetic characteristics of the particular individuals of ssp. coronaria from which it descended and the consequences of genetic homozygosity resulting from its self-pollination.

The evolutionary process of adaptation in the species may have been truncated by a human-caused fire at its Oregon locality in 1972, which killed old stands of the locally dominant shrubs (big sagebrush and green rabbitbrush) and made it possible for the weedy cheatgrass (Bromus tectorum) to invade and dominate the site within a few years. Before the fire, the population size of S. malheurensis during wet years was estimated in the hundreds. After the fire, its population and that of ssp. coronaria initially increased dramatically, presumably aided by the sudden availability of additional nutrients from the burnt shrubs and several climatically favorable years. However, the same factors also contributed to a very great increase in cheatgrass establishment. Subsequent aggressive competition from cheatgrass, which has been very well documented in the Great Basin (Mack & Pike, 1983, Thill et al., 1984), led directly to a very rapid decline of both species as well others at the site. Cheatgrass is an alien species from Eurasia that invaded the Great Basin in the nineteenth century and overwhelmed many native species. These species were adapted to the recurrent fires that periodically swept the region and maintained open habitats, but not to the dense growth and deep roots of the new weed. Stephanomeria malheurensis was given federal protection as a rare and endangered plant species (listed November 12, 1982). Large seed stocks are maintained in several botanical garden storage facilities and there have been several attempts to reestablish the species in areas cleared of cheatgrass at its native site.


Genetic changes that result in speciation reduce or eliminate gene exchange between formerly interbreeding populations. The changes described in the three examples above involved repatterning of chromosomes, a new breeding system, new morphological characters and adaptation to a new habitat. However, none of the changes seem profound in the sense that they involved large numbers of genes, although there is not enough information to be certain about this and there is no information about the sorts of changes that have occurred, be they in coding, binding, signaling or regulatory patterns. Our intuition that not much has happened reflects the overall genetic similarity between the newly arisen species and its parent, and illustrates the value of identifying species that have a P–D relationship. The conclusion that they represent P–D relationships rather than sister species reflects the concordance of many lines of evidence, chief among them being genetic similarity and apparent discrepancy of the relative ages of the species in each pair. However, it may be that our recognition of a P–D relationship simply reveals that their divergence was relatively recent and that most closely related species would be considered P–D had we come upon them similarly early. This realization does not lessen the usefulness of an identified P–D pair for understanding speciation but does tell us how precious they are. Our assessment of relatively few modifications may be wrong, but this is less likely than would be the case with most pairs of species that are less recently diverged. The similarity between progenitor and derivative species in these examples is important because it has bearing upon our concepts of speciation and suggests that their differences evolved in the derivative concurrent with or after it acquired genetic independence and not in its progenitor.

Table 1 provides a list of proposed examples of progenitor and derivative species pairs. For most of the examples, the available evidence is significantly less complete than in the three cases discussed here.

Table 1.  Proposed examples of progenitor and derivative species pairs
ProgenitorDerivativeTypes of evidenceReference
  1. In many cases, only one or two types of evidence are available; nevertheless, these cases are excellent candidates for additional study. The types of evidence are abbreviated as follows: C, cytogenetics and/or reproductive compatibility; E, electrophoretic assays; G, genetics; I, internal transcribed spacer (ITS)/external transcribed spacer (ETS) sequence; M, morphology; P, PgiC sequence; R, cpDNA restriction enzyme analysis; S, cpDNA sequence; AFLP, amplified fragment length polymorphism markers; RAPD, random amplified polymorphic markers. 1The original conclusion is modified or rejected by new evidence. However, criteria to decide which data set is more appropriate have not yet been established.

