Evolution, 2(1): 1-9 (Mar., 1948)
Missouri Botanical Garden, St. Louis 10, Missouri
Received November 7, 1947

It has been the experience of most biologists that hybridization between species is rare in nature. Many biologists encounter interspecific hybrids in the field so rarely as to doubt if they really occur there at all or else find them under special circumstances (Epling, 1947) which raise serious doubts as to the importance of hybridization under natural conditions. After a series of investigations (Anderson, 1936b, 1936d; Anderson and Hubricht, 1938; Anderson and Turrill, 1938) it has been my own experience that clearcut out-and-out hybrids are seldom met with, even when a deliberate search is made for them, and that hybrid swarms of bizarre recombinations are found, if at all, only under peculiar circumstances.

It was once the common opinion (Zirkle, 1935) that this lack of evident hybridization was caused by the sterility of interspecific hybrids. Experimental evidence has not confirmed this judgment and modern advocates of reproductive isolation (Mayr, 1940) as a species criterion have had to phrase their definitions to permit semi-fertility between distinct species. For the higher plants there is an impressive amount of experimental evidence on this point though it is widely scattered. There are the papers of the early hybridizers (Focke, 1881 ; Roberts, 1929; Zirkle, 1935), much work in genetics (see for instance East, 1913, 1916) and experimental taxonomy (Clausen, Keck, and Hiesey, 1946), and the experience of numerous plant breeders. The latter, by far the largest of these three bodies of evidence, is not too accessible to most scientists since it has to be dug out piecemeal from such compendia as Rehder's Manual of Trees and Shrubs (1940). A critical summary of all this evidence is badly needed. Lacking one, the evidence for the higher plants may be roughly summarized as follows: Well-differentiated species of the same genus may or may not be interfertile when tested experimentally. On the whole it seems to vary with the genus. There are certain genera in which interspecific hybrids are difficult to make and are sterile. There are others, equally exceptional, in which the widest possible crosses within the genus will yield fertile or semi-fertile hybrids. (Anderson and Schafer, 1931). The yellow trumpet Narcissi are so completely fertile with the flat, white-flowered Poets' Narcissi that the whole business of supplying new garden hybrids has been founded on it (Calvert, 1929). The Poets' Narcissi (in themselves a whole group of species and sub-species) are so interfertile with the species complex making up the long-crowned yellow daffodils that it has been possible to accomplish such recombinations as the transfer of the deep red-orange pigment from the rim of the tiny central eye in a Poets' Narcissus to the brilliant flaring orange trumpet of such modern "red trumpets" as the variety "Fortune" (Anderson and Hornback, 1946). The commonest condition among the higher plants seems to be that in which crosses within a species group are easy to make while intergroup crosses are difficult or impossible.

If, therefore, we base our explanation of the rarity of hybrids under natural conditions on the experimental facts we can make the following summary: 1. In a minority of the cases where related species occur near each other, they are completely intersterile. 2. In certain cases hybrids can be formed but are sterile. 3. In a surprising number of cases the experimental evidence shows that the species can be crossed readily, in the breeding plot, but no hybrids appear under natural conditions. 4. In many cases which have been carefully investigated, hybrids, though usually rare, do occasionally occur but leave no apparent descendants except under unusual conditions.

This experimental evidence, such as it is, justifies the generalization that species maintain themselves as recognizable units even where they are interfertile and even when there is a considerable opportunity for them to hybridize. Why should this be so? Before we can answer the question we must first have clearly in mind the known results of such hybridization under experimental conditions. If we ignore the complications due to polyploidy and other important but more-or-less specialized phenomena, the usual results of hybridization are readily summarized. They were well established by the early hybridizers and have been abundantly confirmed by modern genetic research.

1. The first hybrid generation is intermediate between the parents and is as uniform as they are, or even more so. It is usually more vigorous than either and more robust in nature.

2. The second hybrid generation, while on the average intermediate, is extremely variable. Usually no two individuals are alike. The variation is usually bewildering, but it can be shown to have a general trend from a few individuals more or less like one of the parental species, to a great bulk more or less like the first generation hybrid, to a very few more or less like the other parental species.

3. The facts with regard to the backcrosses are equally well established but unfortunately have been so little stressed in modern times that they are not as generally familiar to biologists. If the first hybrid generation is crossed back to either parent, the first backcross is made up largely of plants which resemble the species to which they have been backcrossed. If we take one of these and cross it back a second time to the same parent, not only will the seedlings resemble this species very much indeed but some of them may he nearly or quite indistinguishable from it. In all the cases with which I am personally familiar, many, if not most, of the first and second back-crosses, if found in nature, would by taxonomists be accepted as varieties or slightly aberrant individuals of the species to which they were back-crossed (see fig. 1). Few taxonomists, even those specializing in that group of plants, would even suspect that many of these back-crosses were of partially hybrid ancestry. For the genus Apocynum we have a detailed experimental record of back-cross morphology and the taxonomic reaction to it (Anderson, 1936b).

