Hen’s Teeth and Horse’s Toes

by Stephen Jay Gould

  In October 1981, virtually all leading experts on extraterrestrial impacts, iridium spikes, paleontological evidence, etc., met in Snowbird, Utah, to discuss both the Alvarez hypothesis and the general issue of the role of extraterrestrial catastrophes in geological history. The proceedings of this conference will be published by the Geological Society of America in late 1982 or early 1983. This volume will be an important source for any reader who wishes to pursue this fascinating subject further.

  26 | Chance Riches

  IN LITERATURE, the idea of randomness is often coupled with chaos, lawlessness, or disorder to conjure up a vision of ultimate terror. In the Essay on Man, Alexander Pope evokes the image of an entire universe falling apart if its lawful order should ever be broken.

  Let earth unbalanc’d from her orbit fly,

  Planets and suns run lawless thro’ the sky;

  Let ruling angels from their spheres be hurl’d,

  Being on being wreck’d, and world on world;

  Heav’n’s whole foundations to their center nod,

  And nature tremble to the throne of God.

  Pope’s facile and hopeful solution simply banished the very idea of chance by a decree of faith:

  All nature is but art unknown to thee;

  All chance, direction which thou canst not see;

  All discord, harmony not understood;

  All partial evil, universal good.

  Chance also got Darwin into trouble more than 100 years later when he granted it an important place in his evolutionary theory. Many critics, impelled by a knee-jerk negativism toward randomness, made no attempt to understand the strictly limited role that Darwin had awarded it. They argued: Darwin must be wrong; nature is so harmonious, animals are so well designed. This order cannot be the work of chance.

  Darwin would not have disagreed. He constructed his theory in two parts, letting chance prevail in the first but strictly excluding it from the second for the conventional reason that chance could not, in his opinion, yield the order so prevalent in our world. In Darwin’s theory, populations must first develop a large amount of heritable variation to provide raw material for the later, directing action of natural selection. Darwin viewed this pool of variation as random with respect to the direction of adaptive change—that is, if a species would be better off at smaller size, variation tends to arise with equal frequency at sizes both larger and smaller than the current average. The raw material for evolutionary change—and the raw material only—arises by a process of random mutation.

  Natural selection then enters for the second part, and it acts as a conventional, deterministic, directing force. Natural selection fashions the raw material of variation by preserving and fostering individuals that vary in adaptive directions. In our hypothetical population, small organisms will, on average, rear more successful offspring. Slowly, but inexorably, the average size of individuals in the population will decline.

  Darwin expressed himself with admirable clarity, yet critics have been misunderstanding this fundamental point for more than a century, from the Rev. Adam Sedgwick (Darwin’s geological mentor, not a dogmatic and antiscientific theologian) to Arthur Koestler. The litany is ever the same: Darwin must be wrong; order cannot arise from chance. Again, Darwin never said that it could. Chance only produces raw material; natural selection directs evolutionary change.

  Evolutionary theory is now stirring from the strict Darwinism that has prevailed during the past thirty years or so. While critics have not seriously challenged Darwin’s mechanism on its primary turf of explaining adaptation, they have rallied around a claim for pluralism. Must all evolutionary change be viewed as adaptation and ascribed to natural selection? Randomness has become a central focus for critics because Darwin’s strict dichotomy seems to be breaking. Randomness may not act only in generating variation; it may be an important agent of evolutionary change as well. The specter of chance is now truly intruding where Darwin’s critics had falsely detected it before. Given both the surpassingly poor reputation of randomness in general, and the specific Darwinian tradition of limiting its role to the production of raw material only, this development in evolutionary theory is both exciting and, to many, distressing.

  I should enter some disclaimers, if only to reduce disturbance. Randomness is making its bid as an agent of evolutionary change, but it is not threatening natural selection in the realm of adaptation. The beauty and aerodynamic efficiency of a bird’s wing, the grace and good design of a fish’s fins are not lucky accidents. Also, I use the specific meaning that “random” has long maintained in evolutionary theory: to describe changes that arise with no determined orientation. I do not use it as a general metaphor for chaos, disorder, or incomprehensibility (more on this later).

  Evolution operates at three major levels: populations change as certain genes become more or less common because individuals carrying them have more or less success in rearing offspring; new species arise by the splitting of descendant populations from their ancestors; and evolutionary trends occur because some species are more successful than others in branching and persisting. Randomness is challenging the determinism of natural selection as a cause of evolutionary change at all three levels.

