Introgressive Hybridization (1949)
Edgar Anderson

Chapter 6

Special Techniques for the Study of Introgression

For the most part this chapter will deal with the special techniques that have been developed for apprehending introgression in the field. It should be emphasized at the outset, however, that, although these are powerful techniques and although they allow us to make reliable estimates of the probability of hybridization from field data alone, they will be more fruitful if combined with the more traditional techniques such as transplant experiments, progeny tests, cytological examination of species and hybrids, and the experimental repetition of the suspected cross. Where it is feasible to carry on this kind of experimentation it is particularly important to study artificial backcrosses of the hybrid to each parent. Until these have been made, one does not have even a rough estimate of how much undetected hybridization there might be in supposedly unmongrelized populations of the parental species. Of all the kinds of experimental evidence which might be gathered on such a problem, the production of artificial backcrosses is of outstanding importance. The mere demonstration that such and such a species hybrid can actually take place under natural conditions is no longer of any general significance. That these crosses can sometimes take place is now proved beyond a reasonable doubt. What we do not yet know is the role (or rather the roles) such hybridizations play in evolution. If we are going to measure the effect (or lack of effect) of hybridization in natural populations, then one of the most useful kinds of evidence we can obtain experimentally is an exact understanding of what is to be expected when the hybrid crosses back to either parent.

The chief disadvantage of these orthodox methods of hybrid analysis is that they can be applied only when the parental species are known, or at least strongly suspected. They are useful largely in proving that certain hybridizations might have taken place. They cannot be used analytically as a basis for successful prediction.

For the examination of hybrid populations or of populations in which hybridization is suspected, we need methods that record precisely the extent to which variation in one character is related to variation in other characters.

The human mind is inefficient in judging variation in more than one variable at a time. A good observer may examine three different populations and note them efficiently for their variation in pubescence, in leaf shape, or in flower color, but careful tests have shown (Anderson, unpublished) that scientists cannot look at three populations varying simultaneously in flower color and pubescence and leaf shape and render an efficient judgment of the comparative association between these characters in the three different populations.

What is needed, therefore, in describing populations is some means of recording simultaneously variation in several different characters. Species characteristically differ by slightly different proportions and trends in proportion for several different characters (Anderson and Whitaker, 1934; Anderson and Ownbey, 1939). We can differentiate most effectively between interspecific and intraspecific variation if we have some method for showing the relationships between the main variables in the population.

For such a purpose the methods of conventional biometry are laborious and inefficient. They were developed for other types of problems, and though they are fairly good for analyzing variation in any one character they are not efficient for exploring relationships between groups of characters, particularly when we do not know in advance the general nature of that relationship.

However, any methods with which we replace or precede biometrical analysis must, like it, be exact, objective, and verifiably accurate. The description and analysis of a population is one of those problems that must first be analyzed precisely on a morphological level before we can choose the best methods with which to analyze it on a mathematical level. The most effective methods so far achieved are of various sorts, but they share one feature so universally that they may be grouped under the general name of polygraphic analysis. That is to say that they are all more or less graphical and that they all in one way or another summarize the variation in two or more characters in a population. These various methods of polygraphic analysis may be listed as follows:

  1. Scatter diagrams.
  2. Pictorialized scatter diagrams.
  3. Ideographs.
  4. Hybrid indices.
  5. Radiate indicators.
  6. Standardized photographs.


Scatter diagrams are the simple alignment of dots in a two-dimensional field, such as were used in Chapter 3 in describing the possible relationships of flower color and pubescence. Since one of the steps sometimes employed in calculating the correlation coefficient is the preparation of a scatter diagram, it may be well to point out specifically that for population analysis scatter diagrams are greatly superior to the correlation coefficient as well as much easier to prepare. It is unfortunately not generally realized by most biologists that scatter diagrams may show various kinds of relationships that are ignored or distorted in the calculation of correlation coefficients (see Walker, 1943, pp. 237, 238).


