[PHYLOGENETIC PATTERNS PREDICTED BY THE CASCADING SPECIATION MODEL]
The most difficult conceptual problem that any chromosomal or parapatric speciation model must meet is to adequately explain how any chromosomally or otherwise derived population which remains in contact with its parental species is able to evolve a completed barrier to gene flow. Not only does the evolution of this barrier define the new species, but a complete barrier is needed to protect the chromosomal rearrangement of other intrinsic impediment to genic introgression from eventual swamping and elimination in the face of continued challenges from ancestral chromosomes and/or genes which would otherwise successfully pass through semi-fertile chromosomal heterozygotes or negatively heterotic hybrids. In this context, note that if geographic isolation is required for the protection of the chromosomal rearrangment, there is no need to postulate a function for fixation of the mutation in speciation; yet, the comparative data reviewed above clearly tell us that speciation patterns in chromosomally variable lineages are qualitatively different from those in chromosomally conservative lines. The foregoing discussions of chromosomal differentiation and cascading speciation are therefore relatively futile exercises in logic unless it can be demonstrated that a chromosomally differentiated population has a reasonable probability of completing [the evolution of] its genetic isolation in the face of continued contact with its parental species.
Under circumstances where a niche sympatric to that of the ancestral species is readily available to a newly differentiated deme, and genetic polymorphism of the deme provides an appropriate substrate for selection, selection for differentiation into the available niche and against wasting gametes in less fit hybrids may possibly lead to rapid character displacement and the evolution of premating isolation, which could allow the derived species to begin expanding into the sympatric niche within a comparatively few generations. However, in most cases I doubt that the chromosomally derived "founder" population would contain enough of the kind of variability needed to allow the immediate evolution of premating isolation. Where there is no premating isolation, for the differentiated deme to increase its range, it must be able to geographically displace the ancestral population; and for the differentiated population to survive swamping by the continued introduction of ancestral genes and chromosomes it must expand its range enough so that no ancestral individuals can disperse far enough to reach its central demes. Therefore, in most cases of chromosomal speciation, where premating isolation and sympatry cannot immediately be evolved, the nascent chromosomally differentiated species must pass through a period of contact hybridization7 with its parental species.
As demonstrated by the zones of contact hybridization between the different Sceloporus grammicus populations and between Sceloporus woodi and undulatus, when the hybridization of species in peripheral contact is examined in sufficient genetic detail, paradoxically the hybrid zones are found to serve as complete barriers to gene introgression beyond the narrow limits of the hybrid zones themselves--even though hybrids are proven to maintain appreciable levels of fertility. Once the reality of this seemingly paradoxical observation is accepted, a model to explain the surprisingly early genetic isolation of the hybridizing populations is suggested. And most interestingly, the hybridization model generalizes to any situation of contact hybridization, whatever its origin, as long as certain limiting conditions are met by the hybridizing populations.
Bush (1975) supposes that when ancestral and derived populations hybridize in a contact zone, the reduced fitness of hybrids will selectively favor the evolution of premating isolation. This assumption might be reasonable for cases of sympatric hybridization (Woodruff, 1973), but as we will see, it is quite inappropriate for any situation of contact hybridization, because premating isolation not only does not evolve as a result of contact hybridization, but it cannot do so. Hard evidence mounts to show this.
As noted by Bush (1975), who cites many examples, when the geographic variation of perfectly reasonable museum "species" is examined karyologically or with other fine-toothed techniques such as electrophoresis, which allow the use of specific gene products as markers for gene flow, many "species" are resolved into geographic mosaics of chormosomally and/or genetically distinctive, parapatrically distributed sibling species. (Here I will use the term "sibling species" to denote the genetically isolated "sister" units of such parapatric mosaics. The semi-species concept implies a unit in a mosaic of allopatric populations where the status of genetic isolation is frequently indeterminate--Mayr, 1963.) A few additional cases of parapatric mosaics of sibling species not cited by Bush (1975) are: Perognathus goldmani complex (Patton, 1969), Sigmodon hispidus complex (Zimmerman, 1970, Johnston, et al., 1972) Anolis brevirostris complex (Webster and Burns, 1973), Rana pipiens complex (Moore, 1975), Crinia laevis complex (Littlejohn, et al., 1971), Litoria ewinge complex (Watson, 1972), and Podisma pedestris complex (Hewitt, 1975).