  The genetic structure of populations. When natural selection operates in its usual way, genetic variation is reduced: the fit arise, in part, by the elimination of the unfit. The total amount of genetic variation in a population should represent a balance between the addition of new variation by mutation and the removal of unfit variants by natural selection. Since we have stores of data on rates of mutation and selection, we may predict the upper limit of variation that a population can maintain if selection acts upon all genes.

  Techniques for measuring the amount of genetic variation in natural populations have been available only for the past fifteen years. Their first and primary result came as a surprise to many geneticists: most populations maintain too much variation to support the usual claim that all genes are scrutinized by natural selection.

  To be sure, natural selection does not always eliminate. In some cases, it may act to increase or maintain genetic variation. The most common mode of maintenance is called “heterozygote advantage.” Suppose that a gene exists in two forms, a dominant A and a recessive a. In sexual species, each individual has two copies of the gene, one from each parent. Now suppose that the mixture, or so-called heterozygote Aa, has a selective advantage over either pure form, the double dominant AA, or the double recessive aa. In this case, selection will preserve both A and a by favoring the heterozygote Aa individuals.

  But even these modes of preservation have their limits, and many geneticists feel that populations still maintain too much variation for selective control. If they are right (and the issue remains under intense debate), then we must face the possibility that many genes remain in populations because selection cannot “see” them, and therefore cannot either mark them for elimination or remove other variants by favoring them. In other words, many genes may be neutral. They may be invisible to natural selection and their increase or decrease may be a result of chance alone.

  Since “change of gene frequencies in populations” is the “official” definition of evolution, randomness has transgressed Darwin’s border and asserted itself as an agent of evolutionary change. (This process of random increase or decrease of frequency is called “genetic drift.” Contemporary Darwinism has always recognized drift, but has proclaimed it an infrequent and unimportant process, mostly confined to tiny populations with little chance of evolutionary persistence. The newer theory of neutralism suggests that many, if not most, genes in large populations owe their frequency primarily to random factors.)

  The origin of species. Species are defined as populations that are reproductively isolated from all others. Placed in contact with other populations, a true species will maintain itself as a separate evolutionary entity and will not amalgamate with another population by hybridization. The key question for the origin of new sp
ecies then becomes: how does reproductive isolation evolve?

  In the traditional view, an ancestral population is split by a geographical barrier (continents may drift apart, mountain ranges rise, or rivers alter their course, for example). The two descendant populations then evolve by natural selection to fit the local environments of their separated places. In due time the populations become so different that they will no longer be able to interbreed, should they reestablish contact. Reproductive isolation is a by-product of adaptive evolution by natural selection.

  During the past decade, the predominance of this mode has been challenged by a variety of new proposals advocating an interesting twist or reversal of perspective. They all argue that reproductive isolation can arise rapidly as a result of historical accidents with no selective significance at all. In this case, reproductive isolation comes first. By establishing new and discrete units, it provides an opportunity for selection to work. The ultimate success of such a species may depend upon the later development of selected traits, but the act of speciation itself may be a random event.

  Consider, for example, the process called chromosomal speciation. Taxonomists have discovered that many groups of closely related species do not differ much in form, behavior, or even general genetic composition. But they do display outstanding differences in the number and form of chromosomes, and these differences produce the reproductive isolation that maintains them as distinct species. An obvious, and old, hypothesis suggests that each new species arises when a major chromosomal change occurs by accident and manages to establish a new population.

  But this hypothesis suffers an obvious difficulty: the major change arises in a single individual. With whom shall it breed? The hybrid offspring of this mutant and a normal member of the population will almost surely be at a strong selective disadvantage. The mutation will therefore be swiftly eliminated.

  Recent work on the structure of populations indicates that some reasonably common forms of social organization might facilitate the origin of new species by rapid and accidental chromosomal change. If populations are “panmictic,” that is, if each female has an equal chance of breeding with each available male, then the chromosomal mutant cannot spread. But suppose that populations are subdivided into small groups of kin that breed exclusively with each other for several generations. Suppose also that these kin groups are harems, with a dominant male maintaining several females, and with brother—sister mating among the offspring.

  If a chromosomal mutation arises in the dominant male, all his children will carry both the mutant (from their single father) and the normal form (from their various mothers). They will probably be at a strong selective disadvantage relative to normal offspring in other harems. But this may not matter, especially if they have limited contact with other groups. The offspring may die at a higher rate, but we only require that a few survive to mate with their similarly afflicted siblings.

  One-fourth of these second generation offspring will be “pure” and contain two copies of the chromosomal mutation. If they can recognize and mate preferentially with each other in the next generation, all offspring will be pure mutants and members of a new species already (if they continue to mate with each other and if any hybrid offspring with normal forms suffers a strong selective penalty). The new chromosome is selectively neutral; it provides neither advantages nor disadvantages in itself. By establishing itself rapidly and accidentally in a small group, it permits the origin of a new species, again by chance. The new species may require substantial adaptive retooling for its subsequent survival, but this is a different, and later, matter.

