In the mid-forties, on the fourth floor of the American Museum of Natural History, there stood the remains of a tyrannosaurus. Towering above hordes of awestruck kids, this pile of bones inspired two of the best-known careers in twentieth-century science–that of a writer and that of a researcher. The most impressive thing about these careers, though, was that they were both pursued by the same person: Stephen Jay Gould.

Gould, who died this May at sixty, always dated his conversion to paleontology to his encounter with the dinosaur at age five. If this event marked the psychological genesis of his dual careers, their public debut followed three decades later, in 1977, when he published two books, one popular, “Ever Since Darwin,” and the other technical, “Ontogeny and Phylogeny.” Though both received good notices, it was, more than anything, their simultaneity that turned heads. Scientists do not as a rule burst upon the popular and the professional scenes at the same moment. (Far more often, as dotage approaches, a scientist will try to leverage a lifetime of scientific achievement into popular success.) The Times, in a rave review, made much of Gould's double debut, and he was off and running.

Gould came to real fame through his writings on evolution for Natural History. For a quarter century, he produced three hundred monthly columns, he proudly observed, “without a single interruption for cancer, hell, high water, or the World Series.” These and other pieces were bundled into ten books, among them many best-sellers. And, like Carl Sagan before him, Gould even went on to assume a place in the pop landscape, appearing everywhere from the cover of Newsweek to “The Simpsons.” At the same time, Gould, who spent most of his career at Harvard, achieved extraordinary professional distinction. He was awarded a MacArthur “genius” prize, elected to the National Academy of Sciences, and made president of the American Association for the Advancement of Science.

Then, with odd symmetry, Gould ended his career as he began it. With the release, this year, of “I Have Landed: The End of a Beginning in Natural History” (Harmony; $25.95), he brought his monthly essays to a planned close. And with the release, at the same time, of his magnum opus, “The Structure of Evolutionary Theory” (Belknap; $39.95), on which he toiled for some twenty years, he summarized a lifetime of thinking on evolution.

Gould was a modern master of the scientific essay, the inheritor of a tradition shaped by the likes of T. H. Huxley, J. B. S. Haldane, and Martin Gardner. True to form, the essays in “I Have Landed” delight in the unlikely intersection of science with you-name-it–the Alamo, the Red Sox, Nabokov, the mourners at Marx's funeral. In one piece, Gould brings together Darwin's reluctance to utter the word “evolution” and the reopening of the Hayden Planetarium, in 2000, to launch a marvellously lucid explanation of the difference between biological and stellar evolution. “When astronomers talk about the evolution of a star, they clearly do not invoke a . . . theory like Darwin's,” he writes. “Stars do not change through time because mama and papa stars generate broods of varying daughter stars.”

Yet Gould was perhaps at his best when on the attack. He warred relentlessly against what he viewed as bad science. His chief enemy was genetic determinism, the view that it's all in the genes. He battled this cant on two fronts. The first was sociobiology and its stepchild evolutionary psychology, and their often soaring speculations on the evolutionary basis of human culture. Gould charged the champions of these creeds with both a vulgar hereditarianism (they were given to saying things like “Consider a gene for gathering behavior in women”–even when no such gene has ever been found) and an addiction to untestable Just So stories (“Gathering behavior is favorable because . . .”). He went on to argue that all such “adaptationist” tales ignore the possibility that some features of animals and plants are simply by-products of how organisms are built, not the direct, designed products of natural selection. Such features, as Gould and his Harvard colleague Richard Lewontin famously put it, are like spandrels in church architecture–the triangular spaces that appear automatically between arches. Speculation about the “purpose” of these unplanned spaces is both futile and foolish. Likewise for speculation about the purpose of, say, the color of blood: as Lewontin and others have pointed out, blood looks red when it carries oxygen, but surely it's the oxygen that natural selection cares about, not the red.

The other front in Gould's war was the I.Q. industry. A large and apparently sophisticated literature claimed that I.Q. measured a single real thing called intelligence, and that this thing showed profound genetic differences across races. Over the years, these conclusions were invoked to justify a number of racist policies, including the Johnson-Reed Actof 1924, which ultimately barred entry to millions of Jews attempting to flee prewar Europe. In work that culminated in “The Mismeasure of Man,” Gould levelled city blocks of this literature, exposing its appalling intellectual shoddiness. His book enjoyed enormous popular success and earned him a National Book Critics Circle Award. This was terrific stuff, and it's too bad we won't have more of it.

