Chapter 11

Concepts of Species and Genera

G. Ledyard Stebbins

A Brief Recent History of Species Concepts and Methods of Analysis

Before 1838, the date of publication of the first widely used regional flora by a North American botanist, species and species concepts were shaped by contemporary European botanists. As specimens poured into eastern herbaria from explorers returning from the West, however, American botanists cataloged and described them. Engelmann, Gray, and Torrey followed standards set by Europeans, producing monographs and floras similar to previous publications. Others, like E.L. Greene, Aven Nelson, P.A. Rydberg, and M.L. Fernald, as well as M.0. Malte and A.E.Porsild in Canada, adopted narrower concepts of species, describing as a new species any specimen that appeared different from others with respect to one or more characteristics of external morphology.

Consequently, regional floras appeared that differed strikingly in the number of species recognized and the morphological differences used to define them. Floras or partial floras that followed the more classical concepts included B.L. Robinson and M.L. Fernald (1908), N.L. Britton (1901), W.L. Jepson (1925), and L.Abrams and R.S. Ferris (1923--1960). Those that followed the narrower concept of species included P.A. Rydberg (1917) and various monographs, written mostly by Rydberg, that formed the basis of the series North American Flora.

In 1930 the Plenary Symposium on the Nature of Species held at the 5th International Congress of Botany, Cambridge, England, addressed the problem of species. Despite a lack of consensus, the dominant belief emerged that a species is any group of individuals (or of dried specimens) that an experienced taxonomist decides to call a species. Among some taxonomists, this opinion still persists (A.Cronquist 1978).

To resolve the dilemma and establish better connections between taxonomy and biology, two new disciplines came into being: experimental taxonomy and biosystematics. Both were favored by geneticists and cytologists who were increasingly aware of the potential impact of cytogenetic research on taxonomic classification.

Experimental taxonomists sought to determine to what extent visible morphological differences are based on environmental modification, and to what extent these differences reflect genetic differences. In Sweden, G.Turesson (1922) conducted several experiments in which adult plants derived from a single seed (a genet) were divided into several separate divisions (ramets) and raised side by side under the same controlled conditions. He concluded that every widespread species consists of a number of genetically different races, or ecotypes, each of which is adapted to a particular environment. In this ecotype concept, Turesson emphasized the discontinuities between these ecotypes, but other experimental botanists recognized that in most natural species, genetically different individuals can be arranged into a continuous series, within which discontinuities in the environment are reflected to a greater extent than those based on genetic differences between adjacent individuals.

J. Clausen, D.D. Keck, and W.M. Hiesey (1940), following a design set up by H.M. Hall, considered the method valid only if the experiments were frequently repeated in different environments. Consequently, they conducted tests at three field stations (Stanford, 50m above sea level; Mather, 1200m; and Timberline, 3000m). Their results have been widely quoted in the evolutionary literature. For taxonomic classification, the following conclusions are significant: (1) The most modifying effect of the environment is on general vigor and survival, followed by time of flowering and length of winter dormancy. (2) The environment also greatly affects characteristics of general form, such as stem length, leaf size, overall leaf shape, and overall architecture of the inflorescence, including number of flowers. (3) Least affected are more detailed characters such as dentition of leaves, nature of pubescence (glandular versus nonglandular), shape of sepals, size, shape, and color of petals, and size and shape of fruits.

The second series of experiments concerned morphological discontinuities between populations that form the primary basis for recognizing different species. Planned programs of artificial hybridization were pioneered by J.Clausen (1926) in Viola, E.Baur (1932) in Antirrhinum, R.E. Clausen and T.H. Goodspeed (1925; T.H. Goodspeed and R.E. Clausen 1928) in Nicotiana, as well as J.Clausen et al. (J.Clausen 1951) in the Compositae subtribe Madiinae. The possible factors considered were ease versus difficulty of producing hybrids, pollen and seed fertility of F1 individuals, chromosomal behavior, and in some instances, vigor and fertility of subsequent generations. Their interpretations were aided by cytogenetic analyses of hybrids between chromosomally different individuals belonging to the same species (performed by B.McClintock [1932], as well as others [H.B. Creighton and B. McClintock 1931]). As a result, the principal chromosomal differences that cause hybrid sterility were identified as differences in chromosome number and in major patterns of structure, e.g., translocations and inversions.

Biosystematics (W.H. Camp and C.L. Gilly 1943, although the term was often used informally during the 1930s) encompasses all uses of biological information, including observations of gross morphology made on dried specimens, that aid the understanding of species and their relationships. It therefore includes experimental taxonomy and interdisciplinary observations, and it requires a careful synthesis of knowledge.

Important biosystematic methods include observations designed to aid in understanding species as systems of populations rather than as groups of individuals. Representative samples ("mass collection") of individuals from a single population are collected and used to estimate the amount and limits of variability with respect to various morphological characters that exist in a single population. For complex characters such as the color patterns of petals, carefully designed diagrams that lead to easily measured parameters are often helpful. An outstanding example of this method is the investigation of the blue flag Iris species of eastern North America (E.Anderson 1936).

Also important are observations and experiments on pollination dynamics, particularly determining the frequency of self- versus cross-pollination and the presence versus absence of asexual or apomictic seed production. An early effort in this direction was the investigation by J.Crosby (1949) on the distribution of heterostyly versus homostyly in English populations of Primula. More extensive observations by H.G. Baker (1948, 1953, 1955) on these characteristics in Plumbaginaceae and other families led to the generalization that obligate outcrossing is most common in populations of a species that occur near its center of origin, while self-fertilization and apomixis are more common in peripheral populations, particularly those that have migrated over long distances.

