Consider a biologically diverse environment such as a forested valley on Kauai, a shallow-water shelf along the shores of Lake Victoria, or a cypress swamp in northern Florida. Some of the resident species are individually specialized for narrow niches and limited to small geographical ranges. Their dispersal ability is poor, their close relatives few or nonexistent. Their vertical evolution creeps along, and speciation is stalled. They have no geographic races and little prospect for multiplication. At the opposite extreme, other species possess flexible food habits and are excellent dispersers. They form new populations readily and evolve quickly into new niches, from diet to habitat to season of activity. Their potential for diversification is high, and they pile up species in the same localities by repeated cycles of dispersal and reinvasion.

In focusing on this last group of actively evolving populations, we are most likely to encounter all the stages of geographical speciation as interpreted by prevailing theory. At the earliest stage, the population is spread continuously across its range, and all the organisms in it freely interbreed. Few if any differences occur from one end of the range to the other. At the next stage, the population is still continuously distributed but divided into subspecies. Where the subspecies meet, they freely crossbreed. Imagine such a race of butterflies in Texas with large spots on their wings, and another in Mississippi lacking spots. Where the two subspecies meet in Louisiana and interbreed, the butterflies have spots of intermediate size. Those near Texas have larger spots, tending toward full Texas size; those near Mississippi have very small spots, tending toward the unadorned Mississippi condition.

Time passes, and at a more advanced stage the subspecies are still able to interbreed freely if they meet, but by now they have diverged in many genetic qualities. Butterflies from the same populations might differ not only in wing pattern but also in size, preferred food plant, growth rate of the caterpillars, and so on through any combination of hundreds of traits subject to genetic variation. The divergence of the subspecies will be hastened if some physical barrier, such as a broad river or dry grassland corridor, separates the two populations and restricts the flow of genes between them.

Finally, the two populations have diverged so far that they do not interbreed when they meet. They have been transformed into full-fledged biological species. Our two butterfly species now coexist in Louisiana, kept apart by differences in breeding season, courtship behavior, or some other intrinsic isolating mechanism singly or in combination - an inherited failure, in other words, to mesh their reproductive activity. Few if any small-spotted hybrids occur in the zone of overlap.

This classical model of geographical speciation makes a tidy picture. It has a core of truth, but real evolution is much messier. In fact, evolution is so messy that a faithful description of real cases converts the science into natural history, in which unique details are as important as the principles by which they are explained.

Consider the subspecies. The category seems an inevitable intermediate step in an Aristotelian progression running from no subspecies to subspecies to species. What exactly is a subspecies? The textbooks define it as a geographical race, a population with distinctive traits occupying part of the range of the species.

What then is a population? We are in immediate trouble. It is easy to say that a clearly defined population, one recognizable by everyone at a glance, occupies an exclusive part of the range of a species. And geneticists like to add, for purposes of mathematical clarity but not as an absolute requirement, that the population is a “deme”: its members interbreed at random, and any member is equally likely to mate with any other member in the population, regardless of its location.

Few such objectively definable populations exist in nature. Most that do look like textbook examples are endangered species, with so few organisms left that there is no doubt as to the boundaries of the population they compose. The last surviving ivory-billed woodpeckers, found in one mountain forest of eastern Cuba, belong to this parlous category. So do the Devil’s Hole pupfish, barely hanging on in a tiny desert spring at Ash Meadows, Nevada. You can stand at the entrance to Devil’s Hole, look down 15 meters to where the water laps over a sunlit ledge, and see the entire species swimming around like goldfish in a bowl.

Most species are not confined this stringently, which is fortunate for conservation and unfortunate for textbook theory. Take the redbacked salamander (Plethodon cinereus), one of the most widespread and abundant salamanders of North America. The species ranges from Nova Scotia and Ontario south to Georgia and Louisiana. Redbacks occur almost continuously through the northern three quarters of this range. It is tempting to classify the whole northern ensemble as one huge population. Salamander taxonomists do just that and call it a subspecies, Plethodon cinereus cinereus (a formal rule of nomenclature: to designate a subspecies, add a third name). But redbacks are far from continuously distributed. They are largely confined to moist lowland forest, which is not a continuous habitat but an irregularly broken filigree laid on the land. Even within habitable forest, the population is divided into local aggregates that slowly expand, contract, and reform into new configurations across the generations. The rate of interbreeding among local demes in forested valleys and woodlots is unresearched and unknown. In short, with more data biologists might be able to distinguish thousands of populations across the vast range of P. cinereus cinereus. A diligent taxonomist might legitimately break the one formally recognized subspecies into large numbers of subspecies with smaller ranges.

To the south, in the mountains of northern Georgia and Alabama, there is another generally recognized subspecies, Plethodon cinereus polycentratus, separated from P. cinereus cinereus by 80 kilometers of redback-free terrain. A third subspecies, P. cinereus serratus, occurs in several widely separated localities in the hill country of Arkansas, Oklahoma, and Louisiana. These two additional races offer the same difficulties as the main northern subspecies. Their triple names are a convenient shorthand, the statement of a rough truth. The classification works so long as we recognize that dicing up the whole species geographically is imprecise and to a large degree arbitrary. Depending on the criteria used, there could be one subspecies of P. cinereus, or there could be hundreds.