Acer pseudosieboldianumAcer takesimenseS, AFLPsPfosser et al. (2002)
Acer monoAcer okamotoanumS, AFLPsPfosser et al. (2002)
Aletes acaulisAletes humilisELinhart & Premoli (1993)
Camassia scilloidesCamassia angustaE, MRanker & Schnabel (1986)
Chaenactis glabriusculaChaenactis fremontiiCKyhos (1965)
Chaenactis glabriusculaChaenactis stevioidesCKyhos (1965)
Cirsium canescensCirsium pitcheriELoveless & Hamrick (1988)
Clarkia bilobaClarkia lingulataC, E, MLewis & Roberts (1956); Lewis (1961, 1962); Gottlieb (1974)
Clarkia borealisClarkia mosquiniiC, MSmall (1971a,b)
Clarkia mildrediaeClarkia stellataC, MMosquin (1962)
Clarkia mosquiniiClarkia australisC, M, PSmall (1971a,b); Gottlieb & Ford (1999); V. S. Ford & L. D. Gottlieb, unpublished
Clarkia mosquiniiClarkia virgataC, M, PSmall (1971a,b); Gottlieb & Ford (1999); V. S. Ford & L. D. Gottlieb, unpublished
Clarkia rubicundaClarkia franciscanaC, E, MLewis & Raven (1958); Gottlieb (1973a)1
Clarkia unguiculataClarkia exilisC, MVasek (1958, 1960)
Clarkia unguiculataClarkia springvillensisC, MVasek (1964); Holsinger (1985)1
Clarkia unguiculataClarkia tembloriensisC, MVasek (1964, 1968); Holsinger (1985)1
Coreopsis nuecensoidesCoreopsis nucensisC, E, M, RCrawford & Smith (1982); Mason-Gamer et al. (1999)1
Crepis neglectaCrepis fuliginosaCTobgy (1943)
Erythronium albidumErythronium propullansE, MPleasants & Wendel (1989)
Gaura longifloraGaura demareeiEGottlieb & Pilz (1976)
Gossypium davidsoniiGossypium klotzschianumC, E, M, RWendel and Percival (1990)
Haplopappus raveniiHaplopappus gracilisCJackson (1962)
Lasthenia minorLasthenia maritimaE, M, ICrawford et al. (1985); Chan et al. (2001)1
Layia glandulosaLayia discoideaC, M, E, G, IClausen et al. (1947); Clausen (1951); Gottlieb et al. (1985); Ford & Gottlieb (1989, 1990); Baldwin (2003)
Mimulus guttatusMimulus cupriphilusGMacnair & Cumbes (1989)
Picea marianaPicea rubensESTsPerron et al. (2000)
Sagittaria isoetiformisSagittaria teresE, MEdwards & Sharitz (2000)
Salix alaxensisSalix silicicoliaEPurdy & Bayer (1995)
Senecio nebrodensisSenecio viscosusE, M, RAPDsKadereit et al. (1995); Purps and Kadereit (1998)1
Stellaria longipesStellaria arenicolaEPurdy et al. (1994)
Stephanomeria exigua ssp. coronariaStephanomeria malheurensisC, E, M, GGottlieb (1973b, 1976, 1979)

Proper assessment of genetic similarity is complex. Historically, it was initially estimated by overall morphological similarity, but this can be misleading when differences are large as, for example, between L. discoidea and L. glandulosa. If interspecific hybrids were made and their meioses studied, chromosomal homology also provided evidence of genetic similarity, as in C. lingulata/C. biloba and the Layias. If the hybrids were fertile, genetic studies sometimes identified genes with individual and quantitative effects, as in L. discoidea/L. glandulosa and S. malheurensis/ssp. coronaria. A progenitor was likely to have a wider geographical distribution and many populations, and a derivative only one or a few populations. However, evidence from such studies did not translate into a quantitative index of genetic similarity.

An index that was widely adopted during the 1970s and 1980s was based on electrophoretic studies of isozyme variability. Electrophoretic comparisons were valuable at that time because they provided a technique to examine directly the products of numerous specific genes in multiple populations and species. It was well understood that allozymes with the same electrophoretic mobility need not be encoded by genes with the same sequences and that electrophoretic differences in mobility do not reveal the number of changes in the sequences. These limitations were accepted because it was not otherwise possible to examine particular gene products. The derivative would have only a subset of the variability of the putative parent with few or no unique allozymes, and the genetic identity between them would be high and similar to that between conspecific populations (reviewed by Crawford, 1990).