Perhaps the most commonly gathered biosystematic data are chromosome numbers and, in species having relatively large chromosomes, gross structural features such as size and length of arms. During the 1920s and 1930s, G.A. Levitzky (1931) and others investigated these characteristics in many taxa: Crepis (E.B. Babcock 1947), Nicotiana (R.E. Clausen 1932; T.H. Goodspeed 1934), Poaceae (N.P. Avdulov 1931), and Ranunculaceae (W.C. Gregory 1941).

The importance of chromosome numbers and morphology goes beyond that of individual characters of gross morphology. For the purpose of placing names on dried specimens, chromosomes can be regarded as "just another character." For biosystematics and evolutionary botany, however, this point of view is too superficial. Although by themselves they rarely provide definitive answers, chromosome numbers can be combined with other characteristics to reach syntheses that are far more significant than those based solely on gross morphology.

Other biosystematic characteristics of particular importance in certain families are relationships to various pollinators, as recorded for the Polemoniaceae (V.Grant and K.A. Grant 1965) and for the Orchidaceae (L.van der Pijl and C.Dodson 1966). In the Poaceae, both N.P. Avdulov (1931) and J.R. Reeder (1957) showed that the biology of seed germination and seedling structure are of primary importance. The biosystematics of Asclepiadaceae are intimately bound up with their relationships to the insects and their larvae that feed on these plants (P.R. Ehrlich and P.H. Raven 1964). The research of D.B.O. Savile (1979) has shown that the nature of fungal parasites can provide valuable biosystematic information about affinities between various groups of angiosperms.

Biochemical kinds of biosystematic information have risen most dramatically in importance during recent years. Much of this information, particularly that concerned with immunological relationships and with gene sequences in chloroplast DNA, primarily concerns relationships above the species level. Significant information with respect to variation within and relationships among species is provided by secondary organic compounds, particularly phenolics and alkaloids, by electrophoretic properties of proteins, particularly enzymes, and to a lesser extent, by some kinds of nuclear DNA.

The importance of phenolics for population studies first became evident from the research on species and hybrids of Baptisia (R.E. Alston and B.L. Turner 1963, 1963b; for a review see J.B. Harborne and B.L. Turner 1984). Variation detected by means of extracting proteins and comparing their mobility on an electrophoretic grid was first recognized as important in humans and Drosophila during the 1960s, and shortly thereafter the method was applied to populations of an introduced grass species, Avena barbata (A.Kahler and R.W. Allard 1970). Since then, the technique has been used with outstanding success (L.D. Gottlieb 1986; D.E. Soltis and P.S. Soltis 1989; D.J. Crawford 1990).

After a few simple hybridizations, progeny analyses of a strictly Mendelian nature were made. Electrophoresis is an accurate method of determining genetic diversity in populations with respect to those proteins that can be extracted and subjected to the technique. These proteins contribute little or nothing to observable morphological differences, but in one way this helps rather than hinders an understanding of the evolution of phenotypic diversity. Although many electrophoretic differences do have some adaptive significance, it is generally lower than that of many other differences, so that correlation of genetic distance between populations with the time elapsed since their descent from a common ancestor can be expected to be relatively high. At the intra- and interspecific levels, protein electrophoresis is currently the most accurate method for estimating rates of evolution.

The use of enzymes that cut the DNA molecule at particular positions in the nucleotide sequence has rendered possible comparisons at the level of the gene structure itself. These methods are relatively laborious, however, and most of the results obtained to date on plant materials are more relevant to differences at the level of genera and families than of populations and species. At this lower level, mitochrondrial DNA has been very useful in animals, but for various reasons, less so in plants. Nevertheless, investigations of the repeated sequences that code for ribosomes have been useful for distinguishing between cultivated varieties of barley (Hordeum vulgare, R.W. Allard et al. 1990) and are likely to become increasingly important in the future.

Chloroplast DNA has, to the present, yielded more interesting results about plant species relationships than has mitochondrial DNA, ribosomal DNA, or the coding DNA of various genes. At the level of higher categories, chloroplast DNA has been a valuable aid to studies of phylogeny in both the Fabales and Asterales due to the presence of inversions, while differences in nucleotide sequence have yielded quite unexpected affinities between the genus Clarkia and the morphologically very different monotypic "genus" Heterogaura, which, if DNA similarity is regarded as the ultimate test, must now be subsumed within the genus Clarkia (K.J. Sytsma and L.D. Gottlieb 1986).

Important results emerging from the application of protein electrophoresis have emphasized the widespread occurrence of mosaic evolution: radically different rates of evolutionary change of different characters within the same evolutionary line during the same time period (G.L. Stebbins 1983, 1983b). G.De Beer (1954) showed that the most famous intermediate between two classes of animals, Archeopteryx, is not intermediate between reptiles and birds with respect to all of its individual characteristics but is rather a mosaic of different characteristics, some of which are typically reptilian and others typically avian. The evolution of birds, as well as of other vertebrate classes, has not been a steady, even progression with respect to all characteristics, but rather a mixture of fast and slow rates, depending on the trait involved. Although the fossil record of intermediate forms between classes of vascular plants is too deficient in diagnostic characteristics to provide a firm basis for determining whether or not mosaic evolution has dominated macroevolutionary changes in plants, the record can be interpreted in this way.

Evolutionists have naturally asked if the evidence from genetic diversity within and differences among populations, as determined by their electrophoretic properties, coincides with or differs from that derived from gross morphology with respect to the rates of evolution that are suggested. Four different comparisons provide an equivocal answer. The first two are in the genus Clarkia (Onagraceae), which consists of many annual species, the majority of which are endemic to California. Upon completion of the first cytotaxonomic investigation of this genus, H.Lewis (1962) proposed that it includes several examples of the sudden origin of new species, which he termed catastrophic speciation, without intervention of polyploidy, which was then widely recognized as a source of this phenomenon.