But, as Gould himself observed, the “history of ideas emphasizes innovation and downgrades popularization.” (If you don't believe it, consider the long lineup of scientists who published pop books on relativity. It's not their faces on all those posters, T-shirts, and mugs.) And Gould did do serious intellectual work. “The Structure of Evolutionary Theory” is his synthesis of these labors, decades of bold and immensely controversial theorizing about how evolution works. Gould's technical work may have been less visible to his fans, but it was as well known within science as his popular work was outside. For his ideas represent the most serious challenge to evolutionary orthodoxy in half a century. Although he concedes that Darwin got much of it right, he had little doubt that his own efforts demand a fundamental change in our view of evolution.

At nearly fifteen hundred pages and five pounds, “The Structure of Evolutionary Theory” is in every sense a weighty tome. In this respect, at least, Gould is at a distinct disadvantage to Darwin. For Darwin had the good fortune of almost being scooped by the naturalist Alfred Russel Wallace, and so having to rush “On the Origin of Species” into print. Darwin always called his book an “abstract” of an intended much longer work. This may well represent Wallace's greatest contribution to science: by speeding Darwin along, he probably spared us an interminable and unreadable book. It's unfortunate that Gould had no Wallace. “The Structure” 's mammoth dimensions are due partly to its range–it surveys both the history of evolutionary thought and Gould's contributions to it–and partly to Gould's remarkably undisciplined prose. A specimen:

The intricate and multifaceted concepts that have nuanced and altered the central logic on all three branches of Darwinism's essential postulates represent ideas of broad ramification and often remarkably subtle complexity, as we vain scientists soon discovered in our fractured bubbles of burst pride–for we had been so accustomed to imagining that an evening in an armchair could conquer any merely conceptual issue, whereas we all acknowledge the substantial time and struggle that empirical problems, demanding collection and evaluation of data, often require.

The result is that, while Gould's popular essays are perhaps the most widely read texts in the history of biology, his magnum opus risks becoming one of the least.

This is a shame, because Gould has something to say. His history of evolutionary thinking, which makes up the first half of “The Structure of Evolutionary Theory,” is particularly impressive. As a rule, though, it's not for history that one turns to Gould. It's for his own unorthodox ideas on the theory of evolution, the topic of the book's second half. The most radical thing about these ideas is that Gould had them. For Gould was a paleontologist, and paleontologists, it had been understood, were there to dig deep holes, not to think deep thoughts. Geneticists and mathematicians sat at evolutionary high table–they alone had the right to theorize about how evolution works–and paleontologists kept to their place. For Gould, that place was the study of snails, an almost painfully unglamorous subject and as far from the rarefied glories of theory as one could get. Gould rejected this intellectual apartheid, announced that paleontology had plenty to say about the process of evolution, and began theorizing furiously.

The biggest of Gould's theories–and the one on which his scientific legacy will surely ride–is known as “punctuated equilibrium.” Gould introduced punctuated equilibrium with Niles Eldredge, of the American Museum of Natural History, in 1972. Their starting point was simple: trust the fossils. The fossil record, they said, shows something surprising. Species look unchanged for vast stretches of time and then–suddenly–they morph. Certain species of African snail, for instance, look the same for millions of years and then abruptly change shell shape. The question was why. The traditional answer among evolutionary biologists was that species change gradually, by natural selection, and if the fossil record says different, so much the worse for the fossil record. This attitude isn't quite as cavalier as it sounds. Evolutionary biologists have always believed that the fossil record is abysmally bad. (Imagine trying to reconstruct Western history from two snapshots, one of Pontius Pilate and the other of Evel Knievel.) Moreover, biologists can see gradual adaptive change happening around them. (Think of antibiotic resistance.) So, the argument went, we're better off extrapolating from what we can see clearly now than trusting a fragmentary record of what allegedly happened then.

Gould and Eldredge believed otherwise. They said that the pattern of long stasis punctuated by sudden change is real. It doesn't reflect gaps in the data; it is the data. They also said–and here we move from the pattern to the theory side of punctuated equilibrium–that this pattern can be explained by two ideas. The first idea is that creatures are robust beings that resist the pressures of changing environments. This conservatism, they argued, reflects the complexities of development, the intricate process by which an organism goes from a single cell to a strapping adult: development is so tightly coordinated that it can't be easily tinkered with without breaking down entirely. The result is no evolutionary change. This idea of “developmental constraints” was the most heterodox aspect of punctuated equilibrium. Darwin, after all, held that animals and plants are nearly infinitely pliant, adapting to the subtlest shift in the environment; Gould and Eldredge held that organisms are stiff and unyielding.

