and Wright 1941), later studies demonstrated correlations with altitude, season, and climate (Dobzhansky 1943, 1947, 1948a, 1948b). Some viability differences among individual organisms bearing different inversions may relate to the effects of temperature and crowding (Birch 1955). In some cases, strong regional differences in seasonal variation are related to climatic differences (Crumpacker and Williams 1974). In the cactus-loving species, D. pachea, a latitudinal cline of inversion frequency is strongly correlated with climate and a change of host plant (Ward, Starmer, Russell, Heed 1974). No evidence proves conclusively that these differences result from characters of the chromosomes per se, as opposed to alleles for specific genes or groups of genes, carried by coincidence.

Dobzhansky (1947) argued that chromosomal polymorphisms were the result of heterozygote (heterokaryotype) superiority. This was based on excesses of heterozygotes relative to Hardy-Weinberg expectations and the convergence to intermediate frequencies of two-inversion populations in laboratory cages. Frequency-dependent selection may also explain this convergence. Studies of overall performance tend to show that heterozygotes are superior, or at least equal in performance, to homozygotes. For example, Moos (1955) showed that homozygotes for the Chiricahua (CH) inversion were inferior in general physiological performance to Standard (ST) homozygotes, which were subequal in performance with CH/ST heterozygotes.

Heterozygote superiority seems to occur only when the inversions come from the same locality (Dobzhansky 1948b). This may indicate that superiority is conferred by favored gene combinations and not by any innate superiority of chromosomal heterozygosity.

Cases of chromosomal polytypism are found in a wide variety of species with limited dispersal ability and usually small population size, such as in small (particularly subterranean) mammals, some Diptera, and in flightless grasshoppers (Bush 1975; Key 1974; Nevo 1982; White 1973). Hybrid zones between chromosomal races, when present, are usually extremely narrow (e.g., Key 1974; Nevo 1982). This may testify to reduced gene flow between adjacent populations due to poor viability of chromosomal heterozygotes.

The distinction between subspecies and species status is quite difficult in situations of chromosomal polytypism. In the Central American Peter's tent-making bat, Uroderma bilobatum, two cytotypically characterized populations overlap in a small embayment in Honduras (Baker 1981). The two cytotypes differ by one terminal translocation and two fusions (2N = 44, 38). Gene flow is restricted, although the hybrid zone is claimed to be quite wide. Backcross cytotypes between the two populations occur over a band of 400 kilometers. Nevertheless, of 11 known polymorphic allozyme loci, 9 have markers unique to each of the populations (Greenbaum 1981). This suggests very restricted gene flow between the two populations, despite the claimed widespread "leakage" of chromosomes across the barrier. Barton (1982) suggested that the pattern of variation seen in the differentiation zone is consistent with a hybrid zone being maintained by hybrid unfitness. The apparent great width of the zone can be explained by the obviously large dispersal distance of bats.

Chromosomally distinct races (species?) are especially common in the house mouse, Mus musculus. The differentiation probably occurred in the past few thou sand years and was strongly influenced by migrations of M. musculus with its commensal humans. But high speciation rates are also associated with many other small rodents, including Microtus and Peromyscus (Martin 1993). The unifying feature must be small effective population size, enforced by low dispersal, which is associated with small body size. Hybrid zones are narrow and hybrids often show strongly reduced fertility or malformations. The typical Mus karyotype in most localities consists of 2N = 40 acrocentrics. An isolated population in eastern Switzerland, however, is fixed for 2N = 26, suggesting the fixation of 7 fusions. This sort of isolation is common in Europe (see Capanna 1982). Populations with various degrees of fixed Robertsonian fusions occur in the Rhaetian Alps, the Apennines, and in Sicily. Fixations of different chromosome numbers divide populations according to region. In the Rhaetian populations, four metacentrics characterize all subpopulations. Other fusion types are found in successively more restricted subpopulations, until a given unit population is characterized by unique metacentrics. This suggests a process of isolation of a primordial population, followed by further substructuring into unique groups. Chromosomally distinct regions occur as islands in a sea of all acrocentric mice.

Populations of the mole rat Spalax ehrenbergi in Israel are also strongly polytypic, or show recent speciation (Nevo and Shaw 1972). Robertsonian fusions result in four major cytotypes (2N = 52, 54, 58, 60), which come into contact along narrow (2.8- to 0.7-kilometer-wide) zones. Paleobotanical evidence suggests that migration and subsequent differentiation in Israel occurred between 250,000 and 10,000 years ago. The four cytotypes are distributed along a north-south aridity gradient and are morphologically similar except for an inverse relationship between body size and environmental temperature. The subterranean habits and small population size of the mole rat are very conducive to such differentiation. Although pre-mating isolation mechanisms usually exist between the cytotypes, one case still shows only postmating isolation, which suggests that postmating incompatibility preceded the evolution of premating recognition mechanisms (Nevo 1982).