Two of these examples, Clarkia lingulata (H.Lewis and M.H. Roberts 1956) and C. franciscana, were reexamined by L.D. Gottlieb (1974, 1974b), using electrophoretic techniques. He showed that small differences between C. lingulata and its close relative C. biloba support Lewis's hypothesis (fig. 11.1), but that enzyme differences between C. franciscana and its two nearest relatives, C. rubicunda and C. amoena, which Lewis believed to be older and ancestral, are in fact great enough to justify the assumption that all three species diverged from a common ancestor at about the same time (figs. 11.2, 11.3).

S.I. Warwick and L.D. Gottlieb (1985) investigated enzyme differences between six species of Layia (Asteraceae), another California genus that had previously been carefully investigated from the cytogenetic viewpoint (J.Clausen 1951). In this group they found that evidence from enzyme \ genetics with respect to the age and relative affinities of the species supported completely that obtained from cytogenetics.

Another significant investigation was a comparison by K.Helenurm and F.R. Ganders (1985; Ganders 1989) of differences among species of Bidens (Asteraceae) found in Hawaii as compared to those found on the North American mainland. The Hawaiian species differ strikingly from each other with respect to gross morphology, particularly vegetative characteristics, but enzyme differences between recognized species are no greater than between different populations assigned to the same species. On the other hand, mainland species show the expected pattern: differences between species are considerably greater than those between populations.

These results are to be expected on the assumption that enzyme differences chiefly reflect (1) the length of time during which two populations or species have evolved independently since their divergence from a common ancestor, and (2) fluctuations in population size, since small size of populations favors accumulation of neutral differences via chance events. Meanwhile many characteristics of gross morphology are adaptive and are therefore subject to selection pressures that act independently of size; these pressures may produce changes rapidly or slowly, depending on whether they are strong or weak.

The Problem of Defining and Delimiting Species

The usefulness of a continental flora may depend to some extent on the degree to which contributors are able to adopt similar standards for delimiting species. Ideally, this essay would be most useful if it could prescribe specific standards that all contributors would be expected to follow. Practically, this is impossible for several reasons, the chief one of which is the diversity of biological factors that are responsible for the diverse patterns of intra- and interspecific morphology that exist in the various groups that are treated. On the other hand, subjective opinions differ so much from one botanist to another as to produce anarchy if every contributor were left to his or her own devices. The objective of the present section is not to arrive at a hard-and-fast definition of species, but to review criteria that have been used by various systematists, pointing out their advantages and disadvantages.

Like all floras, the present one is based on the assumption that all populations of the plants that it treats must be accommodated into one of several thousand basic units designated as species. Moreover, with the exception of a few rare, localized entities, each species is a system of populations that differ from each other at least quantitatively with respect to some of their visible and measurable characteristics. Although some evolutionary botanists might wish to recognize as distinct species populations known as "sibling species," which are morphologically alike but either cannot be crossed or produce sterile hybrids if crosses are obtained, recognition of such "species" in a large flora is impractical. Botanists who are treating groups in which siblings have been experimentally demonstrated, as in the genus Holocarpha (J.Clausen 1951), might do well to state that they exist, but not otherwise include them. The current practice of delimiting species chiefly on the basis of morphological discontinuities between several visible traits is justified. Difficulties arise, however, with respect to the extent and nature of these discontinuities.

Many biologists, particularly zoologists, believe that species are nonarbitrary objective entities that exist in nature whether humans are there to describe and delimit them or not. If accepted, this concept assumes that if enough has been learned about patterns of population diversity in nature, the limits of species will be obvious to everyone.

The biological species concept is based on the belief that the great majority of populations consist of groups among which free exchange of genes via hybridization in nature is possible, but between different "species groups" exchange of genes is prevented by the action of one or usually several isolating mechanisms. This appears in nearly all textbooks of evolution (G.L. Stebbins 1950), as well as plant speciation (V.Grant 1971, 1981). The minority of exceptions to this rule are usually regarded as species that will evolve into distinct species in the near future, based on the evolutionary time scale.

For many groups of animals, this belief may be completely justified because of the much greater effectiveness in animals than in plants of two kinds of isolating barriers. One, highly specific courtship patterns, prevents sexual union even between entities that, if cross-fertilized artificially or allowed to mate under unnatural conditions, can produce viable and fertile hybrid offspring. A second is the far more complex patterns of development in animals than in plants. Therefore in hybrids, disharmonious gene action in development is far more common and is likely to produce complete inviability or sterility in at least one sex, that in which the sex-determining mechanism is heterochromosomal (X--Y in males, or W--Z in females).

Another difference in higher animals is the almost universal presence of separate sexes, and the condition that sexual reproduction must always be biparental, while in higher plants, which are usually hermaphroditic (with perfect flowers or monoecy), uniparental reproduction via self-fertilization is often possible. In some evolutionary lines, particularly weedy annuals, it predominates. A cursory review of North American genera suggests that in about 25% of them, most or all species are distinct as judged by the presence of actual or potential gene flow among different populations, balanced by the absence or rarity of gene flow between populations of different species because of hybrid inviability or sterility. In another 25%, the common presence of self-fertilization and/or partial fertility of F1hybrids, and apomixis in a few examples (e.g., Crataegus, Rubus subg. Eubatus), often renders clear definitions of species difficult. In about 50% of the genera, two or three of these different conditions exist in different species groups of the same genus.

Data are not available that would allow us to determine whether or not these deviations from expectation, according to the biological species concept, are correlated with different evolutionary ages of the groups. The conditions mentioned exist among recently evolved groups. The "normal" pattern found in Layia(Asteraceae) probably reached its present condition during the past 10,000 years, with the desiccation of climate that began after the Pleistocene pluvial (glacial) epoch. The same is probably true of many groups of self-fertilizing annuals in California and the Southwest. Most examples of extensive hybrid swarms are also recent; many have occurred since the onset of disturbances caused by human agriculture.