The second idea is that sudden change in the fossil record happens at the very moment that a species splits into two species. According to conventional theory, “ordinary” evolutionary change and species-splitting have little to do with each other. If we consider two lizard populations–one on one side of a river and one on the other–traditional theory says that either population could start to change, say, tail length at any time. If we could leap forward and look at these populations a million years hence, we might find that although they began with identical tail lengths they now have tails of markedly different lengths. This wouldn't mean, however, that these lizards now belong to separate species. They would belong to separate species only if the two populations could no longer mate with each other and produce fertile offspring; only, that is, if they could no longer share genes. The genetic changes underlying this process of “speciation” are thought to accumulate slowly and gradually between geographically separated populations. But the essential point is that, under this traditional view, speciation and change in ordinary features like tail length are not necessarily simultaneous.

This is the view that Gould and Eldredge rejected. Echoing arguments made by the naturalist Ernst Mayr, they claimed that speciation involves “genetic revolutions,” episodes of extensive genetic change that shake up much of an organism's genome. Going even further, Gould and Eldredge argued that only speciation–only passage through a genetic revolution–is sufficiently violent to break the binds of developmental constraints. The result is that all evolutionary change is restricted to rare moments of species-splitting. Lizards can't just go changing tail length; they can do so only when splitting into different species. So much for Darwinism.

Punctuated equilibrium ignited enormous controversy. The scientific literature erupted into attack and counterattack; conferences melted down. Even by the indelicate standards of science, the debate got ugly. Punctuated equilibrium was branded “evolution by jerks,” and Gould was, as he says, “reviled in many quarters.” The ferocity of the proceedings had several causes. One was that a great deal was at stake. Punctuated equilibrium struck at the foundations of evolution, and scientists do not, as a rule, like being told that their theoretical edifice teeters on a wobbly base. Then there was the matter of Gould's style. Gould was, by his own admission, the “most arrogant of literati,” and his outsized confidence seemed to provoke an escalation in his (and everyone else's) rhetoric. In 1980, he brashly pronounced the modern theory of Darwinism “effectively dead,” and predicted the emergence of “a new and general theory of evolution.” Last, the theory of punctuated equilibrium was at odds with the evidence, unsalvageably so. There was simply no reason to think that speciation is a tumultuous event during which evolutionary change is especially likely to take place. Worse, mountains of data contradicted the claim that organisms are bound by developmental constraints, unable to change unless they speciate. The best of this evidence is no farther away than the corn on your plate or the Chihuahua on your sofa. If man can shape these things from their wild ancestors by artificial selection–and without speciation–talk of strong constraints is in a bad way.

Gould ultimately conceded all this, though at an agonizing pace that spanned decades. (In this respect, if in no other, he was a gradualist.) In “The Structure of Evolutionary Theory,” he admits that “we made mistakes,” and that prominent among these were the claims about both constraint (idea one) and speciation (idea two). The theory part of punctuated equilibrium was thereby gutted. But what of the pattern part? If a punctuated pattern characterizes the fossil record–and Gould insists that it does, and spends a good deal of “The Structure” trying to show that new species often appear abruptly in the fossil record alongside ancestral ones–what explains it?

Gould concludes that the best guess was one made by the evolutionist Douglas Futuyma in the eighties. Futuyma's idea starts with the fact that different populations of a species often adapt to local features of their environments. Mice in the north, for instance, might evolve to be bigger than those in the south because big bodies lose proportionally less heat. But such local differences are usually too short-lived to show up in the fossil record. That's because populations are apt to come back into contact with each other, and when they do they begin to mate and so swap genes. As a result, any differences among the populations tend to get blended away: close encounters between northern and southern mice yield mid-sized mice. There's only one way to prevent this washing-out: populations must not only adapt but speciate. For separate species, by definition, can't swap genes and so won't lose their distinctive looks by genetic blending when they come into contact. This, then, might explain why, in the fossil record, change and speciation seem to go hand in hand. It's not that change is rare and happens only when species split; it's that change is common, but only if species split do the resulting differences last long enough to have a shot at showing up as fossils.

Futuyma's idea, though, had one embarrassing blemish: it's pure Darwinism. Gould and his allies thus found themselves in a peculiar position: the red-hot revolutionaries suddenly seemed staid traditionalists. This was not, of course, lost on evolutionists, many of whom were still steaming over the Darwinism-is-dead business. By the nineties, most evolutionary biologists had simply stopped paying attention to punctuated equilibrium. At best, the theory looked like a moving target, veering now at breakneck speed toward Darwinism. At worst, its chief advocate seemed muddled, a mixture of radical rhetoric and malleable ideas. Punctuated equilibrium was down, if not out.