Fixation of chromosomal variants in populations. In a randomly mating deme, the rate of fixation of rearrangements with large heterozygous disadvantage is minuscule unless effective population size is very small (Lande 1979a). Given known rates of fixation of chromosomal variants and spontaneous rearrangement rates, Lande estimated that long-term effective deme sizes must be on the order of 30 to 800 individuals for a wide variety of mammals, lower vertebrates, and insects. This suggests the ubiquitous occurrence of genetic drift in animal populations. The spread of chromosomal variants has probably occurred usually by random local extinction and colonization. Therefore, most locally fixed variants disappear when their population disappears via random extinction. Owing to their relatively minimal effect on heterozygotes, inversions and Robertsonian fusions seem to predominate over reciprocal translocations.

Rapid speciation of mammals, which seems to have occurred over time spans of thousands of years, is inconsistent with the average fixation rate of new chromosomal rearrangements, which is maximally on the order of one per lineage per million years (Bush, Case, Wilson, and Patton 1977). It may be that most chromosomally differentiated populations are geographically restricted and thus have a high probability of extinction. The more extensive populations, such as the all-acrocentric "standard" populations of Mus musculus, may survive owing to their abundance. In the random case, the probability of spread of a given variant over the entire species should be 1/N, where N is the number of demes. This probability, multiplied by the probability of fixation of a variant within a deme, would give the probability of spread through the entire species. To take the house mouse as an example, we can imagine that the number of demes must be sufficiently high that the probability of survival of the all-acrocentric karyotype is assured. New populations with novel genotypes will, for the most part, go extinct.

Chromosomal incompatibility may not prove to be the primary mechanism of isolation. Genic mechanisms of postmating isolation may be important even in chro-mosomally polytypic populations. In the rampantly speciating Hawaiian drosophilids, many groups of species are chromosomally monomorphic. On the other hand, extensive regional differentiation in chromosomal variants can occur with minimal reproductive isolation, as in the pocket gopher Thomomys bottae (Patton 1972). The phenomenon of hybrid dysgenesis in Drosophila is now known to result from genic disruption by transposable elements. Crosses between different strains result in accelerated mutation rates and chromosomal disruptions (Kidwell, Kidwell, and Sved 1977). Although the phenomenon is unrecognized outside of Drosophila, otherwise unknown alleles may be found in hybrid zones in mammals (Hafner, Hafner, Patton, and Smith 1983) and in other groups (Sage and Selander 1979; Whitt, Childers, and Cho 1973; Woodruff and Gould 1980; Woodruff and Thompson 1980). This may also reflect intragenic recombination in hybrid zones that may generate novel alleles.

A strong case can be made for gene-based sterility as a mechanism of postzygotic incompatibility. In Drosophila melanogaster, fixation of new genes in a population has a considerable probability of ensuring reduction of viability of offspring produced from crosses with other populations. At the X chromosome, fixations at about 9% of the genes would result in major sterility problems in females (Gans, Audit, and Masson 1975). In considering genes that affect more subtle aspects of mating behavior and reproduction, the potential for incompatibility is greater. Male hybrid sterility is explained similarly by several to many genes, particularly at the X chromosome (Coyne 1984). It may be, however, that the autosomes harbor as large a proportion of sterility factors as the X chromosome (Hollocher and Wu 1996). Genetic difference per se does not, however, guarantee postmating incompatibility. Cases of extensive chromosomal differentiation are known where reproductive compatibility is high. In the goodeid fish Ilyodon furcidens, extensive variation in the number of metacentric chromosomes occurs within a single river basin, despite minimal allozyme divergence and full viability of laboratory crosses and backcrosses (Turner, Grudzien, Adkisson, and Worrell 1985).

The extrapolationist hypothesis is consistent with the chromosomal differences observed among species. The mechanisms of geographic differentiation are easily related to speciation mechanisms involving the establishment of postmating isolation. Variants can be traced across subsequently differentiated populations and species

(Olvera et al. 1979; White 1973). There is no intraspecific-interspecific dichotomy. It is not clear, however, that chromosomal incompatibility is a major genetic mechanism of speciation. Chromosome differentiation does, however, bear the signature of genetic drift.

Comparisons with morphological and allozyme divergence. Extensive chromosomal race formation can occur with little concomitant allozymic differentiation. In the peripheral relict pocket gopher species Geomys tropicalis, major changes in chromosome number are not accompanied by an unusual degree of allelic substitution (Selander, Kaufman, Baker, and Williams 1974). This seems to be common in cases of extensive chromosomal differentiation (e.g., Greenbaum 1981; Nevo 1982). An interesting exception can be found in Rocky Mountain populations of Geomys, where extensive among-population chromosomal differentiation is accompanied by strong allozymic differentiation (Penney and Zimmerman 1976). Such major differences within one genus suggest a degree of unpredictability of an allozyme-chromo-some correlation, but the latter case argues for local drift.