For all their excellencies, scatter diagrams are a somewhat limited form of polygraphic analysis because the relationships of only two characters can be considered at a time. We can get around this handicap by letting the shape of the dot represent a third character, and the color or intensity of the dot a fourth. These pictorialized scatter diagrams are of very general usefulness in analyzing for oneself some of the main relationships in a population that one is just beginning to study. In studying variation in fields of North American maize, kernel width was diagrammed (Fig. 18) on the horizontal axis, and number of rows of kernels on the vertical axis; the shape of the dot represented the degree to which the kernel was pointed at its apex, and the intensity of the dot was proportional to the amount of soft starch in the kernels.

In making a population analysis by this method one takes a random sample of 25 ears from each corn field and records for each ear the kernel width, row number, amount of soft starch, and shape of the kernel. In the resulting diagram, each clot represents 1 ear. From the diagram as a whole, one can tell at a glance the range of variation and the average for each of these characters, as well as the relationships among all 4.

It is possible to demonstrate the reliability of the above method, though not in a quantitative way. If repeated samples of 25 are drawn from the same population, one can see at a glance that the diagrams are essentially similar. At the top of Fig. 18 are 2 samples from the same variety, with and without the addition of artificial fertilizer. At the base of the figure are 2 other varieties grown in the same Guatemalan town. It will he seen that these pictorialized scatter diagrams distinguish between varieties but give consistent results for the same variety even under somewhat different environmental conditions. This is not just a happy circumstance; 5 years of preliminary studies of many kinds of maize under various conditions of growth had been carried on before these 4 characters were finally chosen as the most reliable.

These pictorialized scatter diagrams are particularly useful because they also lend themselves to summarization. In Fig. 18 each dot represents a single ear. It is possible to calculate an average ear from each of these samples. One can then compare the averages of fields, town by town or region by region. By this method it was possible to demonstrate (Anderson, 1946) in an exact and objective summary, how the prevailing corn type changes, within 300 miles, from the wide-kerneled, few-rowed types of western Mexico to the many-rowed, small pointed kernel types of central Mexico. By choosing appropriate characters and symbols this method can be adapted to any kind of material. On page 97, in a demonstration of the method of extrapolated correlates, pictorialized scatter diagrams are fitted to Riley's data on introgression in Iris.

FIG. 18. Pictorialized scatter diagrams for 4 samples of maize, all from the town of Quezaltenango, Guatemala. Above: the same variety grown in a manured and in an unmanured plot. Below: two very different varietics grown in adjacent fields. In all four samples each of the 25 spots represents 1 ear of maize, the shape of the spot representing the degree to which time kernels are pointed, and the blackness indicating the relative amounts of hard and soft starch in the kernel. These four diagrams demonstrate that superficial differences due to environmental effects are scarcely apparent (note the similarity of manured and unmanured plots), while fundamental differences are made conspicuous. Though there is much variation in each, "Nueva Cuartel White" differs from "Nueva Cuartel Yellow" in having on the average more pointed kernels, more soft starch, higher row numbers, and narrower kernels.


Though these have been employed in a number of different problems, they are not so generally useful in population studies as scatter diagrams. They are laborious to make and difficult to reproduce in quantity. However, in certain problems in which it is important to demonstrate all the relationships between a number of different measurements they are greatly superior. Ideographs are even more pictorial than scatter diagrams. In making them the original measurements are recombined in a diagram that is a more or less conventionalized representation of the object measured. They have been used extensively by Alpatov (1929) in his work on geographical differences in bees and in Anderson's studies of iris (1936c). In this latter work, the four measurements (length and width of petal; length and width of sepal) were combined to produce a figure (Fig. 19) that represented a conventionalized white petal lying on top of an equally conventionalized black sepal.

Though they are laborious to construct, the importance of ideographs lies in the fact that they show so many things at once. For the iris ideographs, each one shows fifteen separate facts. That is, if the ideographs were to be replaced with statistics, it would be necessary to employ fifteen separate measurements and ratios for each ideograph. There are first of all the four original measurements—sepal length, sepal width, petal length, and petal width; then there are the six proportions between these four, taken two at a time (the length of the petal in proportion to its width, the width of the petal in proportion to the width of the sepal, etc.); then there are four three-way relationships (such as the length-width of the petal in relation to the length of the sepal); and finally there is the relationship of all four measurements taken at once.