In many cases where such sibling species complexes have been geographically mapped in enough detail to locate exactly the zones of contact between their constituent parapatric populations, extremely narrow zones of contact hybridization have been found. Usually these hybrid zones are much too narrow to be zones of intergradation or introgression. But, on the other hand, given the high levels of hybridization observed (e.g., Nevo and Bar-El, 1976; my observations of grammicus hybrid zones), there is equally little or no evidence of premating isolation between the hybridizing populations, even though in many cases the "pure" populations are seen to differ by ecological, behavioral, or other potentially discriminable features which might easily be enhanced by selection to achieve this isolation. Morphological differentiation is exemplified in the morabine grasshoppers by the cereal hook of P24(XO), which presumably could be adapted to mechanically prevent copulation with the parapatric viatica19 with which it normally hybridizes (White., et al., 1969). Behavioral differences are exemplified by the strikingly different "mating" calls of Crinia laevis (Littlejohn, et al., 1971) and lesser but still significant differences between members of the Litoria ewingi complex (Watson, et al., 1971). However, in an old and seemingly forgotten paper, Moore (1957) presents several difficulties which he believes are sufficient to negate the effects of any antihybridization or character displacement selection resulting from contact hybridization which might otherwise favor the evolution of premating isolation. Two of his ideas are especially cogent:
A minor point of Moore's still worth mention is that as the efficiency of any isolating mechanism based on character displacement or antihybridization selection improved, the selection pressure to further improve it would decline, such that complete isolation could be approached asymptotically, but would never be reached. As noted by Moore, these objections would not apply only in cases where a substantial proportion of an entire species risks hybridization, as in sympatric hybridization; or as would be probable for a small deme in which a heterozygously semi-sterilizing mutation had just achieved fixation. However, if the mutant population of the nascent species survived to spread to any extent, but still had not evolved premating isolation, a situation of contact hybridization would result, and the arguments given above would apply; as they must apply to any case of contact hybridization, whatever its origin.
On the other hand, no matter how relevant the arguments above may be in raost cases of contact hybridization they should not apply to the contact on the eastern divide of the Valley of Mexico where the P1 and F6 populations of the Sceloporus grammicus complex hybridize (Hall and Selander, 1973) or to the hybrid zones between Sceloporus undulatus and woodi (Jackson, 1973a, 1973b). In both cases, highly convoluted hybrid ozones completely surround populations in which no individual may be more than 10 to 20 times the width of the hybrid zone itself from the risk of hybridization. Probably 5 to 10 percent of the total of some P1 and woodi populations risk hybridization in every generation.
If the hybridization resulted in reproductively fit offspring, there would of course be no reason to evolve premating isolation. But this is not the case: Although F1 hybrids in both hybrid zones reproduce, there is some evidence for recombinational breackdown in the backcross generations and there is good evidence that there is no introgression into any of the pure populations outside of the narrow hybrid zones. In other words, any interpopulation mating represents a genetic dead loss to any genetically pure individual which becomes involved in such a mismating. However, in the grammicus case, although P1 and F6 differ ecologically and show significant (but not completely diagnostic) differences in male head-bob patterns, coloration, and adult size, and where any or all of these differences could be shaped by selection to improve premating isolation, there is no indication either in the field or in the lab that either of the hybridizing species successfully discriminates against mating with the other (Moody, 1971). Moore's (1957) arguments discussed above might apply to the failure of widely disctributed F6 or undulatus populations to avoid mismatings, but they do not plausibly fit the P1 or woodi populations, where a substantial fraction of the entire species risks hybridization in every generation. Alternatively one might argue that these arecases of recent secondary contact, and that the populations simply have not had enough time to evolve premating isolation. However, this doesn't fit either. There is every indication that woodi and undulatus would have been in contact at least since the last sea-level periods allowed the present emergence of low-lying areas of the Florida Penninsula (Jackson, 1973b), and it is my belief that P1 and F6 were never out of contact since P1 became a distinct population (Hall and Selander, 1973). If we then accept for the sake of the argument that these contacts are old and that Moore's arguments do not apply to them, we are left with the unexpected conclusion that something very fundamental is happening in the hybrid zones themselves which retards the evolution of premating isolation.
Mayr (1963) suggests that introgression through the hybrid zone might serve to prevent the evolution of isolating mechanisms by breaking up gene complexes that would otherwise evolve to prevent the hybridization. However, even in situations of parapatric contact where the hybridization is demonstrably relatively free and backcrossing is observed, in several cases there is still no evodence for the introgression of genic markers from either hybridizing population into the other beyond the limits of the narrow hybrid zone itself. Two parapatric complexes provide the most precise documentation for this.
In Sceloporus grammicus genic or chromosomal markers in two unrelated and precisely mapped contact zones allow F1 and backcross genotypes to be clearly distinguished (P1 x F6 [Hall and Selander, 1973; Hall, Moody, and Selander, unpub.] and S x FM2 in the ruins of the ancient city of Teotinuacan [Hall, Stamm, and Reichlin, unpub.]).
Similarly, in the Spalax ehrenbergi complex (Wahrman, et al., 1969), although genic markers are lacking (Nevo and Shaw, 1972), chromosomal differences in two of the four types of hybrid zones possible allow this distinction to be made (Nevo and Bar-El, 1976). In 3 transects across the two appropriate hybrid zones in Spalax, the high frequency of backcross gene or chromosomal complements proves indisputably that hybrids successfully backcross with both parental types, so we know that hybrid sterility alone cannot be a barrier to gene flow in any of these cases. In the F6 x P1 case, where the several genic differences allow the distinction to be made, it appears that recombinational breakdown in the backcross generations may contribute to the genetic barrier (Hall and Selander, 1973), but even here it remains to be proved that the backcross barrier is complete. Most interestingly, in the Spalax examples, the widths of the hybrid zones are inversely proportional to the number of chromosomal differences between the hybridizing populations (Nevo and Bar-El, 1976). Also, in the narrowest hybrid zones of Spalax, the frequency of backcross chromosome numbers exceeds that which would be expected from a random-mating model for the karyotypic differences.