An even more fundamental difficulty of subspecies is the discordance of the traits by which the subspecies are defined. Suppose that we agreed to ignore the population problem for the moment. Imagine that easily definable populations exist in an idealized species (which I will continue to call the redback salamander for clarity). One species comprises thousands of small populations across North America. Individuals from the southern half, Georgia to Virginia, have stripes over most of the body; those from the northern half, Maryland to Canada, lack stripes. On the basis of this one character, there are two subspecies, two geographic races: striped southern redbacks and plain northern redbacks. We notice, however, that western individuals of the species are larger. These two characters, stripedness and size, are obviously discordant-they break along different geographic lines. They can be used to define four subspecies: big striped in the southwest, little striped in the southeast, little unstriped in the northeast, and big unstriped in the northwest. Next we find that the eyes of juveniles are amber southwest of a line running from the Great Lakes to Georgia and yellow to the northeast of the line. Two more subspe- cies are added to produce a total of six overall. Looking still more closely we find...

Here is the point of this exercise in geometry: most traits varying geographically in a given species are discordant. They change at different places and in different directions. It follows that subspecies are recognized according to whatever traits taxonomists choose to study. It also follows that the greater the number of traits, the larger the number of the subspecies that must be recognized.

The uncertainty of the limits of populations combined with the discordance of traits means that the subspecies is an arbitrary unit of classification. That uncertainty is reflected in the confusion over human races. In past years anthropologists struggled hopelessly with attempts to define human races. Estimates of the number of races made by researchers during the 1950s ranged from six to more than sixty. The variation in numbers is due precisely to the fact that Homo sapiens is a typical evolving species.

Anthropologists, like biologists, have now largely abandoned the formal subspecies concept. They prefer a convenient shorthand to designate a certain part of a population with reference to one or two traits. They say, for example, “northern Asians tend to have more prominent canthal folds,” knowing full well that canthal folds differ in geography from blood types, which deviate in turn from average height, lactose intolerance, Tay-Sachs syndrome, eye color, hair structure, infant passivity, and so on through hundreds of other more or less discordant traits prescribed or at least influenced by 200,000 or so human genes scattered through 46 chromosomes. The emphasis in research in anthropology and biology has passed from the description of subspecies to the analysis of the geography of separate traits and their respective contributions, singly or in combination, to survival and reproduction.

The demotion of the subspecies should carry with it a word of caution, in the service of moderation. Real populations do exist, however difficult to define. Genetic traits still vary. It may be artificial to divide up and label redback salamanders from the southern United States as subspecies, but they nevertheless differ in many genetic traits and compose a reservoir of unique genes. It is further true that some populations of widespread animals and plants are sufficiently isolated and genetically distinct to compose objective subspecies even in the abstract textbook sense. It is useful to label such populations formally as subspecies. Stephen O’Brien and Ernst Mayr, for example, have proposed guidelines to that end for use by conservation biologists and policy makers. They suggest that subspecies be defined as individuals occupying a particular part of the range of the species, with genes and natural history distinct from those of other subspecies. Members of different subspecies can freely interbreed. They can arise either as populations adapting to local conditions or as hybrids between subspecies.

The delimitation of subspecies, an occasional bureaucratic necessity when the U.S. Endangered Species Act or its equivalent is invoked, will usually be difficult and even controversial. Evolution, to repeat, is messy. The Florida panther offers a case in point. Once the panther, also known as the mountain lion or cougar, occurred throughout the southern United States. Now it is down to about fifty individuals in southern Florida, the subspecies Felis concolor coryi. Biochemical tests have revealed that this tiny population was derived from two stocks: the final survivors of the original North American panthers that once roamed Florida, and seven animals of mixed North American and South American origin released from captivity into the Everglades between 1957 and 1967. The present population is thus of hybrid origin, but it contains a unique ensemble of genes of partial North American origin and deserves protection as a native mammal.

The ambiguity of the subspecies as a taxonomic unit creates an interesting dilemma in evolutionary reasoning. We have before us an idealized sequence that starts with a geographically isolated population still identical to other populations of the same species. The population then evolves into a subspecies, still capable of interbreeding with the other populations, if they could somehow breach the geographical barrier and meet along the edges of their ranges. The subspecies finally evolves into a full species, meaning that if it meets the other populations it would no longer freely interbreed with them. The dilemma is this: if subspecies are usually amorphous and cannot be defined by a single objective criterion, how can such an arbitrary unit give rise to the species, which is sharply defined and objective?

The answer to the puzzle tells us a great deal about the origin of diversity. In order to spring forth as a species, a group of breeding individuals need only acquire one difference in one trait in their biology. This difference, the innate isolating mechanism, prevents them from freely interbreeding with other groups. It does not matter if the limits of the populations as a whole are poorly marked. Nor does it matter if all the other traits vary in a crazy-quilt manner across the populations that are splitting off as species. What counts is that somehow a group of individuals occupying some part of the total range evolves a different sex attractant, nuptial dance, mating season, or any other hereditary trait that prevents them from freely interbreeding with other populations. When that happens, a new species is born. The truly objective unit, the closed-gene pool of future generations, is the group of individual organisms that acquires the isolating trait. This new species can be defined by a single isolating trait. Other characters that vary geographically-hairiness, color, cold hardiness, whatever—can show any pattern of geographical variation whatever, either concordant with the isolating trait or totally different, without changing the outcome. Once segregated in this manner, the species will inevitably evolve away from other species, becoming ever more different in a steadily enlarging suite of traits as time passes.

- Edward O. Wilson, 1992.  The Diversity of Life.  Harvard University Press, Cambridge, MA. pp 63-69.  (Buy this book.)