In the genus Tradescantia, by rigidly experimental methods (Anderson, 1936a, 1936d; Anderson and Sax, 1936; Anderson and Woodson, 1935), including the production of experimental backcrosses, it has been possible to demonstrate that the principal result of hybridization, in those cases where it did occur, was a series of such backcrosses to one or to both parents. To this phenomenon I gave the name introgressive hybridization (Anderson and Hubricht, 1938), since it provided a mean by which elements in the germplasm of one species might introgress into the germplasm of the other. The chief result of hybridization under such conditions is the enrichment of variation in the participating species. Such hybridization is cryptic and only by very specialized techniques can we measure its exact importance in any particular case. Since the first publications on the subject, the phenomenon of introgressive hybridization has been confirmed, with experimental verification, in numerous genera. (A comprehensive bibliography on introgression by Dr. Charles Heiser is now under way. The following papers are representative: Goodwin, 1937b; Riley, 1938, 1939a, 1939b; Dansereau, 1941; Marsden, Jones and Turrill, 1946; Epling, 1947; Stebbins, Matzke, and Epling, 1947; and Heiser, in press.) Circumstantial evidence indicates its importance in many more genera. How important is it on the whole in the higher plants and in other groups of organisms? We do not yet have any exact evidence. From those cases where experimenters have gone to the trouble of making experimental backcrosses (Epling, 1947), it is clear that many of the second backcrosses could not be recognized as of mongrel origin purely by their appearance. It is therefore clearly indicated that gene flow from one species to another may go far beyond any point which could be detected by ordinary morphological techniques. We shall not be able to assess the real importance of introgression until we can study genetically analyzed species in the field and determine the actual spread of certain marker genes. Until such data are available, any generalizations are based on mere opinions.

Introgression therefore gives us a partial answer to our original question as to why hybrids are so seldom met with in nature. It is because when hybrids do occur, they usually perpetuate themselves, if at all, in backcrosses to one or the other parental species and the mongrel nature of their descendants is not apparent to the ordinary biologist. The commonest result of hybridization is introgression, and introgression must be excessive before it will produce results conspicuous enough to impress biologists who are not making a deliberate search for such phenomena. This is only half an answer. Why do interfertile species limit themselves very largely to backcrosses when they meet under natural conditions and most particularly why is the backcrossing

largely in areas where natural conditions have been very much disturbed? There are at least two main reasons; one resides in the germ plasm itself, the other in the habitat. As to the internal one, the total effect of all the forces which make for specific cohesion is very great, much greater than one would expect until he made careful calculations (Anderson, 1939a & b). Following the arguments used in these calculations it can furthermore be shown that in well-differentiated species the total effect of linkage is so strong that two well-differentiated but interfertile species, meeting in an idealized environment favorable to hybridization, would remain recognizable units in spite of their interfertility. The details of the argument are largely mathematical and are shortly to appear elsewhere, but the general conclusions can be tersely put. Linkage by itself is a force strong enough to prevent the complete swamping of interfertile species. As a factor in specific cohesion it is proportional to the differentiation of the two hybridizing entities: the greater the differentiation, the stronger the cohesive force of linkage.

The effect of the habitat, however, is also important, and it usually operates in exactly the same direction. The argument is as follows: it is now known that the physiological differences between species segregate in the same way as do the morphological ones. In Neurospora (Beadle, 1945) the mode of inheritance of scores of physiological differences are precisely understood. In yeasts the Lindegrens (1947) have demonstrated with laboratory precision the inheritance of various differences in habitat preference. The higher plants are not so amenable to precise physiological analyses of their nutritional requirements but there is abundant circumstantial evidence that a similar situation prevails and there are precise data for a few characters such as maturity (see for instance Goodwin 1937c, footnote 13, or Marsden Jones and Turrill, 1946), response to day length, etc. Nearly everyone who has grown and studied the second generation from a species cross has noted the segregation and recombination of such physiological differences as length of blooming season, resistance to diseases, dayblooming habit vs. night blooming, ease of wilting, resistance to cold, light tolerance, etc. If, therefore, we cross two species differing in their ecological requirements we may expect these physiological differences to segregate as follows: The first hybrid generation will be uniform in its requirements and on the whole they will be for conditions intermediate between those required for the two parents. The second generation will be made up of individuals each of which will require its own peculiar habitat.

Let us repeat this last statement; it is the crux of the argument. THE SECOND GENERATION WILL BE MADE UP OF INDIVIDUALS EACH OF WHICH WILL REQUIRE ITS OWN PECULIAR HABITAT FOR OPTIMUM DEVELOPMENT. As a whole the requirements of the second generation will range from a need for something more or less like one parental habitat to something more or less like the intermediate habitat of the F-1 to something more or less like the habitat of the other parent.