Diagram showing typical flower of I. virginica and resulting ideograph.

Diagram showing typical flower of I. versicolor and resulting ideograph.
FIG. 19. Diagrams showing how measurements for sepal length and width and for petal length and width can be grouped into "ideographs" for analyzing variation in two species of Iris.


This type of polygraphic analysis has been used by several students of populations, notably by Norman Fassett (1941) and by Carson and Stalker (1947), but apparently has never yet been dignified with a name. Radiate indicators are useful in presenting for a number of different populations the occurrence of certain different traits or subtypes.


One of the most difficult types of population to analyze is one in which two or more species have hybridized freely and produced second-generation hybrids and backcrosses. Suppose, for instance, that the two species differ principally in flower color, in petal shape, and in plant height. In the second generation of hybrids and in backcrosses there will be various and multitudinous recombinations of flower colors, shapes, and heights, and no two plants will look very much alike. If we are to make an efficient comparison of two such populations, or a series of them, we must have some means of getting an overall picture of each population so that, roughly at least, we can equate one to another.

For such situations there was evolved (Anderson, 1936d) a method so crude that it was published only after its general usefulness had been demonstrated in a number of different problems. It consists in drawing up a list of differences between the hybridizing entities. All the plants in the hybrid population (or a random sample of them) are then scored individually for all these characters. Attributes like sepal length or petal length are measured; colors can be recorded by comparison with a graded series as on the Munsell and Fischer color charts. Differences in shape can be scored as essentially like one species, or like the other, or intermediate. Raunkiaer (1925) had used and published such a method for showing the great variety of character combinations to be met with in Crataegus populations. By the simple additional step of throwing all these differences together into a composite index, it was possible to extend the usefulness of this method into the domain of analysis. One could then employ it not merely to report the condition he had discovered in a certain hybrid colony but also to inquire into the forces that had produced the variation.

* See Chapter 1, pp. 2-11.

In the simplest application of this method each character (sepal length, petal color, height of plant, numbers of nodes, etc.) was scored in three grades: (1) similar to one species, (2) intermediate, and (3) similar to the other species. One of the species was arbitrarily selected for the low end of the scale, the other for the high end of the scale. Each character, therefore, was scored 0 if it was like the former, 2 if it was like the latter, and 1 if it was intermediate. Supposing 6 characters had been chosen for study, we would then have had a scale running from 0 to 12. Plants exactly like the first species would have scored 0 in every character, and the total score of each plant would have been 0. Plants exactly like the second would have scored 2 for each of the characters, and their total score would have been 12. Plants that were exactly intermediate would have scored 1 for each character, and their total score would have been 6. In actual practice it is usually advisable to give different score values to certain characters, either because they can be more accurately measured and therefore deserve more consideration as criteria, or because they are known to rest upon a wider genic basis and hence are representative of a large portion of the germplasm. In Riley's study of introgression in Iris (1938),* tube color, sepal length, petal shape, stamen exsertion, size of style appendages, and presence of a crest were all scored as like Fulva, like HGC, or intermediate. The color of the sepal was scored in five grades from 0 to 4, and the length of the sepal in four. This gave an index running from 0 for plants like Iris fulva to 17 for plants like Iris giganti-caerulea. Riley has given a meticulous description of the way in which the hybrid index was constructed in this particular study (loc. cit., pp. 727-734), to which the interested reader is referred for further details.