More recently, gene sequencing has provided more precise information about genetic similarity and a new method of recognizing P–D relationships. If genes are sequenced from a representative sample of individuals of two closely related species, the sequences from sister species are expected to form disjunct clades. The sequences from a P–D pair may show a more complex relationship, with those from the derivative forming one or more subclades within the whole. When obtainable, this may now be the most convincing evidence of a P–D relationship, especially when supported by similar results with more than one gene and by concordance with other categories of evidence. However, lineage sorting will eventually eliminate sequence evidence of a P–D relationship, depending on population size and other factors, and rapidly evolving genes may not retain sufficient phylogenetic information, making the choice of appropriate genes critical. When sequence evidence reveals a P–D relationship between species that have a similar range and population size, it may be difficult to determine the direction of morphological evolution. Ontogenetic studies or, eventually, studies of the genes responsible may be useful. Baldwin's work (Baldwin, 2003) is providing convincing evidence from sequence analysis for the origin of L. discoidea from within L. glandulosa. Clarkia lingulata and S. malheurensis have not yet been tested.

Understanding the origin of the reproductive isolation between pairs of species is also complex. The task might be simplified by certain recent points of view that call for changes in the definition of reproductive isolation itself. Based on the frequent observations that plant species often exchange genes where they are geographically sympatric without merging, it may be that individual genes or small chromosome segments that are not exchanged are the critical integrating factors that hold a species together. If so, the unit of reproductive isolation is genes and not the entire genome (Rieseberg & Burke, 2001). The three cases presented here show a range of different modes of reproductive isolation. Layia discoidea is isolated by habitat and distance, not by hybrid sterility barriers; consequently, genes governing its serpentine adaptation are probably key. Clarkia lingulata is strongly isolated by differences in chromosomal structure. Stephanomeria malheurensis is isolated by a mixture of factors, including differences in breeding system, chromosomal structure and presently undefined genic incompatibilities. As more species are studied, it is expected that the mechanisms of reproductive isolation will prove diverse.

The identification of adaptive differences between progenitor and derivative species is the most difficult challenge. It is not sufficient to point out that they grow in different habitats and then claim they are adaptively distinct. Such a claim may be correct but it is not informative. In the context of studies of speciation, the issue is to identify the specific adaptive difference and learn how it was assembled, what genes were involved, and what changes in the genes resulted in the morphological and physiological traits that are the basis of the adaptation. At the same time, one must carry out ecological studies to learn how the adaptation operates in the habitat.

Of the three species described here, L. discoidea is perhaps the most interesting in this regard. The fact that its genome and that of its progenitor can apparently be freely recombined with no evident disruption and no reduction or loss of viability or fertility suggests that it might be possible to determine the basis of its adaptation to serpentine habitats. We do not know how various recombinants would fare in nature but the discovery of plants with the gibbous floret phenotype suggests at least some may do well. Further, the means by which L. discoidea is successfully pollinated despite the lack of showy rays can be determined and the possibility that the loss of rays is itself adaptive for a new pollinator can be tested. The evidence of close relationship to its progenitor suggests the entire suite of morphological and physiological differences that led Clausen to regard the species as an ancient relict may be governed by a small enough number of genes that they can be mapped and individually identified. The species shows clearly that striking morphological differences can be assembled from relatively few genes without adverse pleiotropies and the change might not be profound from a genetic standpoint.

Clarkia lingulata is very strongly isolated from its progenitor because of numerous differences in its chromosomal structure, but it has only a single, distinct morphological trait and seems to have the same set of adaptations as its progenitor. The lesson from C. lingulata is that a plant species may not require novel adaptations to survive in its parent's habitat and, if this is true, then reproductive isolation can be fortuitous as Lewis thought. Nevertheless, its genetic independence and small population size may have set the stage for future evolutionary changes.

Stephanomeria malheurensis has a distinctive morphology that may confer a novel capability in the habitat that it shares with its parent. If so, the niche for large seeds or large roots appears to be infrequent at its present site. The difficulty of interpreting the ecological significance of different morphologies is well known. The lesson from S. malheurensis may be that a novel adaptation does not necessarily increase long-term fitness, especially in a climatically harsh and unpredictable environment. More generally, the example reminds us that species originate in an ecological context so their habitat at the time of their origin plays a major role in the probability of their persistence.