On the other hand, fossil evidence indicates that in woody species the parental taxa that entered into the hybrid swarms have been distinct from each other since the middle or late Tertiary period, 15--25 million years before the present (M.Y.B.P.) (D.I. Axelrod 1958). No convincing evidence exists to confirm that the majority of species that are poorly defined on the basis of observations of natural populations are more recent than are the better defined species. In plants, the condition of arrested speciation or stasis can, in some groups, persist for millions of years.

From these observations, two questions arise. First, what, if any, are the factors that arrest speciation versus those that permit the completion of the process via reproductive isolation? Second, how can these situations be incorporated into species concepts and taxonomic treatments that will both be useful for identification and produce evolutionary information?

Growth Habit, Population Structure, and Reproductive Isolation

Ever since the formulation of the synthetic theory, plant evolutionists have known that in some genera, species recognized by most taxonomists often hybridize to form interspecific hybrid swarms, while in others barriers of reproductive isolation exist between populations usually assigned to the same species (G.L. Stebbins 1950; J.Clausen 1951; V.Grant 1971, 1981). Most genera in which conventional species boundaries may be blurred (Quercus, Arctostaphylos, Ceanothus) are woody, but some are herbaceous (Iris, Apocynum, Elymus). The latter, however, consist of individuals that are relatively long-lived, so that the partial sterility of F1hybrids is not as serious a detriment to reproductive fitness as it is in short-lived perennials or annuals.

On the other hand, clusters of sibling species that are not usually recognized by taxonomists have been found in annuals, both self- and cross-fertilizing. These correlations suggest by themselves that biological properties other than the fertility versus sterility of hybrids should be recognized when species concepts are being formulated and species are being delimited.

Particular attention must be given to examples of species pairs that are sympatric over large areas and differ with respect to absence or presence of reproductive isolation in different parts of their areas of contact. Situations of this kind are particularly common in Quercus (J.M. Tucker 1952; M. K. Cooperrider 1957; W.B. Brophy and D.R. Parnell 1974; R.J. Jensen and W.H. Eshbaugh 1976; V.Grant 1981). Quercus is particularly important because its Tertiary fossil record is very good and indicates that the species involved have been distinct from each other for millions of years, except for local hybridization (D.I. Axelrod 1958). Where evidence of this kind is available, longtime persistence of distinctive morphological and ecological characteristics should be considered as part of the biological properties of species.

A complete analysis of reasons why these differences exist is impractical. Nevertheless, one point appears to be clear. The relationship of growth habit to patterns of speciation is not direct, but acts via the different kinds of population structure, both spatial and temporal, that result from different growth habits. Annual species, particularly in regions where climatic stress varies from one season to another, are most successful in the long run if they always produce the maximum amount of seed that the environment permits. In poor seasons the individual plants are few and small; in good seasons, numerous and large. Great fluctuations in population size are the rule, not the exception. On the other hand, woody plants and long-lived perennials respond to stress conditions by producing little growth and often no seeds at all.

A similar situation holds for migration and the colonization of new habitats. If the colonists are annuals and self-fertilizers or apomicts, the first individual colonist will usually produce a large crop of seeds, and its descendants will capture the area via a rapid increase in population size. In contrast, if the new colonist is a tree, shrub, or rhizomatous herb, it may survive by vegetative growth and propagation for many years, and its descendants may capture the environment by a relatively slow increase in population size, accompanied by high, aggressive competitiveness of individual plants.

The final connection between growth habit and degree of reproductive isolation can be established by the properties of populations first proposed by S.Wright (1931) and reaffirmed by many other population geneticists. Repeated reductions of populations to small size promote the establishment of neutral differences between them both in genes and in chromosome structure. As many experiments of interspecific hybridization have shown, the latter are the commonest source of reproductive isolation between plant species. In this connection, the same species groups that contain few and weak reproductive isolation mechanisms also are usually homoploid and relatively homogeneous in karyotype. The same ecological characteristics that are associated with strong reproductive isolation barriers between species having the same chromosome number are also associated with polyploidy.

Toward a Broader Biological Species Concept

Before considering the bearing of the above information on species concepts, two specific examples need to be reviewed. They represent situations that are not uncommon and that constitute serious barriers to the acceptance of the biological species concept as usually formulated.

The first of these is the complex of Vicia sativa. It consists of weedy, self-fertilizing annuals, native to southern Europe, but extensively introduced throughout much of North America. Most American manuals recognize two introduced species, V. sativa and V. angustifolia, and one or more varieties. In Flora Europaea (P.W. Ball 1968), both of these, plus several other entities often regarded as species, are placed as subspecies in the single species V. sativa.

A thorough review by D.Zohary and U.Plitmann (1979) of the cytogenetic literature on this group states that several entities within it are separated from each other by barriers that are usually regarded as delimiting species. Three different somatic chromosome numbers exist, 2n = 10, 12, and 14. Furthermore, entities that have the same number may differ greatly with respect to gross chromosome morphology. Not unexpectedly, therefore, F1 hybrids between such entities have irregular meiosis and low fertility of both pollen and seeds. Nevertheless, the plants are so vigorous that they produce some seeds via self-fertilization that give rise to progeny of F2 and subsequent generations. These are derived from individuals that quickly recover full fertility while retaining intermediate phenotypes with respect to the characteristics by which the original parents differ.

This result is possible because the genes responsible for chromosome differences segregate independently of those that code for different morphological characters, a condition that has been well known ever since progeny derived from partially fertile interspecific hybrids were analyzed (G.L. Stebbins 1950). The persistence of such intermediates is the basis for treatments by P.W. Ball (1968) as well as by D.Zohary and U.Plitmann (1979), in reducing to subspecies rank Vicia sativa and V. angustifolia. Zohary and Plitmann pointed out further that some of the subspecies that they recognized include nonweedy natural populations that differ from each other more widely than do the different subspecies that are more common and weedy. Apparently, adaptation to habitats created by humans, by favoring hybridization plus establishment of fertile derivatives, has promoted evolutionary convergence rather than divergence.