In such cases as hybridization between the Louisiana irises, in which the differences between the species are conspicuous and many of them are easily measured, this method is simple to apply and yields satisfactory results. When the contributing parental species are closely similar or only vaguely different, it is much less satisfactory. Hubbs and Hubbs (1943) have replaced it in their studies of hybridization in fishes with a similar but statistically more elegant method that is superior for their material. At the present time, at least for plant material, the Hybrid Index Method is a powerful means of analysis. It is efficient in exploring a complex situation and pointing out the general overall picture. In my own estimation its main application is in digging into such a problem. When the main facts have been secured, one can then work out a more precise technique adapted to any particular case. From a statistical point of view it is a crude device, and although it could easily be turned into something more respectable mathematically, for the higher plants at least, the time is premature. When we know more about hybridizing populations than we now do—when, in other words, the general problem has been more thoroughly explored on a biological level—we shall then be ready to work out more precise and elegant methods for dealing with such phenomena.

To understand the value of methods as mathematically crude as the Hybrid Index, one needs to keep in mind the general principle behind the doctrine of significant figures: A chain of evidence is no stronger than its weakest link. Precise methods of analysis can be applied effectively only when the nature of the problem is critically understood. In dealing with anything so complicated as hybridization under natural conditions, we need a quick method for roughing out the problem. To take an actual instance, the employment of this method in the field demonstrated effectively that what at first sight appeared to be a large, more or less freely interbreeding hybrid swarm was instead a series of highly localized populations each with its own micro-environment and its own direction of selection. Until our understanding of the dynamics of vegetation is much more precise than it is at present, we shall need simple, diagnostic field methods for summarizing in populations variation trends that are too complex for the unaided mind to grasp efficiently.


The invention of the miniature camera has made it possible to take large numbers of photographs at minimum expense. Properly standardized, such photographs become an efficient record of population variation, but they have been little used. Their earliest employment was by A. J. Wilmott of the British Museum in his studies of population differences in Salicornia. To date, their only published demonstration has been in Erickson's studies of Camassia (1941) and in the studies of maize from this laboratory (Anderson, 1947; Brown and Anderson, 1947), but they have been used extensively in various laboratories for population analysis on a variety of material.

Though it is a basically simple technique, it can be given greater precision. The first point to be borne in mind is that standardized photographs are something more than just photographs. They are exact, standardized records and need to be made in as routine a fashion as possible. Since large numbers of them will be very much alike, it is an absolute necessity to photograph the title on each picture, near the edge if need be, so that it can be cut out if the photograph serves as a published illustration. The background should be neutral, identical for each series, if possible, and the scale should be photographed in each picture. Two examples will show the ways in winch this technique may be adapted to population problems. (1) As worked out by Dr. W. L. Brown (Brown and Anderson, 1947) for Zea Mays: A 10-foot white board (hinged in the middle for more ready storage) is securely fastened to the north side of a field laboratory. At 25-centimeter intervals, lines of black adhesive lantern slide tape are stretched across it to provide a scale. Down the center of the board a series of nails driven part way in and with their heads filed off provide a rack by which the corn plants can be quickly affixed to the board. Labels give the year and the record number of each plant. The leaf above the ear (usually on a sister plant) is traced on wrapping paper and photographed in a standardized position at the left of the photograph. (2) In studying Nicotiana hybrids the calyx and corolla and the dissected limb of the corolla were photographed in a standardized fashion against a frame just one half natural size. By printing these pictures on an enlarger equipped with a frame of natural size, it is a simple matter to produce a large number of exact, standardized records all of them just twice natural size.

This is one of those simple techniques that are more important than they seem. Everyone who has tried it has learned unexpected things about the material he was studying. When one sits down afterwards with a set of standardized photographs of variable populations, it is possible to see slight trends in variation or regional differences, which had completely escaped one in the field.


The methods described above have been used in the field, in the experimental plot, and in actual plant breeding with a great variety of hybrid material. At first in a very tentative way, and later with increasing confidence, they have been employed to determine the putative parentage of hybrid swarms. The general method, which is here formally designated for the first time as the Method of Extrapolated Correlates, has a sound theoretical basis (Anderson, 1939b; see particularly p. 692, where the theory's application to criteria of hybridity was specifically pointed out). It was presented pragmatically by Anderson and Turrill in 1938, its application to a particular example being illustrated step by step.