The rate of chromosomal variant fixation is inversely proportional to body size (Bengtsson 1980). This may relate to the longer generation time or to greater vagility of larger mammals. With relatively infrequent reproduction and few young per brood, a chromosomal abnormality would cause a significant loss of offspring. In fecund animals, loss of a few young might be matched by increased health of survivors or accelerated production of a successive brood. Thus, change might be accommodated more easily in small-bodied species. Gene flow among larger and possibly geographically wide-ranging mammals is probably not an explanation for reduced divergence, because behavioral deme structuring is common among larger-bodied species (Bengtsson 1980). Although deme structuring does not guarantee reduced gene flow, it can permit such a restriction.

Relation to morphological evolution. Chromosomal polytypism can be correlated with geographically related reproductive isolation. The chromosomal differentiation itself probably represents random fixation in relatively small populations with low vagility. The accumulation of such fixations in isolated populations may contribute to reproductive isolation. Despite the widespread occurrence of chromosomal races, the evidence does not support any extensive concomitant morphological differentiation. For example, the three classic morphologically recognizable species of the mole rat genus Spalax, ranging from Russia to North Africa, represent at least 30 karyotypes, most of which seem to be distinct species (Nevo 1982). Although the chro-mosomally distinct eastern Switzerland population of Mus musculus was once recognized as a different species on traditional grounds as M. poschiavanus (see Capanna 1982 and references), numerous other isolated races are morphologically indistinguishable except by karyotype. Great karyotypic disparity among species with few morphological differences can be observed in some rodents, foxes, insectivores, horses, and gibbons (see cited literature in Marks 1983).

I should emphasize that morphological correlations can be found with karyotypic differences. As an example, body size in grasshoppers seems related to the presence or absence of given inversions (White et al. 1963). This could be due to the presence of a few contributing genes, however. Most rearrangements seem unrelated to mor phological differentiation. In Drosophila, rearrangements found in natural populations do not show any relationship to characters of taxonomic significance (Spieth and Heed 1972).

Chromosomal evolution has been claimed to be a cause of morphological evolution (Bush et al. 1977; Wilson, Sarich, and Maxson 1974; Wilson, Carlson, and White 1977). Under this hypothesis, chromosomal variants are regarded as having gene regulatory and morphological significance. If speciation is a cause of, or a concomitant process with, chromosomal evolution, then we would expect a correlation among speciation rate, karyotypic diversity, and morphological evolution. Rate of chromosomal evolution is assumed to be related to a measure of karyotypic diversity among extant species. Following the method of Stanley (1979), Bush et al. estimated speciation rate by taking the number of extant species and the time of origin in the fossil record for the group and calculating a splitting rate assuming constant dichotomous splitting. There was a positive correlation between the rates of chromosomal evolution and of speciation in a study of extant species of various reptilian groups and orders of mammals. From the correlation, they inferred that karyotypic evolution is a source of morphological evolution.

Although this is possible, a casual inspection reveals inconsistencies. Horses have the highest speciation rate and corresponding rate of chromosomal evolution. But the living forms whose karyotypic differences are extensive constitute a rather morphologically homogeneous group of mammals. A consideration of the fossil record of closely related horses does not increase the morphological diversity very much. Although correlated changes in body size, relative length of limbs, and hypsodonty characterize the grazing equine genera, morphological similarity is very strong, to the degree that minor reinterpretations of features in fossils have caused the systematic position of various groups to change radically (Woodburne and MacFadden 1982).

In contrast, the morphologically diverse Cetacea are lowest among the mammals in karyotypic diversity and speciation rate. The Sei whale, Balaenoptera borealis, and the common dolphin, Delphinus delphis, have nearly identical karyotypes (2N = 44), yet they must have diverged 40 million to 50 million years ago (Arnason 1972). Although the fossil record is too sparse to make an estimate of speciation rate, the Pinnipedia are similarly chromosomally homogeneous (Arnason 1972). As Bengtsson (1980, p. 38) noted, "...a relationship between karyotype evolution and the evolution of regulatory genes is, at most, of highly indirect and weak nature." The correlation observed by Bush et al. probably relates to the expected population genetic processes at reduced population size that occur during the speciation process. Karyotypic divergence is thus probably an effect of speciation, or even an occasional cause of reproductive isolation. It is not likely to be a major cause of morphological evolution.

Correlations between morphological and karyotypic evolution may occur, but only coincident with the speciation process. As an example, cladistic analyses of chromosomal banding patterns from 48 species of cryptodiran turtles, combined with fossil-based methods for estimating rates of karyotypic change, demonstrate that karyotypic evolution was twice as fast in turtle groups arising in the Mesozoic as in more recent splits and involved different forms of rearrangements. The decel eration in rate of change is correlated with decelerated morphological change. Some chromosomes have remained unchanged for at least 200 million years (Bickham 1981). Although chromosomal changes might be involved in adaptive changes, it is likely that the initial rapid radiation of turtles was accompanied by divergent morphological evolution, which must have involved speciation among geographically separated populations. In other words, speciation could have been an effect of divergent adaptation; the tempo of karyotypic evolution would probably have tracked speciation. Chromosomal change, therefore, was likely not the cause but was more likely the effect of evolutionary radiation and speciation.