The second example is from the genus Quercus (fig.11.4). Quercus douglasii is the commonest species of tree growing on the foothills that surround the Central Valley of California. In various parts of its range, it is sympatric with at least four other species of Quercus, and hybrids with all of them have been recorded with varying frequency. Its leaves are deciduous, moderate in size, shallowly lobed, and bear trichomes.

For 250 km along the western margin of its range, extending from northwest to southeast, Quercus douglasii is sympatric with Q. dumosa (including Q. turbinella, which hybridizes freely with Q. dumosa and also with Q. berberififolia and others), shrubs having evergreen leaves with spiny margins and stellate trichomes having 2--9 branches, as compared to 3--5 in Q. douglasii. This species pair was studied by J.M. Tucker (1952b). He found that in the northern part of the range of sympatry, the two species form occasional hybrids, without producing hybrid swarms. Therefore, the two sets of populations are completely discontinuous and easily separated from each other both morphologically and ecologically (Q. douglasii grows on open grassy hillsides, and Q. dumosa among other shrubs on shady canyon sides). In the central and southern parts of their ranges, however, hybridization is more common, and the distinctness of the two species depends on the degree of overlap.

No evidence suggests that these species are becoming merged into a single population system, in spite of the extensive hybrid swarms that occur in certain areas. Fossil leaves that closely match both species are known from formations that date back to the early Pliocene and late Miocene epochs, 15--9 M.Y.B.P. Moreover, the other woody species associated with these fossils are the same as those that are associated with each species in its typical modern habitat, indicating that the adaptive properties of the late Tertiary ancestors of the modern species were the same as those of their present descendants.

These two examples show that over evolutionary time, given certain habitat changes, well-developed barriers of reproductive isolation may not promote the divergence of populations. On the other hand, poorly developed genetic barriers may persist for long periods of time, even when populations are spatially sympatric, while the populations involved still remain, to a large degree, distinct from each other.

Each of the two examples described above has counterparts in several genera characteristic of the North American flora. Perhaps the largest species group that resembles Vicia is Panicum subg. Dichanthelium (R.Spellenberg 1975). Ceanothus, a woody genus having large numbers of species in California, is similar to Quercus (H.McMinn 1944; M.A. Nobs 1963). A striking example is Mimulus in a broad sense. The annual species belonging to the complexes of M. guttatus and M. glabratus (figs. 11.5, 11.6) show a pattern similar to Vicia, but the shrubby species that constitute subgenus Diplacus resemble Quercus (V.Grant 1971; M.R. Macnair and P.Christie 1983). This difference is particularly important with relation to the explanation given above, that stronger barriers of reproductive isolation in short-lived plants are due to their occupation of pioneer habitats and a greater influence of chance events, while shrubs, often predominating in climax situations, are more strongly and directly affected by natural selection. Although all species of Mimulus are cross-pollinated by insects, annuals of the M. guttatus complex are often self-pollinated and can rapidly form new populations from a single founder individual, while the shrubs belonging to subgenus Diplacus are rarely selfed, and each flowering shrub is accompanied by several neighbors.

Botanists do not need to revert to a purely subjective species concept, nor do they need a concept that is based entirely on the presence versus absence of reproductive isolation. Therefore, this author agrees with opinions expressed recently by D.Levin (1979), L.D. Gottlieb (1986), B.D. Mishler (1985), B.D. Mishler and M.J. Donoghue (1982), P.H. Raven (1986), and others. The divergence between zoologists and botanists is justified by differences between animals and plants and the populations that they form. In animals, the difference between presence versus absence of gene exchange between populations is reinforced to such a degree by the combination of isolation via different courtship patterns, more complex individual development, strongly developed differences between the sexes and chromosomal heterozygosity in one sex, plus limited and often short life spans, that it provides almost universally effective barriers of genetic discontinuity, which can easily be made to coincide with species boundaries.

Conversely, in plants numerous conditions combine to eliminate in many genera the sharp contrast between presence versus absence of gene exchanges among populations that serves so well to delimit animal species. These conditions include the presence of both biparental and uniparental reproduction, which may alternate within the same pair of individuals depending on environmental conditions, random cross-pollination by wind or by unspecialized insects, simple developmental patterns that often allow the formation and establishment of hybrids in nature, great differences in the degree of fertility of F1 hybrids, and not infrequent vegetative persistence of a single individual during hundreds of years. Populations belonging to closely related plant species possess an entire spectrum of intermediate degrees of gene exchange that may vary greatly depending on different environmental conditions. Plant species are not bounded by any single limiting condition, as B.D. Mishler (1985) has pointed out. Nature forces botanists to adopt a pluralistic species concept.

The need for a pluralistic, but still biological, species concept has recently been recognized by the zoologist A.R. Templeton (1989). His cohesion species concept is similar to those discussed above. It can be applied more widely to different groups of organisms than can a biological concept that is based entirely on the presence versus absence of reproductive isolation and gene flow.

A broad definition of species that is in accord with these principles is as follows. Species are basic units of systematics and evolution. They consist of systems of populations that resemble each other in morphological, ecological, and genetical properties. The populations are held together by various cohesive forces, principally by gene flow but sometimes by partial autogamy, and normally by similarity due to common descent as well as by similar, complex adaptive syndromes that elicit parallel responses to environmental influences. Species boundaries may be sharply defined to produce complete genetical and physiological isolation from all other species, or they may be locally and temporarily weakened by partial breakdown of interspecific barriers.