The method of extrapolated correlates is based on the demonstration (set forth in detail in Chapter 3) that in a species cross all the multiple-factor characters are linked with each other (Anderson, 1939b). When well-differentiated entities hybridize, we may expect their cohesive forces to continue to operate for many successive generations in hybrid swarms. Certainly for scores, and perhaps for hundreds, of generations, we may expect to find the characters that went into the cross together still tending to stay together. By a precise and detailed examination of such populations we can discover the cohesive centers of variation still existing within them. By comparative, quantitative methods we can draw up descriptions of the original entities that must have operated to produce these centers of variation. It is possible, working with a single variable population of a species previously unknown to the investigator, to draw up a precise description of the other species which is introgressing into that population. The subsequent discovery that such a species does actually exist and could have operated in that area cannot be dismissed as a remarkable coincidence; when the prediction has been verified for a complicated series of technical details, it then becomes proof. It is even possible by this method to work with a hybrid swarm and draw up detailed descriptions of both parents when neither of them are known to the observer. Crude examples of such a prediction are given in Anderson and Turrill (1938) and in Anderson and Hornback (1946). The method has since been considerably refined. It will be illustrated below from the data presented in Riley's paper on introgression in Iris (Riley, 1938).

A portion of the data from Tables 1, 2, 3, and 4 of Riley's paper were presented (page 3) in Table 1 in a slightly simplified form. The figures for sepal lengths have been rounded off to the nearest centimeter. In Riley's paper the method of attack was to examine the two species first, and from a study of them attempt to analyze what was taking place in the hybrids. Using the method of extrapolated correlates, we shall demonstrate from these same data how one may work backwards from the introgressants, to the species from which they were derived. For the purposes of the illustration, therefore, let us suppose that only Iris hexagona var. giganti-caerulea is known to us and that we have come upon Colony H-2, which is much like that species on the whole yet is more variable and shows several variants outside the ordinary range of that species. In the discussion below, following the convention established in Chapter 1, we shall use HGC to designate Iris hexagona var. giganti-caerulea and Fulva to represent Iris fulva.

For the analysis, what we need is some simple method of determining for the whole population what characters are tending to stay together and in what patterns. We shall work with pictorialized scatter diagrams, choosing for time horizontal and vertical scales two characters each of which can be measured fairly exactly in a series of grades. In Riley's data these conditions are met by petal length and by color of sepal blade. The latter, thanks to the particular chart used by Riley, was scored in a series arranged with increasing redness from violet blue through blue violet, violet, and red violet to red. Diagramming increasing redness on the vertical axis and petal length on the horizontal axis, we produce the dots of Figs. 20 and 21 for a population of HOC and for our problem population I-I-2. From an inspection of these dots it is apparent that redness and petal size are tending to stick together, particularly in those individuals at the left of Fig. 21 which are outside the range of ordinary HGC. We accordingly examine Riley's table to see what other characters are varying and to see how these two extreme individuals fit into this other variation. There are five such characters, each one of which Riley scored in three grades. We add these to our large dots (each one of which represents an individual plant) by using much smaller bars at five different positions around their circumferences. Tube color is represented directly above, petal shape horizontally to the right, stamen exsertion directly below, style appendages horizontally to the left, and the presence of a crest diagonally to the left. Each of these characters can be represented with no bars for one extreme grade, with a short bar for an intermediate development, and with a long bar for the other extreme.

FIG. 20. Pictorialized diagram of 23 plants of Iris hexagona var. giganti-caerulea, scored by the symbols shown in Fig. 23 from H. P. Riley's published data.

On the hypothesis that, if redness and small petal size came into this population from the same source, other characters may have come in with them, we assume that the peculiarities which we find tending to stay together in the two individuals at the upper left of the diagram are doing so because their genes were introduced into the population together. Since all seven of these characters are apparently multiple-factor characters, the chances are inconceivably small that the genes for all could vary simultaneously. That redness, smallness, yellow tube color, petal shape, stamen exsertion, a small style appendage, and absence of a crest all are tending to stay together in this population is most readily explained as due to the influx of whole chromosomes or of chromosome segments from a species in which these characters were tied up together.

