This optimistic notion—that diligence in selection can produce almost any desired trait by artificial selection of domesticated animals or cultivated plants—has inspired the customary extrapolation into nature’s larger scales, leading to a conclusion that natural selection must work even more inexorably to hone wild creatures to a state of optimal design. As Darwin wrote:
How fleeting are the wishes and efforts of man! how short his time! and consequently how poor will his products be compared with those accumulated by nature during whole geological periods. Can we wonder, then, that nature’s productions…should be infinitely better adapted to the most complex conditions of life, and should plainly bear the stamp of far higher workmanship. It may be said that natural selection is daily and hourly scrutinizing, throughout the world, every variation, even the slightest; rejecting that which is bad, preserving and adding up all that is good; silently and insensibly working.
This common claim for organic optimality cannot be reconciled with a theme that I regard as the primary message of history—the lesson of the panda’s thumb and the flamingo’s smile: The quirky hold of history lies recorded in oddities and imperfections that reveal pathways of descent. The allure of perfection speaks more to our cultural habits and instructional needs than to nature’s ways (good design inspires wonder and provides excellent material for boxed illustrations in textbooks). Optimality in complex structure would probably bring evolution to a grinding halt, as flexibility disappeared on the altar of intricate adaptation (how might we change a peacock for different environments of its unknown future?).
In any case, leaving aside the abstractions of how nature ought to work, we have abundant empirical evidence that enormous effort in husbandry does not always bring its desired reward. Poultrymen have never broken the “egg-a-day barrier” (no breed of hen consistently lays more than one egg each day), and we are now trying to produce frost-resistant plants by introducing foreign DNA with techniques of genetic engineering because we have not been able to develop such traits by selection upon the natural variation of these plants. We do not know whether such failures represent our own stupidity or lack of sufficient diligence (or time) or whether they record intrinsic structural and genetic limits upon the power of selection. In any case, selection, either natural or artificial, is not the agent of organic optimality that our newspapers and textbooks so often portray. Limits are as powerful and interesting a theme as engineering triumph.
Francis Galton himself, in the same book that promised so much for human futures by controlled breeding, presented our most incisive metaphor for the other side of the coin—the limits to improvement imposed by inherited form and function. (Darwin was also intrigued by the subject of limits and devoted as much attention to this aspect of growth and development as to natural selection itself—see his longest book, the two-volume Variation in Animals and Plants Under Domestication, 1868.) Following the optimistic notion of unrestricted molding, we might view an organism as a billiard ball lying on a smooth table. The pool cue of natural selection pushes the ball wherever environmental pressure or human intent dictates. The speed and direction of motion (evolutionary changes) are controlled by the external force of selection. The organism, in short, does not push back. Evolution is a one-way street; outside pushes inside.
But suppose, Galton argues, organisms are not passive spheres but polyhedrons resting upon stable facets.
The changes are not by insensible gradations; there are many, but not an infinite number of intermediate links…. The mechanical conception would be that of a rough stone, having, in consequence of its roughness, a vast number of natural facets, on any one of which it might rest in “stable” equilibrium…. If by a powerful effort the stone is compelled to overpass the limits of the facet on which it has hitherto found rest, it will tumble over into a new position of stability…. The stone…can only repose in certain widely separated positions.
Galton proposes no new force. The polyhedral stone will not move at all unless natural selection pushes hard. But the polyhedron’s response to selection is restricted by its own internal structure; it can only move to a limited number of definite places. Thus, following the metaphor of Galton’s polyhedron to its conclusion, the actual directions of evolutionary change record a dynamic interaction of external push and internal constraint. The constraints are not merely negative limits to Panglossian possibilities, but active participants in the pathways of evolutionary change. St. George Mivart, whom Darwin acknowledged as his most worthy critic, adopted Galton’s polyhedron as the basis of his argument and wrote (1871):
The existence of internal conditions in animals corresponding with such facets is denied by pure Darwinians…. The internal tendency of an organism to certain considerable and definite changes would correspond to the facets on the surface of the spheroid.
If Galton’s polyhedron ranks as more than mere verbiage, then we must be able to map the facets of genetic and developmental possibility. We must recast our picture of evolution as an interaction of outside (selection) and inside (constraint), not as an untrammeled trajectory toward greater adaptation. We can find no better subject for investigating facets than Darwin’s own prototype for evolutionary arguments—changes induced in historical time through conscious selection by breeders upon domesticated animals. I can imagine no better object than our proverbial best friend—the dog—Galton’s own choice for comparison in the very first sentence of his manifesto for human improvement.
We should begin by asking why dogs, cows, and pigs, rather than zebras, seals, and hippos are among our domesticated animals? Are all creatures malleable to our tastes and needs, and do our selections therefore reflect the best possible beef and service? Or do some of the strongest and tastiest not enter our orbit because selection cannot overcome inherited features of form or behavior that evolved in other contexts and now resist any recruitment to human purposes?