Subspecies and Semispecies

Botanists must clearly pay attention to infraspecific categories. Moreover, these must include entities that are sympatric as well as those that are allopatric. As indicated earlier, sympatric population systems that are incompletely isolated from each other are of two kinds: (1)groups within which self-pollination predominates, as in the Vicia sativa complex, and (2) groups within which the formation of hybrid swarms depends on local ecological conditions, as in woody genera and in some long-lived perennial herbs. A model of a polytypic species within which allopatric subspecies can be recognized is Potentilla glandulosa, as classified by J.Clausen, D.D. Keck, and W.M. Hiesey (1940). Allopatric infraspecific categories are usually designated as subspecies. In local floras (see discussion of the Vicia sativa complex), some authors recognize as separate species, sympatric populations that in many regions keep distinct from each other but that elsewhere form localized hybrid swarms. Other authors designate them as "varieties." Examples like Quercus douglasii and Q. dumosa could be regarded as semispecies, or examples of arrested speciation, but for purposes of a flora they are assigned binomials, as are full species.

Recombinational Species

The example of Vicia sativa demonstrates the independence of segregation between morphological characteristics and the elements that compose barriers of reproductive isolation, chiefly chromosomal differences. Because of this independence, partially sterile interspecific hybrids can give rise in later generations to fully fertile, true breeding descendants that are partially isolated from both of their parental species. This course of events has been experimentally documented in Nicotiana (H.H. Smith and K.Daly 1959) and is responsible for the process designated as recombinational speciation (V.Grant 1981). Plant taxonomists who have used experimental methods have long recognized that at the level of population systems and species, phylogeny is reticulate. This should be kept in mind in cladistic treatments, especially with groups that include polyploid complexes.

The Relevance of Polyploidy and Apomixis to Taxonomy

One of the most significant factors that affects species patterns and speciation among higher plants is polyploidy. Its frequency has been variously estimated, depending on the criteria used to determine whether or not the ancestors of a particular species have undergone chromosome doubling at some time during their evolutionary history. A conservative estimate is that at least 30% to 40% of modern species are involved. Some of this doubling has accompanied the origin of the genus to which species belong. In other examples, entire subfamilies have originated via polyploidy, including Rosaceae subfamily Maloideae (basic number x = 17), Oleaceae subfamily Oleoideae (x = 23), Salicaceae (x = 19), and Magnoliaceae (x = 19). The origin of these groups has most probably involved reticulate evolution at the level of higher categories. Only recognizable polyploid series within genera are considered herein. The percentage of polyploidy among North American species varies over the entire spectrum from 0% to 98--100% (table 11.1).

TABLE 11.1. Approximate Percentages of Polypoid Species in the Larger, More Widespread Genera of the North American Flora1

Percent Polyploid Species Poaceae2 Asteraceae Ranunculaceae Rosaceae Scrophulariaceae Fabaceae Other
90-100 Agrostis
Elymus3 Iris
Poa Senecio Thalictrum Viola
70-89 Festuca Crataegus4 Saxifraga
Potentilla Silene5
50-69 Eragrostis Ranunculus Rosa Veronica Plantago
Muhlenbergia Polygonum
Paspalum Salix
30-49 Bromus Artemisia Anemone Rubus4 Castilleja Galium
Panicum Opuntia
Vaccinium
10-29 Aster Delphinium Penstemon Lupinus Allium
Erigeron Arabis
Haplopappus Lomatium
Helianthus Phacelia
Solidago Phlox
0-9 Melica Baccharis Clematis Pedicularis Lotus Asclepias
Brickellia Trifolium Berberis
Hypericum
Lilium
Lonicera
Populus
Quercus
Viburnum
1The Poaceae tribe Andropogoneae and the Genus Oenothera were omitted because of difficulty in coordinating published chromosome numbers with genera and species currently recognized.

2Family names are used only when two or more genera, as recognized in all manuals, are included in the table.

3The genus is recognized in a broad sense, including many species listed elsewhere under Agropyron, Sitanion, Hystrix, and various newly adopted names.

4The percentages assigned to the genera Crataegus and Rubus are somewhat arbitrary due to the difficulty of delimiting species among the apomicts of these genera.

5The high frequency of polyploids applies only to North America species. The great majority of Old World species, including some introductions into North America, are diploids.

All aspects of polyploidy have been carefully discussed in the volume edited by W.H. Lewis (1980). Additional information comes from studies of genetic diversity based on electrophoretic marker gene loci on allopolyploid Tragopogon (M.L. Roose and L.D. Gottlieb 1976), on genera of Saxifragaceae (D.E. Soltis and P.S. Soltis 1989), and on autopolyploid Dactylis (R.Lumaret 1988). For refinements, particularly with respect to ecology and distribution, see G.L. Stebbins (1984, 1985, 1986) and G.L. Stebbins and J.C. Dawe (1987).

The existence of polyploidy presents two important difficulties in classifying species and infraspecific taxa along traditional lines. First, it promotes reticulate evolution, creating situations in which systematists who would not know the chromosome numbers of the entities with which they are dealing would be likely to regard a common intermediate polyploid species as ancestral to a group. The actual diploid ancestors, which are less common and often more localized, would be considered as derivative. Second, polyploidy usually occurs in association with hybridization, genetic segregation, and natural selection. Hereditary differences within a polyploid complex involve not only differences in chromosome number, but also in structural rearrangements of segments within individual chromosomes and allelic differences at many gene loci. Some differences affect the diagnostic characters used to distinguish between species; others do not, although they may affect considerably the adaptation and distribution of the populations.

Consequently, the classification of populations in polyploid complexes deals with highly complex genetic and cytogenetic segregations, only a small part of which can be followed by the conventional methods of morphological taxonomy. In order to make these difficult problems accessible to those who are not cytogeneticists, some kind of simplification is necessary. Nevertheless, one cannot extrapolate from simplified treatments to determine actual relationships. With polyploid complexes, a case can be made for keeping the working taxonomy that is useful in floras separate from the analytical cytogenetic and biochemical taxonomy that is necessary for understanding evolutionary relationships.