FIG. 21. Pictorialized diagram of 23 plants from a hybrid colony studied by Riley (see Plate 1). Diagrammed from his data according to the symbols of Fig. 23. The upper-left-hand star-shaped dot represents the hypothetical species responsible for the introgression, as determined by the "method of extrapolated correlates." Further discussion in the text.

From hybrid population H-2 there are indications that these characters are so correlated. By diagramming similarly the other hybrid population H-1 (Fig. 22) in the same way we can demonstrate that these correlations hold for it and are even more strongly apparent there.

FIG. 22. Pictorialized diagram of Hybrid Colony H-1 of Plate 1, plotted from Riley's data, using the symbols of Fig. 23.

Having demonstrated the repeated existence of these complex correlations, we now proceed on the hypothesis that they are the result of introgression from a species in which all these characters were united. We can, therefore, extrapolate our data on the correlates in the hybrid population and produce a conception of what species would have been required to create such an effect. Population H-2 was very similar to HGC on the whole, and even H-1 bore a strong resemblance to it. Therefore, we need to imagine what kind of iris when crossed with HGC would yield such variants. If it produced reddish blue descendants in its cross with HGC, then it must have been redder still. If it produced small flowers in combination with HGC, then it must itself have had very small flowers. In this way we may extrapolate character by character from HGC to the hybrid to the other putative species. It would have had to have been an iris with very narrow, red petals, strongly exserted stamens, a yellow tube, no crest, and small stylar appendages.

FIG. 23. Within lower-right-hand box are the symbols used in all the pictorialized scatter diagrams of Figs. 20 to 23. Upper left: 23 plants of Iris fulva, plotted from Riley's data. Note the exact correspondence with the predictions of Fig. 21.

Such a species having been predicted, if we can find exactly such a one in this same area, its very existence will constitute strong evidence for the suspected hybridization. Our hypothetical introgressant, of course, proved to be Fulva. The diagram of its population plotted from Riley's data (Fig. 23) agrees exactly with our extrapolations. A series of such predictions successfully made forms almost indisputable evidence for the validity of the method of extrapolated correlates and confirms the hypothesis of introgression.

The ease of extrapolation will vary with the number of easily measured differences separating the species under observation. In a genus like Fraxinus, in which species are separated for the most part by vague and inconstant differences in texture, pubescence, etc., extrapolation will be difficult, though not impossible. The more closely related the entities involved and the more similar they are morphologically, the more difficult will it be to find differences that lend themselves to precise description and measurement. In time higher plants, however, with persistence, it has always proved possible to find suitable characters. It must be admitted that the techniques of putting such differences as leaf shape, leaf texture, and branching patterns into measurable form are still in the exploratory stage, but several that have been worked out for particular cases seem to be rather generally applicable. How far these methods can be used with other kinds of organisms it would be difficult to say. Because of the relatively simple nature of their development, plants exhibit their species differences in less complicated ways than does, for example, an insect wing or a vertebrate tooth.

In trying out such a method as that described above, one elementary fact is of great importance. If possible the work should be done in the field, at least in a preliminary way. By taking squared paper to the field it will often be possible to measure at least a few of the more obvious differences in a population and make a preliminary determination of what characters are tending to cohere in that population. As the cohering center is apprehended more and more closely, the sets of characters that go together will be more and more clearly seen. One will thus be able to collect those specimens and to concentrate on the study of those characters that are the most effective.

In interpreting and measuring the results of interspecific introgression, one of the most difficult and challenging problems is the effect of a few genes from one species when introduced into the genetic background of the other. The greater the morphological hiatus between the two hybridizing entities, the more difficult does it become to predict the impact of such a recombination or to interpret it after it has been observed. One can comparatively easily estimate the probable outcome of crossing one inbred line of maize with another and then backcrossing one or two times to the original line. It takes more experience to suggest what might be the result of such an operation upon well-differentiated species. When totally different genera (such as Zea and Tripsacum) may be concerned, the possible effects of introgression of either into the other is a research problem of no mean dimensions. One may have studied genetics for a lifetime and still be totally unable to answer the question "What would be the result of any one or two genes from Drosophila if they were introduced into Zea Mays?"