From the first—or at least since Western traditions abandoned the idea that God had designed creatures explicitly for human use—biologists have recognized that only certain forms of behavior predispose animals to domestication and that our successes represent a subset of available species, not by any means an optimally chosen few amidst unlimited potential. In particular, we have recruited our domestic animals from social species with hierarchies of rank and domination. In the basic “trick” of domestication—what we call “taming” in our vernacular—we learn the animal’s own cues and signals, thus assuming the status of a dominant creature within the animal’s own species. We tame creatures by subverting their own natural behavior. If animals do not manifest a basic sociability and propensity to submit under proper cues, then we have not been able to domesticate them, whatever their potential as food or beast of burden. As Charles Lyell wrote in 1832:
Unless some animals had manifested in a wild state an aptitude to second the efforts of man, their domestication would never have been attempted…. It conforms itself to the will of man, because it had a chief to which in a wild state it would have yielded obedience…it makes no sacrifice of its natural inclinations…. No solitary species…has yet afforded true domestic races. We merely develop to our own advantage propensities which propel the individuals of certain species to draw near to their fellows.
The dog is our primary pet because its ancestor, the wolf Canis lupus, had evolved behaviors that, by a fortunate accident of history, included a predisposition for human companionship. Thus, our story begins with a push onto a facet of Galton’s polyhedron. Domestication required a preexisting structure of behavior.
We might readily admit this prerequisite, yet marvel at the stunning diversity of domestic breeds and conclude that any shape or habit might be modeled from the basic wolf prototype. We would be wrong again.
We can usually formulate “big” questions easily enough; the key to good science lies in our ability to translate such ideas into palpable data that can help us to decide one way or the other. We can readily state the issue of li
mits versus optimality, but how shall we test it? In most cases, we approach such generalities best by isolating a small corner that can be defined and assessed with precision. This tactic often disappoints nonscientists, for they feel that we are being paltry or meanspirited in focusing so narrowly on one particular; yet I would rather tackle a well-defined iota, so long as I might then add further bits on the path to omega, than meet a great issue head-on in such ill-formed complexity that I could only waffle or pontificate about the grand and intangible.
A standard strategy for the study of limits lies in the field of allometry, or changes in shape associated with variation in size. Two sequences of size differences might be important for studying variation in form among breeds of domestic dogs: ontogeny, or changes in shape that occur during growth of individual dogs from fetus to adult; and interspecific scaling, or differences in shape among adults of varying sizes within the family Canidae, from small foxes to large wolves. We might search for regularities in the relationship between size and shape in these two sequences and then ask whether variation among dog breeds follows or transcends these patterns. If, for all their stunning diversity, dogs of different breeds end up with shapes predicted for their size by the ontogenetic or interspecific series, then inherited patterns of growth and history constrain current selection along channels of preferred form. Growth and previous evolution will act as facets of Galton’s polyhedron, favored positions imposed from within upon the efforts of breeders.
The biological literature includes a large but obscure series of articles (mostly auf Deutsch), dating to the early years of this century, on allometry in domestic breeds. These themes have been neglected by English and American evolutionists during the past thirty years, primarily because an overconfident, strict Darwinism had so strongly emphasized the power of adaptation that the older subject of limits lost its appeal. But exciting progress in our understanding of genetic architecture and embryological development has begun to strike a proper balance between the external strength of selection and the internal channels of inherited structure. I sense a welcome reappearance of Galton’s polyhedron in the primary technical literature of evolutionary biology. As one example, consider a recent study of ontogenetic and interspecific allometry in dogs by Robert K. Wayne.
Wayne asks how inherited patterns of allometry might constrain the variety of domestic breeds. He finds, for example, that all measures of skull length (face, jaws, and cranium) show little variation in three senses: First, the ontogenetic and interspecific patterns are similar (baby dogs look like small foxes in the portions of length elements); second, we note little change of shape as size increases (baby dogs are like old dogs, and small foxes are like large wolves); third, we find little variation among breeds or species at any common size (all young dogs of the same size have roughly equal length elements).
These three observations suggest that natural variation among canids offers little raw material for fanciers to create breeds with exotic skull lengths. Wayne has confirmed this suspicion by noting that few adults of different breeds, from toy poodle to Great Dane, depart far from the tight relationship predicted by ontogeny or interspecies scaling: The length elements of a small breed may be predicted from the proportions of puppies in larger breeds or from small adult foxes.
Wayne points out that the criteria of artificial selection in domesticated races (the quirky human preferences imposed upon toy or fancy breeds, for example) must differ dramatically from the basis of natural selection in wild species—“the dog’s ability to catch, dismember, or masticate live prey.” If length elements are so constant in such radically different contexts, then their invariance probably reflects an intrinsic limit on variability rather than a fortuitous concurrence in different circumstances. Wayne concludes:
Despite considerable variability in the time, place and conditions of origination of dog breeds, the scaling of skull-length measurement components is relatively invariant. All [small] dog breeds are exact allometric dwarfs with respect to measures of skull length. It is unlikely that such a specific morphological relationship has been the direct result of selection by breeders. Rather, a lack of developmental variation seems a better explanation.