Although mating in the hybrid [zone] is obviously not random, because of the strong geographic component, the excess number of backcrosses may imply that backcrossing continues for more than one generation--which would prove that the backcrosses were at least partially fertile. Additionally, as supposed by the parapatric and chromosomal speciation models, there is no evidence that the sibling species in the Spalax complex have only recently come into secondary contact.. Israli populations of Spalax sort themselves out clinally along a climatic gradient from mexic to xeric, and it does not seen reasonable that the different chromosomal populations used only parts of this gradient to allow geographic barriers between their populations. The situation in grammicus is precisely similar. Therefore, in both Sceloporus and Spalax, we have convincing evidence for free hybridization and backcrossing in the zones of contact hybridization, and seemingly contradictory but equally convincing evidence that there is no introgression or gene flow out of the zones. Although other cases have not been studied as thoroughly as those reviewed above, let it be assumed that this paradoxical situation is the normal situation for stable zones of contact hybridization.
Summarizing all of the apparent paradoxes associated with interactions of species meeting in zones of contact hybridization will make the problem clearer:
In other words, paradoxes or no, contact hybridization allows the hybridizing populations which meet in peripheral contact to evolve independently as perfectly good biological species from the moment a contact zone stabilizes in such a form. And, so long as the hybrid zone remains narrow, and does not break down into intergradation; or the evolution of premating isolation outside of the contact zone does not allow sympatry; then one species can have no effect on the other beyond determining the location of their mutual contact and geographically one species from the range occupied by the other. In any event, irrespective of the actual situation in nature, all of these seemingly paradoxical conditions can easily be explained by a theoretical model (Hall, 1973).
Key's (1968) "surface tension" idea described the phenomena implied by argument 7 above. If a hybrid zone is curved so that the radius of curvature of its central axis is not large with respect to the width of the zone, the interface or surface of the convex population with the hybrid zone will have a significantly shorter length per included angle of the curve than will the surface of the concave population. Holding all other population parameters equal, the shorter interface of the convex population will contribute fewer migrants into the sink per unit angle of curvature than will the longer interface of the concave population. Consequently, the convex population will be pushed back to a less curved distribution which will again be able to acheived an immigration pressure that is balanced with that of the opposed population.
This model also easily accounts for Key's (1968) observations that "tension" zones interact to merge when they are close together to become more effective: A population caught between two hybrid zones will have to use its fixed rate of intrinsic increase to feed migrants into two sinks; while each outer population, facing only one zone, can put its entire surplus of migrants into the single zone it faces. Therefore, if the central population is narrow enough so that its central demes must feed both hybrid zones, the two outer populations will begin to push their respective hybrid zones together to eliminate the intervening population. Then, when the two different hybrid zones merge, their negative effects on hybrid fitness should be additive to increase the effectiveness of the sink, as supposed by Key (1968). Exactly this phenomenon appears to be demonstrated by the negative relationship in Spalax between width of the hybrid zone and number of chromosome mutations differentiating the hybridizing populations (Nevo and Bar-El, 1976).
Note that any small population which initially becomes fixed for a heterozygously semisterilizing chromosome mutation immediately faces this surface tension effect. This, in turn, has interesting consequences for the predicted effects of chromosomal speciation on phylogenetic patterns. If the nascent chromosomally differentiated population is not at least partially protected from surrounding parental populations by local barriers to migration, its fitness in the local area must be considerably higher than that of the adjacent parental populations if it is to survive the constricting effects of the surface tension of the hybrid sink. Probably many nascent chromosomally differentiated populations are quickly swamped by this constrictive effect. On the other hand, given the effects of selection for better adaptations to the local environment, and that the population structure required for chromosomal fixation will also encourage drift at many gene loci, it is reasonable to assume that a chromosomal differentiate will occasionally and independently from the chromosomal mutation achieve a particularly fortuitous gene combination which confers a superior local fitness sufficient to counterbalence the surface tension. These populations would be the only ones to survive as incipient species. Then, once the range of such a nascent species increased to where its curvature flattened enough to significantly reduce the constricting effect of its surface tension, the superior fitness of its population relative to the ancestral species would fuel an expansion out of the local area. Expansion would then continue until a point of ecological balance was reached with the parental population or until the parental population was exterminated. I
t can be seen that the surface tension effect of the hybrid zone will function as a group selection filter which will allow only populations with greater fitness than surrounding ancestral populations to survive as nascent species. It obviously follows that the end products of a cascade of speciation will very likely be superior competitors which would be able to replace or displace chromosomally more primitive species in their radiations, as has obviously occurred several times in the radiation of Sceloporus.