Cherry, Case, Kunkel, Wykles, and Wilson (1982) used a metric, D, to estimate proportional differences in homologous skeletal measurements and found no substantial differences among species within genera of frogs, lizards, and mammals. Generic longevity of mammals is substantially less than for the others, and this might suggest that speciation rate accelerates mammalian morphological evolution. Alternatively, phyletic morphological evolution might be greater for mammals. Using Van Valen's (1973a) compilation, we can calculate the ratio, R, of D within a genus to the number of species per genus. If one assumes that the average number of extant species in a genus is proportional to the number of speciation events required to generate the species richness, then the divergence-to-species richness ratio gives a rough estimate of the relative amount of change realized per speciation event. One gets the following: mammals: R = 1.90; lizards: R = 1.08; amphibia: R = 0.76. A given speciation event or anagenic change during a species' history in mammals may therefore entail more morphological change than in reptiles or amphibia. The relationship between the rates of morphological divergence and of speciation may therefore be coincidental, or morphological evolution might even accelerate speciation.

This would solve a paradox well known to evolutionary biologists: The greatest amount of divergent evolution of morphology occurs near the beginning of the fossil record of a group. But this cannot be a time of maximum absolute number of speciation events, if any sort of exponential model of species increase applies. Thus, it is not the sheer number of speciation events but a qualitative difference in rate of morphological change that increases the degree of divergence per speciation event more toward the beginning of the history of a radiation. Usually, this seems correlated with the prior elimination of a competing group by a mass extinction (see chapter 7). It is therefore doubtful from this perspective that speciation per se accelerates morphological evolution.

Allozymes and interspecies divergence. Allozyme polymorphisms are ubiquitous in natural populations, although different groups may have characteristically different levels of variability (Avise 1994). Allozyme polytypism is also common among many species (e.g., Christiansen and Frydenberg 1974; Koehn, Milkman, and Mitton 1976; Schopf and Gooch 1971), though geographic homogeneity in allele frequency is also common (Ayala, Powell, and Tracey 1972; Prakash et al. 1969). Although the strength of selection at any locus is difficult to calculate, it is clear that natural selection plays a major role in the maintenance of variability, by means of fitness differences depending upon metabolic efficiency (Eanes 1999). A large number of studies, mainly in Drosophila species, permit estimates of the degree of intraspecific and interspecific differentiation.

Nei's (1972) index of genetic distance is commonly used to estimate allelic divergence at allozyme loci. If Ix is the average sum of the squares of the allelic frequencies over all loci for species x, Iy is the corresponding sum for species y, and Ixy is the average over all loci of the sum of the cross products of allelic frequencies for a locus, then distance D is

Ix and Iy measure the average probabilities of identity over all loci of two randomly chosen homologous genes from species x or y, whereas Ixy is a measure of the average probability of identity of two randomly chosen homologous genes from the two species.

The data on the willistoni group of Drosophila (Ayala, Tracey, Hedgecock, and Richmond 1974) suggest a smooth transitional increase in D from geographic populations to morphologically different species. This seems to hold generally for animals (Nei 1975): Average D = 0.00-0.06 between races, 0.00-0.20 between subspecies, 0.1-1.5 between sibling species, and 0.1-2.5 between nonsibling species. The species barrier does not seem to be a special level of rectangular divergence in genic identity. If these degrees of differentiation correspond to stages in speciation, divergence seems to continue smoothly after speciation has progressed from the sibling species stage to a later stage of morphological divergence.

The smooth transition within a group of Drosophila might suggest an overall correlation between speciation rate and allozyme divergence among related groups. A correlation of genetic distance with the number of speciation events is compatible with punctuated equilibrium, but so is the greater accumulation of phyletic evolution in species that are undergoing speciation. This is especially relevant to ecologically driven speciation, where adaptation to environments leads to separation and eventually establishment of crossing barriers between newly established daughter species. Thus, although correlations between speciation rate and genetic divergence have been established (Mindell, Sites, and Graur 1990), it is only lack of correlation that proves anything, and such a lack of correlation falsifies a prediction of punctuated equilibrium. For example, the North American minnow family Centrarchidae is depauperate in species, whereas rapid speciation has been the rule for the sunfish family Cyprinidae (Avise 1977; Avise and Ayala 1976). In a study of 24 gene loci, average D = 0.63 for centrarchids, whereas average D = 0.65 for the cyprinids. This suggests a lack of relationship between speciation rate and divergence rate. The neutral theory would predict similar divergence among species if timescales since divergence were similar. Douglas and Avise (1982) extended the work to morphology and demonstrated that divergence among species is about the same between the rapidly speciating minnows and the more slowly speciating sunfish. Smith (1981) criticized this work, as the fossil record suggests a higher speciation rate for centrarchids than previously believed. Mayden (1986) thoughtfully analyzed the conclusions of Avise and colleagues and, although not disagreeing with their conclusion, pointed out many systematic difficulties, among which is the probable lack of monophyly of the minnows employed in the analysis. Avise (1994, p. 269) replied that if the minnows do not constitute a monophyletic clade, then they should be even older than now thought and should be even more genetically and morphologically distant among species.