The usual separation of polyploids into two categories, auto- and allopolyploids, must be regarded in the light of this complexity. The terms were originally defined on the basis of chromosomal differences (H.Kihara and T.Ono 1926). Allopolyploids contain two or more different genomes, defined as sets of chromosomes different enough from each other to prevent regular pairing at meiosis in an Fl hybrid. Kihara and Ono believed that the presence versus absence of such pairing is an absolute difference, usually correlated with species boundaries. Many cytologically intermediate situations were found later. Depending on the parental combination, different interspecific hybrids present a spectrum of conditions between perfect and highly irregular pairing. The degree of regularity can be increased or decreased by the action of different genes.

Furthermore, chromosome pairing in F1 hybrids is poorly correlated with species distinctness as determined by taxonomic methods. Consequently, a common taxonomic usage---to designate as autopolyploids those entities that fall within the range of morphological variation delimited as a species boundary by taxonomists, and as allopolyploids those which do not---bears little or no relationship to the categories as originally defined and named by Kihara and Ono. Polyploid complexes will be better understood if these terms are either discarded or assigned secondary importance, particularly by taxonomists who have not made careful cytogenetic investigations.

As D.E. Soltis and P.S. Soltis (1989) and R.Lumaret (1988) have shown by their electrophoretic-genetic investigations, all successful polyploids, even those that are strictly autopolyploid, are genetically more heterozygous than their diploid relatives. Relatively homozygous diploid genotypes, when subjected to artificial somatic doubling, produce tetraploids that, either immediately or in the long run, are adaptively less fit than their diploid progenitors (G.L. Stebbins 1985). Consequently, the evolutionary, genetic, and morphological structure of most, if not all, polyploid complexes is best expressed by the pillar model.

Among the complexes at least partially represented in North America, the genus Dactylis (Poaceae) and Epilobium subgenus Zauschneria (Onagraceae) are based on autopolyploidy (G.L. Stebbins 1971), Antennaria (Asteraceae) (R.J. Bayer 1985, figs. 11.7, 11.8; 1985b) is based on intermediate or segmental allopolyploidy, and the complex of Clarkia purpurea (Onagraceae) is based on genomic allopolyploidy (H.Lewis and M.E. Lewis 1955). Other polyploid complexes may be of a mixed nature.

Many polyploids have achieved great success, but that is by no means universal. Table 11.2 shows that exceptions exist to any rule that might propose a general advantage for polyploidy over diploidy, even when these conditions are present within the same genus. The same conclusion can be drawn from the more complete data compiled by G.L. Stebbins and J.C. Dawe (1987) from Flora Europaea (T.G. Tutin et al. 1964--1980).

Polyploidy in itself does not increase the resistance of plants to cold, drought, and other forms of stress (G.L. Stebbins 1980, 1984, 1985, 1986). The most cogent statement on this point is that the highest levels of polyploidy among cultivated plants are the sugar canes (Saccharum) native to the moist tropical lowlands of New Guinea; among woody species in genus Clerodendrum, native to the forests of tropical Africa; and among ferns in species of Ophioglossum, native to southern India. The correlation between percentage of polyploidy and latitude in European floras on which the hypothesis was first based does not hold for Pacific North America (G.L. Stebbins 1984, 1986; J.G. Packer 1969). Furthermore, better knowledge of the principal genera involved, such as Calamagrostis, Draba, Poa, Potentilla, and Saxifraga, has revealed the presence in each of these genera of diploid species or cytotypes at high latitudes, or at high elevations in mountain areas. The arid regions of North America are notable for their low percentages of polyploids.

Lists have been compiled of species having very high chromosome numbers (2n = 96 or higher) in each of the floristic provinces of North America as defined by A.Cronquist (1982). Interpretation of these results is somewhat complex and will be presented in a subsequent publication. Briefly, the numbers of such species do not differ significantly among the provinces that have the greatest area and the maximum recent history of advances and retreats of floras, i.e., Arctic, Canadian, and Appalachian. They provide support for the secondary contact hypothesis: successful polyploids are usually formed after reunion or secondary contact between previously separated and ecologically divergent races, subspecies, or species.

In a few genera, polyploidy is associated with seed production by asexual means, or apomixis. The principal features of this process, including the various ways by which the events of meiosis and fertilization can be circumvented, have been well known for many years (G.L. Stebbins 1950). Recent research has done no more than refine our knowledge and provide additional examples. In the North American flora, variation patterns dominated by apomixis are best known in Antennaria, Crepis sect. Psilochaenia, Crataegus, some species of Erigeron, Poa, Potentilla, Rubus, and Taraxacum (both introduced weeds and native species of the western Cordillera and arctic regions). In the genus Hieracium, the largest and most complex series of apomicts known, several of the species introduced from Europe are apomictic, but so far as known, the endemic North American species of the subgenus Stenotheca are all diploid and sexual.

An early model for the treatment of agamic complexes (E.B. Babcock and G.L. Stebbins 1938) has been widely accepted and, with revision based on new knowledge, is still valid. It has been recently applied to the two groups of Antennaria species found in eastern North America. With the aid of experimental taxonomy (interspecific hybridization and detailed analysis of F1 hybrids), biosystematic techniques (quantitative determination of the extent of morphological variation via multivariate analysis and cytological studies of meiosis in species and hybrids), and electrophoretic investigations of enzyme variation patterns, the model has yielded a reasonably complete picture of species relationships (R.J. Bayer 1985, 1985b). The absence of gene exchange between apomicts renders impossible decisions about the delimitation of species and subspecies based on presence versus absence of gene flow, so that species boundaries may become more subjective.