When we turn to skull widths, however, we note variation where lengths showed constancy: Puppies differ from adult foxes of the same size; puppies also turn into dogs of greatly altered shape, and small foxes are easily distinguished from large wolves by proportions of width elements. The material available to dog breeders should be extensive.
Wayne finds that dog breeds do vary greatly in width elements. (We might be tempted to say, “So what; doesn’t everyone know this from a lifetime of casual inspection?” Yet our intuitions are often faulty. Wayne shows that small, snubnosed breeds have wide faces, not short skulls or jaws. Readers might be chuckling and saying, what’s the difference—doesn’t overwide amount to the same thing as undershort, as in the fat man’s riposte that he is only too short for his weight? But the statements are not equivalent, for we are comparing lengths and widths with a common standard of body size. Small breeds are the right length, but unusually wide, for their size.) The great variation among dog breeds is not uniform among all parts, but concentrated in those features that supply raw material in growth and evolutionary history. Dog breeds form along permitted paths of available variation.
This study of internal potential helps us to predict which features will form the basis for variety among breeds (not simply what the selector wishes, but what the selectee can provide). But we can extend this insight much further to encompass the great differences in variety among domesticated species. If some features of dogs differ more among breeds because internal factors must supply the requisite variation, then perhaps, by extension, some entire species develop more diversity, not because human selectors have been more assiduous or because human needs require such variety, but because the internal facets of Galton’s polyhedron are many, closely spaced, and varied.
Why do breeds of dogs differ so greatly, and those of cats relatively little? (Cats vary widely in color and character of coat but not much in shape. Nothing in the world of felines can approach the disparity in skull form between a stubby Pekingese and an elongated borzoi.) Before we speculate about diminished human effort or desire to explain Garfield versus Lassie, we should consider the more fundamental fact that available variation in the ontogeny and interspecific scaling among cats offers very little for breeders to select. Lions differ from tabbies far less than large from small dogs while, more importantly, kittens grow to adult cats with only a fraction of the change in shape that accompanies the transformation of puppy to grown dog. Consider the accompanying figure taken from Wayne’s article and comparing, at the same size, neonates and adults of cats and dogs. Dogs have contributed to their own flexibility; cats, as ever, are recalcitrant.
Wayne makes a persuasive case that comparative diversity among domesticated species depends more upon available variation in the growth of wild ancestors than in the extent of human efforts. Horses change relatively little in shape during growth, and the heads of Shetland ponies do not differ much from those of the largest workhorses. Pigs, on the other hand, are second only to dogs in diversity of breeds. They are also unmatched among farm animals for marked change of shape during growth.
Skulls of dogs (upper row) and cats (lower row) show the differences between neonates (left) and adult animals (right). Note the much greater range of change in dogs vs. cats. Evolution 40 (2), 1986.
Amounts of intrinsic variation therefore set limits and supply possibilities to breeders. But even the most variable of wild species do not become putty in the hands of breeders. Pigs and dogs vary in definite ways during their growth, and only certain shapes are available for selection at definite sizes. Wayne’s most persuasive case for internal limits lies in his demonstration that sets of traits in a standard “ontogenetic trajectory” (a sequence of stages from puppy to adult) tend to hang together. Dog breeds are not a hodgepodge of
isolated traits, each taken at will from any stage of ontogeny. Traits of juvenility remain associated, and many breeds, particularly among small dogs, continue to look like puppies when adult—an evolutionary process called paedomorphosis (child-shaped) or neoteny (literally, “holding on to youth”).
Wayne has shown that—without exception—adults of small breeds resemble the juvenile stages of large dogs more than the adults of other wild canid species (small foxes)—a convincing demonstration that inherited patterns of growth set possibilities of change. Dogs resolutely stick to their own trajectories of growth. “To some extent,” Wayne concludes, “many dog breeds represent morphological snapshots between these developmental endpoints…. This suggests that small domestic dogs differ from foxes because puppies of small dogs cannot grow out of their distinctive neonate morphology.”
We know, of course, that breeders can do many wonderful and peculiar things, from making a dachshund into a frankfurter to turning a chihuahua into a hairless rat or a sheepdog into a woolly mimic of its charges. But these peculiarities are imposed upon a basic and unaltered pattern set by constraints of inherited growth. The trajectory of ontogeny provides, as Raymond Coppinger states, a “rough first draft” for all breeds.
If ordinary variation in growth provides the main source for breeders, then a wild species’ own juvenile stages are the primary storehouse of available change. Under this basic theme of limits, we may understand an old and otherwise puzzling observation about domesticated versus wild species. Over and over again, we note that domestic species develop more juvenile proportions than their wild ancestors. We cannot explain this difference by smaller size (since domestic breeds are often larger than their ancestors) or by conscious selection on the old theme of planned optimality, for what possible common adaptive advantage could have inspired breeders to produce such a similarly shortened face in Middle white pigs, the Niatu oxen of South America, and the Pekingese of the Chinese imperial court (see figure). The only sensible coordinating theme behind these similarities is retention by neoteny of juvenile traits common to most vertebrates.