There is no necessary relationship between morphological divergence and allozymic divergence. It is true that a good correlation exists in many species of mammals and fishes (Avise 1976). But many exceptions suggest that this may be due to rather constant correlated rates of morphological and allozymic divergence with time, with no causal relationship between the two sets of traits. In the desert pupfish Cyprinodon macularius, significant among-river morphological differentiation is not accompanied by allozymic differentiation much greater than is usually found in intraspecific comparisons of other teleosts (B. J. Turner 1983). This suggests that morphological differentiation can be rapid and independent of an allozymic scale. A similar discordance exists between patterns of color and banding and allozymic differentiation in Pyrenees and Welsh populations of the land snail Cepaea nemoralis (Jones, Selander, and Schnell 1980).

Morphology. It is difficult to summarize adequately the evidence for intraspecific versus interspecific divergence in morphology. In the case of chromosomes and allozymes, one has at least the confident feeling that identifiable markers can be traced across intraspecific and interspecific barriers. In the case of morphology, different parts of the genome can exert significant control on a given trait. There is also no uniform criterion by which one can draw equivalence between any two external morphological traits. If intraspecific variation of color morphs in butterflies can be extrapolated to interspecific comparisons, what relationship does this have to wing shape or to time of pupation?

Consider a character that has two different states in two different species. Is there a leap in character state that can be explained only by a speciation event, or can intrapopulation polymorphism be used in a simple extrapolationist model to explain polytypism and interspecies differences? Two possible approaches can be taken. First, if hybrids and F2 generations can be obtained between the two species, quantitative genetics can be employed to learn whether the difference between alternative character states is saltatory and based on fixation of alternative alleles at one locus or whether it is polygenically controlled with extrapolation possible from within-population variation. (Remember that discrete morphs can also be polygenically controlled.) Even without a genetic approach, much can be learned from a comparative biometrical study of morphological variation at the intraspecific versus the interspecific level. But if intraspecific morphological variance is much smaller than interspecific morphological differences, one can always argue that strong directional selection occurs during the speciation event. On the other hand, if the levels of variance are about the same, one can argue that yet other characters are saltatory and one has not come across the "species-specific" characters.

Crosses between species and populations have been done extensively, particularly in plants in which interspecific developmental incompatibilities are smaller than in animals. The minimum number of genetic factors controlling the trait is estimated by comparing the phenotypic means and variances in the two parental populations and the F1 and F2 hybrids and backcrosses. Polygenic control and the consequent possibility of extrapolation are often demonstrated by the intermediate phenotypic scores in hybrids and the expansion of phenotypic variability in the F2. In cases in which intermediacy in the F1 is not found, threshold effects and polygenic inheritance usually turn out to be the rule (e.g., Green 1962; Wright 1934a, 1934b, 1935a, 1935b). In most cases, the minimum number of genes for morphological traits is typically estimated as 5 to 10, with occasional values up to 20 (Lande 1983). Ten independent genetic factors were estimated to be operative in an analysis of tomato strains where two varieties differed about 56-fold in mean weight. As the haploid number of chromosomes is 12, the actual number of factors is probably larger, with some chromosome-level linkage.

Though we cannot make any universal statements, some interesting cases of interspecific variation demonstrate that the extrapolationist hypothesis is supportable for most transspecific evolutionary changes. A remarkable case of extreme morphological differentiation has been discovered between two species of Hawaiian Drosophila, D. heteroneura and D. sylvestris (Templeton 1977; Val 1977). The pair are very similar by allozyme and cytogenetic standards but are strikingly different in head shape (Figure 3.8). A genetic analysis shows that at least six to eight independent genetic factors control the phenotypic difference. The effects of the factors are predominantly additive, on which is overlain a sexual dimorphism that is most likely connected by a sex-linked locus or loci whose expression is limited to males. The interspecific phenotypic difference may be quite important in premating isolation, but it could have easily evolved from intraspecific variability.