  My colleague Guy Bush, of the University of Texas, tells me that horses provide, circumstantially at least, a strong case for chromosomal speciation. They all maintain the harem structure of kin breeding. Their seven living species (two horses, two asses, and three zebras) all look and act pretty much alike, despite some outstanding differences in external color and pattern. But their chromosome numbers differ greatly and surprisingly, from thirty-two in one of the zebras to sixty-six in that paradigm of the unpronounceable, Przewalski’s wild horse.

  Major patterns of rise and fall in the history of life. Many readers might be willing to accept chance at the lower levels. A bit of slop for some “invisible” genes within populations, an accidental species here and there. But surely conventional reasons rule the grand ebb and flow of major groups in the history of life. Trilobites must not be as good as the “advanced” arthropods (shrimps and their allies) that now populate our seas. The hordes of brachiopods from times past must have been pushed out by other creatures, clams in particular, that look like them but work better. Dinosaurs must have been bad at something that mammals can do. At this level, natural selection must reign and life must improve.

  The facts of mass extinction proclaim the fallacy of this argument. If groups slowly replaced other groups, some gaining in species over millions of years while others lost just as steadily, a scenario of selective control might seem irresistible. But most groups disappeared during the episodes of mass extinction that have punctuated the history of life. This fact is not news. It has been acknowledged for decades but explained away with an assumption that differential mortality must have a selective basis. The groups that roared (or even squeaked) through a holocaust must have survived for a reason. They were the tough guys, the good competitors. But some recent data on the extent of mass extinctions must call this comforting explanation into question.

  The granddaddy of all extinctions occurred about 225 million years ago at the end of the Permian period. (We don’t know why, although the coalescence of all continents at about the same time must have set the basic stage.) By wiping out many groups, permanently debilitating others, and allowing some to pass through relatively unscathed, this great dying set the fundamental pattern of life’s diversity ever since. But how profound was the Permian extinction? An old and familiar figure states that half the families of marine organisms (52.0 percent to be exact) perished at that time. But families are taxonomic abstractions. What does 52 percent of families mean for species, nature’s real units? (Inconsistent taxonomic practices and an inadequate fossil record prevent the counting of species directly. Families, as bigger units, are harder to miss.)

  We can be sure that a removal of half the families requires the death of a much greater percentage of species. A family is not gone until all its species die, and many families contain tens or hundreds of species. The extinction of most individual species does not wipe out a family, just as, for example, the random excision of a single entry in a telephone directory rarely removes the name entirely—you’d have to bump off a lot of Smiths. How many species must die before 50 percent of families are gone?

  David M. Raup of Chicago’s Field Museum has recently considered this question (see bibliography). The problem has no easy solution. If all families contained about the same number of species, then a simple formula would suffice. But variation is enormous. Many families contain only a single species. In this case, removal of the species also wipes out the family. Phone books contain their Zzyzzymanskis as well as their Wongs. Other families contain more than 100 species. We must know the empirical distribution of species per family before we can make a proper estimate. And we cannot construct an empirical distribution for Permian families because we cannot count the species directly.

  Raup therefore made tabulations for a group that we do know well, the echinoids, or sea urchins. Echinoids include 894 species distributed into 222 genera and 40 families. How many species must, on average, be removed at random in order to eliminate 52 percent of the families? Raup considered this question both empirically and theoretically and came up with the astounding figure of 96 percent. If the rest of life maintains a distribution of species within families similar to the echinoids—and we have no evidence for major differences in this pattern—then the Permian debacle might have wiped out all but 4 percent of species.

  Since estimates of living species in the l
ate Permian range from 45,000 to 240,000, a removal of 96 percent would leave but 1,800 to 9,600 species as guardians of life’s continuity. Moreover, as Raup argues, we have no strong evidence, despite intensive and specific search, of selectivity in the Permian extinctions. The debacle did not seem to favor any particular kind of animal—bigger creatures, inhabitants of shallow water, more complex forms, for example.

  I am not entirely persuaded by the 96 percent figure. Echinoids may not be a good model for all of life. More important, Raup is assuming that the 52 percent figure is not artificially inflated by biases in the fossil record. We know, for example, that late Permian marine sediments are relatively rare and we may be missing some successful families simply because so few late Permian fossils have been preserved. Nonetheless, even the most conservative figures indicate a removal of 80 to 85 percent of all species.