In nature therefore we might reasonably expect to find first generation hybrids growing in an intermediate zone between the two parental habitats. The persistence of any considerable variety of the various second generation recombinations would require a habitat such as is seldom or never met with, where various combinations of the two parental habitats are found in close juxtaposition to one another.

As a crude example let us consider the adjacent habitats in which one finds Tradescantia subaspera and Tradescantia canaliculata at home in the Ozark Plateau (Anderson and Hubricht, 1938). The former grows in deep rich woods at the foot of bluffs while the latter grows up above in full sun at the edge of the cliffs. As an over-simplified example we can list three of the outstanding differences between these two habitats as follows:

rich loam rocky soil
deep shade full sun
leaf mould cover no leaf mould cover

Tradescantia canaliculata and T. subaspera are well-differentiated species; neither one of them is by any means the closest relative of the other, yet Mr. Hubricht and I have found by actual experiment that not only can they be crossed readily by artificial means but they do cross abundantly when left to themselves in an experimental garden (Hubricht and Anderson, 1941). Even though both he and I were familiar with the appearance of these artificial hybrids and though we searched for them at many points where the species were growing very near one another, we found very few of the first generation hybrids. The habitats of the two species are strikingly different; in the Ozarks one seldom finds the intermediate habitat in which the hybrid is able to germinate and survive. This is a more or less intermediate condition, a gravelly soil, partial shade with some bright sunlight, and a light covering of leaf mould. Imagine, however, the habitat which must be provided if we are to see the second generation recombinations which we obtain in the breeding plot. If we consider only the three contrasting characters of the habitat which have been mentioned above, our recombinations would require the following six new habitats in addition to the parental ones (these six represent only the extreme recombinations; a whole series of intermediates will also be required):

rich loam rocky soil
full sun deep shade
no leaf mould leaf mould
rich loam rocky soil
full sun full sun
leaf mould leaf mould
rich loam rocky soil
deep shade deep shade
no leaf mould no leaf mould

What would have to happen to any natural area before such a set of variedly intermediate habitats could be provided? It has been very generally recognized that if hybrids are to survive we must have intermediate habitats for them. It has not been emphasized, however, that if anything beyond the first hybrid generation is to pull through, we must have habitats that are not only intermediate but which present recombinations of the contrasting differences of the original habitats. If the two species differ in their response to light, soil, and moisture (and what related species do not?) we must have varied recombinations of light, soil, and moisture to grow their hybrid descendants. Only by a hybridization of the habitat can the hybrid recombinations be preserved in nature.

The actual inherent differences in ecological preference will of course be much more diverse than in the crude example given above. The number of different kinds of habitats required by the hybrids will rise exponentially with the number of basic differences between the species. With ten such differences, around a thousand different kinds of habitat would be needed to permit the various recombinations to find a niche somewhere as well suited to them as the original adjacent habitats were to the two parental species. With only twenty such basic differences (and this seems like a conservative figure) over a million different recombined habitats would be needed. Under natural conditions anything like such a situation is close to impossible. Ordinarily it is only through the intervention of man that it is even remotely approached. Even in these cases the new "Recombination Habitats" will largely be limited to habitats pretty much like those required by one of the parental species, but which in a few characteristics approach the requirements of the other parent. We may expect that even in such disturbed habitats there will be back-up recombinations not greatly different from one of the parents which will most readily find an ecological niche suited to them.

One of the best demonstrations yet published of the way in which man can provide strange new niches of hybrid recombinations is that given by H. P. Riley (1938, 1939a) in his analysis of the hybridization between Iris fulva and Iris giganticaerulea, two species which differ strikingly in their color, morphology, and ecological adaptations (Viosca, 1935). In one of the localities which he studied in detail on the Mississippi delta, a series of long, narrow farms run straight back from the highway side by side, in the fashion set by the French settlers, with almost the precision of experimental plots. The original environment at that point was fairly uniform, but each man has treated his farm a little differently. It was strikingly apparent from Riley's study that the numbers and kinds of hybrids varied from farm to farm. Some had few or none, while others, even when adjacent, had hybrids in great quantities; there were significantly more of them where the meadows had been pastured. In one farm in particular, the little depression which ran parallel to the highway had been subjected to a series of operations. The trees and shrubs had not been entirely removed from this area, but it had been repeatedly cut over and had in addition been heavily pastured. It had a swarm of different hybrid derivatives, almost like an experimental garden, and the hybrid area went right up to the fenceline at the border of the farm and stopped there.