For the sibling species of Drosophila, genital morphology is the sine qua non of species-specific morphological characters. They often are the only means of diagnosis and likely constitute a principal mechanism of premating isolation. Differences between species are discrete; otherwise they would not be good species characters! Coyne (1983) analyzed genitalia differences among the siblings D. melanogaster, D. simulans, and D. mauritania, by substitution of different chromosomes in hybrids. Variation in genitalia is under the control of at least four to five genetic factors.

heteroneura si/vestris

Figure 3.8. Discrete differences in head shape, despite polygenic trait control, between two closely related species of Hawaiian Drosophila. (After Val 1977.)

heteroneura si/vestris

There is no need to invoke any unique process in the morphological differentiation accompanying speciation. A QTL mapping of the form of the posterior lobe of the male genital arch in Drosophila simulans and D. mauritaina shows the action of at least 19 loci (Zeng, Liu, Stam, Kao, Mercer and Laurie 2000).

Though species are morphologically distinct, one can find extensive regional differentiation, often equal in magnitude to interspecific differences. This can be shown, for example, in some species of the land snail genus Cerion on Caribbean islands (Gould 1969a; Woodruff and Gould 1980). Over distances of 100 meters, large changes in sculpture, size, and whorl number per unit size occur. Discrete, often major, variation is found commonly within species of marine snails, such as the genus Thais (Palmer 1985). In the three-spined stickleback Gasterosteus aculeata, intraspecific differentiation is pronounced and of the same order as interspecific differences (e.g., Bell 1976, 1981). Within a lake in British Columbia, two probable species coexist that reflect extensions of intraspecific differences (Larson 1976; McPhail 1984). The benthic form has a heavier body, wider mouth, reduced dorsal spines, and reduced lateral plates, relative to a limnetically specialized form. It is not clear that speciation occurred within this particular lake, but the body size, spine and plate polymorphisms are well known within other populations.

One example is of particular interest as it falls within the home territory of the macromutationist-speciation school. Mimicry in butterflies has been discussed above and shown to represent a polygenic system that evolved by accumulation of several new genes of varying relative effect on the phenotype. Can intraspecific variation be extrapolated to interspecific differences? Remember, this case was one of Goldschmidt's (1945a, 1945b) prime examples of the uniqueness of saltatory jumps.

Mimicry probably has the longest pedigree of any work integrating variation in natural populations with speciation. In 1862, Henry Walter Bates published a theory of mimetic resemblance among butterflies, stemming from his observations of intraspecific and interspecific variation in the color patterns of South American butterflies. Using the fabulous diversity of form found in Brazilian faunas, he was able to demonstrate a continuity between geographic varietal variation within a species and the common occurrence of small-ranging groups of species whose ranges were contiguous. To Bateson, this indicated that polytypism preceded speciation.

Turner (1981 and cited references) has investigated patterns of mimicry in the genus Heliconius, where Mullerian mimicry (model and mimic are poisonous) is the rule in both larvae and adults. The butterflies feed on passion flowers (Passifloraceae), which live in shaded forests. The genes involved in mimicry consist of a combination of genes of large and small effect. A large gene bridges a gap that permits the further evolution of stronger resemblance. Racial divergence within species is strong and is easily extrapolated to interspecific differences. The species pair H. melpomene and H. erato co-occur in a range of localities, each with its characteristic and quite different mutually mimetic color pattern. Laboratory crosses demonstrate complete interfertil-ity among populations of the same species, taken from different locales. Nonmimetic relatives of both species have similar yellow and black patterns.

Brande (1979) investigated intraspecific versus interspecific variation within the genus Mulinia (Mactracea). Mulinia lateralis, for example, has a broad geographic extent from New Brunswick to Yucatán. It has given rise to one daughter species in Lake Pontchartrain, Louisiana - M. pontchartrainensis - and several related species also occur in the western hemisphere. Using discriminant function analysis, Brande found that the characters contributing to most of the among-locality variance within a species were also those important in among-species variation. This suggests that the features of the shell involved in intraspecific evolution are also those involved in the evolution of interspecies differences. Because shell characteristics are those expected to be crucial in bivalve adaptation (Stanley 1970), we can conclude that the speciation process is not particularly important here as a threshold in bivalve evolution. Similar results were obtained in an examination of the Miocene scallop Chesapecten (Kelley 1983a). Kelley (1983b) found that in some cases, characters most important in describing the variance within species were not those diagnosing differences between species. This proves little, because the "species" consisted of an ancestor-descendant series with no cladogenesis. How does one tell species apart, in this case, except by morphological change? Even in cases in which true species are examined, finding such a discordance between intraspecific and interspecific variance could also indicate that times of unique ecological change induce changes in characters of otherwise low variation. Unfortunately, Brande's test applies only to confirming the continuity of intraspecific to interspecific variation. A lack of continuity yields an ambiguous result.