The three species all appear to be examples of quantum speciation in the sense that a derived species has budded off and acquired new traits while the parental species continues more or less as before. Although Mayr (Mayr, 1992) came to regard this mode of speciation as the most common one, we still do not have enough evidence to decide one way or the other. Sequence evidence, as discussed, will probably be very helpful. We also have to learn whether the concept of quantum speciation is relevant to the origin of perennial plant species.

Clarkia lingulata served as the primary example of quantum speciation (Grant, 1981, p. 158) because it exemplified a parent–offspring relationship, seemed to have originated rapidly, and involved ‘drastic’ fluctuations in population size in an area peripheral to the parental distribution. Grant paid no attention to Lewis’ evidence that the species was not better adapted than its parent and that its origin seemed fortuitous. Perhaps this is not surprising because Lewis’ view of the relationship between chromosomal rearrangement and adaptation changed during the time he was studying Clarkia. In the course of a vivid overview of the genus (Lewis, 1953, p. 14), he stated the orthodox point of view: ‘Structural rearrangement of chromosomes is of prime importance as a mechanism which permits adapted gene combinations to persist immune from recombination and for this reason is probably the single most important factor in the evolution of Clarkia, particularly with respect to species formation.’ Twenty years later, he changed his mind. He noted that a ‘neospecies’ such as C. lingulata is not better adapted to extreme habitats than ecologically marginal populations of its parent, and is often less well adapted than its parent to the sites it occupies (Lewis, 1973, p. 165). Consequently, its particular chromosome arrangements cannot be acting to prevent the disruption of adaptive gene combinations because there is no evidence that they harbor such gene complexes (Lewis, 1973, p. 166).

Grant placed S. malheurensis in the category of allopatric speciation involving self-fertilization rather than quantum speciation (Grant, 1981, p. 163) and proposed it arose somewhere other than where it is now found. This is unlikely because the species has not been found during diligent searches of the entire region by many collectors, especially after its designation as a listed endangered species. If it had arisen elsewhere and somehow migrated to its present site with ssp. coronaria, it is surprising that electrophoretic study found only a single rare unique allele. The self-pollination of S. malheurensis provides the same isolation as spatial separation.

Layia discoidea, although probably not an ancient relict, may be the oldest of the three species on the basis of its relatively numerous populations, adaptation to a novel habitat and more numerous differences relative to ssp. lutea. However, the evident genetic and developmental compatibility with its parent demonstrates that its origin was not ‘radical in its genotypic or phenotypic effects’, and does not support Grant's (1981, p. 155) characterization of quantum speciation. Presumably, L. discoidea budded off from a peripheral population of its parent and became adapted to serpentine habitat, but this capability was achieved without radical changes. In this example, as in the others, the relationship between a species and its environment is central to understanding what happens during speciation.

The general lesson from the review of these studies is that quantum speciation remains an important and useful concept that can be applied to a greater variety of situations than initially envisaged. The genetic and chromosomal changes need not be as radical as once supposed, and a new species may originate and persist even without novel adaptations. It is also apparent that the mechanisms that facilitate the origins of species are no different from those that lead to changes within species, except in the magnitude of their consequences.

The study of plant speciation should be conceived as an inquiry into both the origin of new species and the acquisition of their characteristic properties. The best materials are species related as progenitor and recent derivative. Clausen (1951, the Introduction) put it this way: ‘One can use living plant species for experimental analysis and try to take them apart and put them together again in order to discover what makes them function as species’. By taking a species apart, Clausen meant that he could clone hundreds of individuals from an F2 progeny between an alpine and foothill population of Potentilla glandulosa and compare their appearance, growth responses and reaction norms in the three Californian transplant gardens. Differences between species could be studied in similar ways. In this regard, it should be remembered that many congeneric perennial species can be hybridized and their progeny are vigorous and fertile. Clausen (1951, p. 161) realized he could take a species apart, but he admitted he did not know very much about the details of what keeps a species together. With our modern genetic and molecular technologies, we are in a much better position to find out. Fifty years later, the analysis continues.