Nor is this an isolated instance. Viosca (1935) and other students of Louisiana irises have worked out in considerable detail the relation between the production of hybrid swarms of these conspicuously different irises and the churning and rechurning of the habitat by ditching, pasturing, lumbering, road-building, etc. It is only where man has hybridized the natural environments of the Mississippi delta that nature can find an appropriate lodging place for the hybrids she has created.

1 Darrow and Camp also considered the reaction of hybridizing polyploid
complexes with the environment. Polyploidy introduces further complications
into hybridization which are beyond the scope of this paper.

This dependence of interspecific hybridization upon the intervention of man has been described by a number of authors (Darrow and Camp,1 1945; Anderson and Hubricht, 1938). It was discussed in some detail by Wiegand (1935) in his paper on "A naturalist's experience with hybrids in the wild." Marie-Victorin has given a vivid description (1922, p. 32; 1935, p. 65) of its operation when the original flora of the St. Lawrence valley was largely replaced by fields and pastures. Epling, Stebbins, Dansereau, and their students have commented upon the connection in a number of different genera. Does this mean that introgression as a phenomenon is limited to the areas disturbed by man and that its results are mere artifacts and not genuine natural phenomena? I think not. Though freely admitting that nearly all the introgression which has been studied experimentally (for one exception see Dansereau, 1941) is of the nature of an artifact, I believe that at particular times, and in particular places, introgression may have been a general evolutionary factor of real importance.

Under the conditions of an experimental garden, natural selection among the progeny of a cross between species is much less severe than it is in nature. Though the optimum environments for the sister hybrids may be quite various, it is possible to raise the majority of them in one plot, providing that they are widely spaced and competition with aggressive weeds is kept at a minimum. The prevalence of iris hybrids on one or two of the farms described by Riley (1938) may have been due in part to the reduction of competition with other plants, particularly grasses, as well as to the variations in shade and moisture brought about by repeated recuttings and overpasturing.

There must have been various times, even without the intervention of man, when species hybridized under conditions which produced varied new habitats and when competition was not too keen, as for instance when newly colonizable areas emerged from the sea or when various floras spread out onto the northern lands denuded by Pleistocene glaciation. At such times introgression would have been an important evolutionary factor. For one area we are beginning to get actual proof that it did occur. Along the coast of California there are peninsulas which once were isolated islands but which are now united with the main land. In their studies of the California knobcone pines, fossil and living, Mason and his students are demonstrating the actual role of introgression (for a general summary see Cain, 1944, pp. 112-118) in forming these pines as we know them today and to determine in some detail how introgression operated at the time when these islands were joined to the coast.

The Edwards Plateau in central Texas is another area in which introgression may have operated on a grand scale. This comparatively small area is a center of distribution and variation for numerous genera. A mere leafing through of a series of monographs of North American genera (Larsen, 1933; Anderson and Woodson 1935; Barkley, 1937) will demonstrate that it is one of the outstanding centers east of the Rockies. For many genera the concentration of species is higher there than at any other point and for the genus Tradescantia we know the even more significant fact that it is a center for the diploid strains of polyploid species (Anderson and Sax, 1936; Anderson, 1937). The geological evidence shows that when the Edwards Plateau came into being, it united older land masses in Mexico and in the United States. Certainly at such a time related species in many genera might have met and hybridized under conditions where competition would not have been keen and where associations of plants were in the making instead of already existing as tightly closed corporations. Tradescantias from Mexico would have met species coming down from the Appalachians in an area conducive to the survival of some of the hybrid recombinations.

Woodson has recently (1947) called attention to the importance of peninsular Florida in the speciation patterns of the eastern United States. During parts -of the Tertiary it was an island or group of islands which finally became attached to the mainland. Species and varieties which became differentiated during the island period must then have had unusual opportunities to hybridize with their relatives on the mainland. Giles' (1942) studies of Cuthbertia in this area have given cytological proof that such hybridization did actually take place. Careful studies of variation throughout the whole area, for a series of species, should yield data with which we could assess the general overall importance there of introgression.


1. Experimental evidence shows that sterility will not account for the rarity of hybrids under natural conditions.

2. Careful field analyses have shown that natural hybridization is largely limited to backcrosses which resemble the parental species so closely that special methods are required to detect them readily.

3. One of the factors limiting hybridization to such introgression is imposed by the habitat for the following reason: Two species differing in their habitat requirements will produce a first generation hybrid adjusted to a uniform intermediate environment. The second generation however consists of individuals each of which requires its own peculiar habitat for optimum development. Such heterogeneous habitats are seldom or never met with, the only approach to them being found in places where man has greatly altered natural conditions.

4. It is concluded that hybrid swarms can survive only in "hybridized habitats." While most of the latter result from human intervention, similar conditions have prevailed in pre-human times when new lands were opened up to colonization by diverse floras. At such times and places introgressive hybridization must have played an important role in evolution.


Edgar Anderson Bibliography