Brande's results follow those of other studies. Clarke and Murray (1969) studied variation in Partula, a genus of terrestrial snails found in the Society Islands of the Pacific. Though it was formerly believed that many species occupied Tahiti and Mooréa, Clarke and Murray showed that only two species were present, with many individual races occupying a series of isolated valleys. Strong morphological differences may occur in direction of coiling, size, shape, and color - yet, many of the identified subspecies interbreed freely in the laboratory. There is good reason to believe that the differences among subspecies are due to genetic drift. In any event, the characters that may be used to distinguish among subspecies are the same that have been used to diagnose different species. In P. suturalis, those mitochondrial restriction fragment length polymorphism (RFLP) genotypes occurring within a population are usually separated by single-step changes. Mitochondrial genotypes can diagnose geographically coherent divergent populations. There is no detectable association between the mitochondrial genotypes and the occurrence of dextral or sinistral populations, showing that chirality does not constitute a genetic barrier (Murray, Stine, and Johnson 1991). The complete local fixation of either right- or left-handed chirality in subpopulations and species, however, may be influenced by difficulties in copulation between forms of different handedness, especially globose forms (Van Batenburg and Gittenberger 1996).

Brande's data on Mulinia allow a comparison of intraspecific versus interspecific variation. In general, the degree of intraspecific variation among populations was less than that among species. This might be explained by either (1) the power of the speciation process in morphological differentiation or (2) the passing of a sufficient amount of time to permit interspecific divergence to transpire via phyletic evolution. Apparently, the latter is the best explanation (Figure 3.9). The recently derived M.

Figure 3.9. Discriminant function plot of first two axes for morphological variation within recent populations of the bivalve Mulinia lateralis (MLR), some Pleistocene populations (MLP), its ancestor the Pliocene M. congesta (MCPLIO), its descendant M. pontchartrainensis (PONTCH), and other recent species (IS and GU). (After Brande 1979.)

Figure 3.9. Discriminant function plot of first two axes for morphological variation within recent populations of the bivalve Mulinia lateralis (MLR), some Pleistocene populations (MLP), its ancestor the Pliocene M. congesta (MCPLIO), its descendant M. pontchartrainensis (PONTCH), and other recent species (IS and GU). (After Brande 1979.)

pontchartrainensis is barely on the morphological fringes of its progenitor, M. lateralis. The Pliocene M. congesta seems to evolve gradually into its descendant M. lateralis. In contrast, seemingly more distantly related species are morphologically more distant as well.

Stasis is used by proponents of the punctuated equilibrium model as evidence for a centripetal force in evolution. Stasis is said to imply a set of "...genetic and developmental coherences that resist selective pressures of the moment and impose a higher level, or macroevolutionary, constraint upon changes within local populations." (Gould 1983b, p. 362).

This argument requires that (1) there be developmental and genetic sources of discontinuity and (2) that these sources be mobilized mainly at speciation. As we have discussed here and will in chapter 4, sources of discontinuity certainly exist. Our evidence, however, suggests that the sources are not associated with speciation. Yet species often are rather constant in morphology. Williams (1950), for example, found little intraspecific variation in cervical articulations in turtles. But some interesting, and quantum, variation could be detected in comparisons among species; is this due to the sort of "resistance" suggested by Gould?

Consider pharyngeal tooth morphology in fishes. Although strong interspecific differentiation is present, intraspecific variation in pharyngeal tooth morphology is relatively slight. This might argue for a centripetal force within the history of the species. A major ecological or genetic crisis might be required to cause the evolution of new forms. This hypothesis can be falsified by examining morphological variation in tooth morphology among clones of unisexual fishes of the genus Poeciliopsis.

Vrijenhoek (1978) found extensive differentiation among clones for trophically significant differences in dentary morphology, involving differences in number and arrangement of teeth (Figure 3.10). These differences coincide with interclonal niche differentiation in feeding behavior. Thus, when clones are formed, many specific and highly differentiated morphologies can be fixed within the geographic range of a species. Major variation typical of interspecific differentiation is thus present, ready to be tapped within any species population. This seems to be common among species of fishes (Vrijenhoek 1978). Stabilizing selection must prevent these combinations from usually appearing.

Therefore, a genetic revolution is not at all necessary to break a pattern of genetic homeostasis. Very likely, natural selection and gene flow prevent the fixation of radically new morphologies. Speciation might be correlated with morphological differentiation, but this is only coincidental with spatial variation in selection pressures. Of course, there are examples in which the introduction of a new allele can destabilize an otherwise regulated (canalized) trait, as in studies of the scute locus of Drosophila melanogaster and at the tabby locus in the mouse. But this does not have any necessary connection with speciation; it can occur as easily within a panmictic population.

Stanley's (1979) monograph on macroevolution, basically a plea for the importance of speciation in morphological evolution, unknowingly revealed the blurred distinction between his conception of speciation and divergent evolution, based on differing selection pressures in ecologically distinctive zones. He asked (p. 72): "Why should all populations of any established species abandon their original niche because adjacent ecological space is free for occupancy? Certainly, expansion of the original niche might be expected, but it is difficult to imagine that this could produce major adaptive shifts without fragmentation into new species. Far more likely would be the rapid invasion of adjacent ecological space in association with divergent speciation."