In introgression, what often seems at first sight to be the appearance of something totally new usually proves to be a recombination that one had not had the wit to anticipate. Hybridization ordinarily results not in the new, but in the unexpected. For example, brilliant-colored stems and leaves often appear when Tradescantia canaliculata suffers introgression from Tradescantia subaspera var. pilosa. Neither of these species has conspicuous plant color. Careful examination, however, shows that T. subaspera has a dull purple pigment in the epidermis—so dull that it gives the leaf and stem a general appearance of very dark green. T. canaliculata has very little color in the epidermis, but what there is has none of the dark purplish cast that characterizes T. subaspera. Introgression, therefore, brings some of the basic genes for colored epidermis into T. canaliculata, and when they operate there in the absence of the dark purple modifiers they produce a brilliant effect superficially quite different from anything in either species.

In the studies of introgression between these species it was not until after the artificial backcrosses had been made that we began to suspect the origin of the subaspera introgressants in T. canaliculata. These two species are strikingly different: T. canaliculata has a few long nodes, the uppermost of which are usually the longest. T. pilosa has many short nodes, and node length decreases progressively upwards. The introgressants of subaspera tend to have brilliant stems and leaves and a much higher node number than ordinary canaliculata. Though their nodes are somewhat shorter than in the latter, the extra number more than compensates, and the introgressants are frequently twice as tall as their unmongrelized sisters. These tallish, bright-stemmed canaliculata's superficially do not look at all like T. subaspera pilosa. It is only when careful studies are made of leaf shape, inflorescence characters, and pubescence that one finds that the whole complex in a greatly diluted form is tending to stay together in these peculiar variants.

After a few examples of introgression have been studied it is much easier to recognize introgression in other genera and in other families. With active introgression, the segregation of whole chromosomes and of chromosome segments produces an overall effect on the variability of the population which, though difficult to describe, is almost unmistakable to those who have learned what it signifies. In such a population several different characters will be varying and recombining to a degree so far beyond what happens without introgression that it is of another order of magnitude. Those who have pioneered in the analysis of introgression are sometimes accused of "seeing hybrids under every bush." The truth of the matter is that, in certain groups of plants and animals, the results of hybridization are more widespread than had previously been suspected by most biologists and that the morphological effects of hybridization upon population variability are of a peculiar sort. With a little practice these peculiarities can often be recognized, even in families of plants and in floras with which the investigator is unfamiliar. By methods like those outlined above, it is possible to apply a series of critical tests to such a varying population and make valid estimates of introgression.


How important is introgressive hybridization? I do not know. One point seems fairly certain: its importance is paradoxical. The more imperceptible introgression becomes, the greater is its biological significance. It may be of the greatest fundamental importance when by our present crude methods we can do no more than to demonstrate its existence. When, on the other hand, it leads to bizarre hybrid swarms, apparent even to the casual passer-by, it may be of little general significance. When, as described in Woodson's studies of Asclepias populations, it produces clines reaching a third of the way across a continent, it is scarcely perceptible in any one locality. Only by the exact comparisons of populations can we demonstrate the phenomenon, yet in such populations the raw material for evolution brought in by introgression must greatly exceed the new genes produced directly by mutation. The wider spread of a few genes (if it exists) might well be imperceptible even from a study of population averages, but it would be of tremendous biological import. Germplasms are proteins, strange and complex substances. The introduction of a single alien gene into a new germplasm would be the introduction of one new unit into a gigantic protein complex. Reasoning purely from chemical facts, we might expect such a mixture to have secondary consequences in addition to its primary ones. But even were there no secondary consequences, the wide dispersal of introgressive genes (perceptible only to the most exquisitely precise techniques) would be a phenomenon of fundamental importance. Hence our paradox. Introgression is of the greater biological significance, the less is the impact apparent to casual inspection.