Guidelines can be obtained from studying species limits in related groups that reproduce sexually. Such analogies originally suggested that if the morphological pattern of a group of apomicts indicated that they were derived from hybridization between two sexual species that are similar morphologically, most or all of the apomictic clones could be admitted within the limits of one or the other of the parental species. If, on the other hand, the putative parental species were very different from each other, or if a series of similar apomicts appeared to have been derived from hybridization involving three or more different diploid species, they would be best understood by grouping them into highly variable, collective "agamospecies" bearing a different binomial from any of the sexual ancestors. This alternative treatment was adopted in both Crepis and Antennaria.

Agamic complexes can be treated taxonomically in two other ways. One method, adopted extensively in Europe for Taraxacum, Hieracium, Rubus, and other genera, is to recognize each apomictic line as a separate species. Aside from the confusion resulting from the enormous number of so-called "species" generated by this treatment, it has one basic defect with respect to both biosystematics and taxonomic philosophy. In all the genera mentioned above, clusters of obligate apomicts exist sympatrically, at least occasionally, with highly heterozygous and variable populations belonging to a related, fully sexual species. If two such clusters or populations are compared with respect to any group of characters, morphological, cytological, or biochemical, each individual of the sexual population will correspond to a large degree with a clone belonging to the apomictic cluster.

Apomictic clones can be recognized as species only when they occur beyond the range of the related sexual species. The confusion produced by giving the same taxonomic rank, in the same section of the same genus, on the one hand to a single genotype, and on the other to an entire population system, produces a confusing taxonomic system.

An alternative treatment might be to split up the sexual species to make each of their "species" units similar to the apomicts, but this would cause even more confusion. At the other extreme, one might argue for recognizing as a single taxonomic species an entire agamic complex. In addition to producing taxonomic confusion of a different sort, such a practice would violate one of the basic principles of modern taxonomy. Because the superstructure of apomictic clones that form the bulk of the complex are "crystallized hybrids" that contain genes derived from several different sexual parental species, some of which may have been isolated from each other for millions of years, the entire complex is, by its very nature, highly polyphyletic. To treat such a complex as a single basic unit, one species, violates not only current taxonomic practices but also basic biological philosophy.

When dealing with apomictic complexes, the taxonomist faces a situation that nature has evolved, fortunately in only relatively few instances, in defiance of the rules that taxonomists would like to follow. Anybody who is dealing with these situations must recognize this fact and solve the problem in a manner that will be most useful to fellow scientists.

New Ideas on the Treatment of Genera

The delimitation of genera has always raised greater problems than that of species. Some evolutionary taxonomists, particularly zoologists, have maintained that species are objective entities that the taxonomist must learn to recognize on the basis of boundaries that become clearly defined when enough facts are known. Regarded by many taxonomists as erroneous even for species, this statement has never been maintained with respect to genera, except by a few taxonomists who regard certain attributes as of overriding importance.

Nevertheless, to many biologists who are not taxonomists, genera are regarded as more important than species. No species of higher plant or animal is as familiar to experimental biologists as are generic names such as Drosophila, Mus, Nicotiana, Lycopersicon, Triticum, Saccharomyces, or Neurospora. Consequently, for general biological purposes such as ecology, physiology, morphology, and evolution, exchange of information is facilitated if genera are delimited to be as nearly equivalent to each other as possible, and to retain their scientific names as long as possible. These considerations should be balanced against the need to change names of genera when new information is obtained about their phylogeny.

In contemporary angiosperm taxonomy, two trends have developed that must be carefully watched. One is the splitting of genera on the basis of microscopic details, as has been done in the Asteraceae tribe Eupatorieae (e.g., R.M. King and H.Robinson 1970, 1987). While microscopic characters should by no means be overlooked, their use to establish new genera that are not distinct on the basis of any other characteristics should be discouraged.

The other trend is to base genera on genomic composition (M.E. Barkworth and D.R. Dewey 1985): taxa with different genomic constitutions are placed in different genera. A familiar example is found within genus Nicotiana. In this example, S and T are genomes, each containing a gametic set of 12 chromosomes that differ from each other so much that they are unable to pair in the diploid hybrid (ST). If the rule stated above is adopted, cultivated tobacco, N. tabacum (SSTT), would be placed in a genus by itself, separated from its diploid ancestors, N. sylvestris (SS) and the N. tomentosiformis group (TT).

Such a proposal has three serious difficulties. First, the decision as to whether two genomes are similar or different is hard to make in many groups because intermediate conditions, partial differences, exist. For instance, the three genomes contained in bread wheat were once regarded as different from each other, so that under the Barkworth-Dewey proposal, the three wheats, aestivum or hexaploid, dicoccum or tetraploid, and einkorn or diploid, would have had to be placed in different genera. R.Riley and V.Chapman (1958), however, have identified genes that prevent partly homologous chromosomes from pairing with each other, so that the wheat genomes can now be regarded as similar. The extent to which genes having this effect exist in other plant groups, including the Triticeae, is unknown.

Second, the amount of labor necessary for the cytogenetic delimitation of genomes is great. If the proposal is adopted, different standards of generic limits would exist for those groups that have been investigated for this purpose and those that have not. For instance, if the genomic composition of a neighboring genus, Bromus, were fully known, and if Bromus were treated according to standards proposed for subtribe Triticinae, it would be split into at least nine genera (G.L. Stebbins 1981). A similar fate would befall Festuca, Poa, Panicum, Agrostis, and other large grass genera should they be analyzed in the same manner.

Third, if the proposal were adopted, then every major cytogenetic investigation of a genus of angiosperms would result in several new and unfamiliar generic names. Beginning students of taxonomy are told that they must adopt scientific names because they have greater stability than common names. If they were then told that names should be changed whenever a comprehensive piece of cytogenetic research was completed, they would be justified in being somewhat skeptical of the first argument.

Nevertheless, a more moderate use of cytogenetic or other phylogenetic information is possible. For this purpose, categories of intermediate rank are recognized. Both the subgenus and the section are commonly used, and could be applied to genomic differences without upsetting the apple cart.