This association of speciation with occupancy of divergent habitats is precisely the same as a model of divergence based on differential selection of a polytypic

Figure 3.10. Two divergent dentary morphs, found in different clones of Poeciliopsis. These morphologies are associated with different algal grazing behaviors. (Courtesy of Robert Vrijenhoek.)

Figure 3.10. Two divergent dentary morphs, found in different clones of Poeciliopsis. These morphologies are associated with different algal grazing behaviors. (Courtesy of Robert Vrijenhoek.)

species. In other words, it is an ordinary neo-Darwinian model. The appearance of a new resource or habitat would exert strong directional selection, particularly if an old resource or habitat is less abundant.

In summary, patterns of geographic variation and genetic analyses of interpopulation and interspecific crosses and backcrosses fail to substantiate the idea that spe-ciation is a special process with regard to morphological differentiation. Although one cannot say much for those characters that cannot be studied effectively, those amenable to genetic analysis only provide support for the intra-interspecific extrapolation hypothesis.

DNA-level molecular variation and interspecies divergence. Although allozyme divergence between species must correspond to molecular differences, the 1990s completely overturned our general approaches to species identification. Systematists and population biologists are rapidly turning to more direct molecular markers to track the differentiation of populations and species. Unfortunately, our disciplines are still at a formative stage and it is difficult to draw broad conclusions.

Molecular markers of divergence at the intra- and sister-species levels fall into the following classes:

1. Restriction fragment length polymorphisms (RFLPs): RFLPs are generated by incubating mtDNA with restriction endonucleases, which cleave DNA at specific recognition sequences four to six base pairs long (see Dowling, Moritz, Palmer, and Rieseberg 1996 for details). Complete digestion yields a set of fragments of different lengths, which can be visualized by means of electrophoresis. Extremely useful in population analysis, RFLPs also have been routinely used to identify isolated populations, presumably on the way to speciation. Indeed, they are one of a spectrum of tools that demonstrate the continuity between population differentiation and speciation, as opposed to there being a discontinuity (Avise 1994).

2. Direct sequencing of DNA: Protein coding genes, their introns, and other sequences provide an immense potential database for the characterization of populations and species. Sequences have the special value of data that are amenable to phylogenetic analysis, as they provide a set of alternative character states for specific sites, providing sequences can be matched among samples with confidence (see chapter 2). In recent years, new techniques and automation have made sequencing accessible to a wide community of population geneticists and systematists. Most data are restricted to a relatively low number of sequence types, owing to the problem of obtaining primers necessary for the PCR to work (Palumbi 1996). Aside from this and other methodological difficulties to overcome (see Hillis, Moritz, and Mable 1996), genes must be chosen that evolve fast enough to discern population differentiation.

It is not my purpose to provide an exhaustive account of DNA sequence differences among species. As a by-product of the study of evolutionary relationships, many trees have been constructed among closely related groups of species. Some examples will be cited below. Species that appear to be sister taxa from morphological criteria usually turn out to be extremely similar in sequence, but admittedly most of our data are confined to a small number of genes, such as carbonic anhy-drase I and 16S rDNA. My own experience with this comes from a study of the crustacean genus Uca, a pantropical genus of semiterrestrial crabs. The most recently derived species, Uca panacea, identified morphologically and by its characteristic mating wave (Salmon, Hyatt, McCarthy, and Costlow 1978), is nearly indistinguishable in 16S rDNA sequence from its likely progenitor, U. pugilator (Sturmbauer, Levinton, and Christy 1996). Given the restricted geographic range of U. panacea, it is likely derived from U. pugilator.

3. Minisatellites: These sequences come from hypervariable regions of repetitive DNA whose variants can be visualized by Southern blot technique and can be used to trace individuals in populations. Stringent methods are used to identify particular minisatellite loci. It is a population level technique useful for tracing paternity and familial relationships in structured populations (Avise 1994).

4. DNA-DNA hybridization: This method estimates the overall similarity of DNA of different taxa by the rate at which separated strands anneal. The argument is that more rapid annealing represents greater sequence similarity. Specifically, one "melts" DNA, separating duplex strands, and then follows the time course of annealing of DNA from different taxa, as compared with annealing of DNA strands from the same taxon. Then, reheating is done, which dissociates duplexes of different-sequenced strands with more ease than complementary sequences. The temperature at which 50% of the hybrid molecules remain in duplex condition is the usual data used in an analysis. One must assume that the number of nucleotide differences is proportional to the degree of sequence difference (repetitive DNA is removed). Unlike direct sequencing, this technique depends on similarity in hundreds of thousands of nucleotide sites but cannot properly estimate the exact sequence differences. It is therefore sort of a hybrid between genetics and something like a DNA set of calipers.

DNA-DNA hybridization was used on a gargantuan scale by Sibley and Alquist (1983) to estimate species relationships among birds. Although there has been tremendous controversy in some of their conclusions, de facto, this method has sunk into history, owing to the ease of DNA sequencing.

0 0

Post a comment