In “The Structure of Evolutionary Theory,” Gould fights mightily to defend his position against charges of collapse. He offers two main defenses. One is to deny reports of his retreat to orthodoxy by claiming that his model was fairly traditional all along. He was, he says, just trying to change paleontology by importing a “fully conventional” biological theory. But this won't do. Gould's model of speciation was far from conventional. And his ideas about strong developmental constraints were clearly and defiantly non-Darwinian. Gould's attempts to take the edge off his earlier views–as when he tells us that his claim that modern Darwinism was “dead” wasn't meant to imply that it was “wrong”–represent one of the least satisfying aspects of “The Structure.”

Gould's second defense is far more important. Early in the debate, he began to reassess just what was revolutionary about punctuated equilibrium. He came to think that the truly outre aspect of the theory was something called species selection. In Darwin's account, natural selection acts at the level of organisms, not at the level of species. Some organisms are better adapted to their environments than others, and so have more progeny. Imagine, for instance, two kinds of moths belonging to the same species. One is white and easily seen by bird predators; the other is brown and often mistaken by birds for a dead leaf. The result is that the brown moths typically have more offspring: brown moths have a higher “fitness” than white ones. The percentage of brown moths will, consequently, increase with each generation. This is normal “organismal selection”–it yields organisms that neatly fit their environments.

But why, Gould asked, can't we extend the selectionist logic down or up the biological ladder, and talk about competition among genes, or cells, or populations, or species? The downward extension has, in fact, been achieved: William Hamilton, Richard Dawkins, and others argued that selection typically acts at the level of genes. Gould, though, was more interested in the up direction–in the idea that selection can act at the level of species. According to this theory, some species split faster or go extinct more slowly than others, and therefore become more common through time. Imagine, for example, two plant species. One, a dandelion, disperses its seeds on the wind–seeds sail away to new locations. The other, a cocklebur, disperses its seeds by hooking them onto the fur of passing animals. Now, plants that disperse by wind might be more likely to speciate than those which disperse by animals: wind-dispersed seeds might be more likely to travel long distances and so found a new population on some distant shore. Such a population–living in blissful isolation from the rest of the species–has a good chance of someday evolving into a new species. If wind dispersers speciate faster than animal dispersers, the percentage of wind-dispersed species will increase over long stretches of time. Thus if we start out with one dandelion species and one cocklebur species, the fast-speciating dandelion will give rise to more “daughter” species, which, in turn, will give rise to more “granddaughter” species, and so on. After millions of years, the earth might well be enveloped by more dandelion-like than cocklebur-like species. This is species selection. The important point is that this process has nothing to do with Darwinian competition at the level of organisms. Wind-dispersed species aren't getting more common because a wind-dispersed organism outcompetes an animal-dispersed organism. They're getting more common because the wind-dispersed species form new species faster than the animal-dispersed ones. Note also that the critical quantity here–speciation rate–is a property of whole species, not of individual organisms. You can't point to one petunia in your garden and ask if it has a higher speciation rate than another. But you can point to wind-dispersed species and ask if they have a higher speciation rate than animal-dispersed ones. Species selection thus involves a higher level of fitness than biologists are used to thinking about–one that reflects the birth and death of species, not of organisms. Gould rounds out his picture of natural selection by arguing that it acts simultaneously at many levels–on genes, on organisms, on species, and perhaps even on higher levels. What he calls “hierarchical selection” is the claim not that species selection is right and organismal selection is wrong but that both act at once.

Gould has high hopes for hierarchical selection, and it takes center stage in “The Structure of Evolutionary Theory.” Hierarchical selection, he says, means that we must “reconceptualize all of evolution, and revise both our worldview and our language, accordingly.” Indeed, he goes on, “we cannot fairly portray these expanded views as pure sweetness and light for orthodox Darwinism. Much that has been enormously comfortable must be sacrificed.” Though the rhetoric is extreme–and it's hard to see what species selection has to do with punctuated equilibrium anyway–Gould's argument for higher-level selection is important. Twenty-five years ago, biologists typically believed that selection was more or less constrained to act on organisms or genes. This is no longer the case, and Gould deserves much of the credit for the change. His defense of species selection–at least as a formal possibility–is in places brilliant. This is Gould at his best. It is also, not coincidentally, Gould on one of the themes nearest his heart: the sheer complexity of the biological world, its stubborn refusal to submit to a simple analysis or a single equation.

But hierarchical selection is not without problems. The chief difficulty, which Gould concedes, is that it's extremely hard to distinguish species selection from other evolutionary forces when you're looking at evidence from the natural world. An increase in the percentage of dandelion-like species might be due to species selection, but it might also be due to a number of other processes, including old-fashioned organismal selection. The consequence, which Gould also acknowledges, is that there are no undisputed cases of species selection in the real world. This does not mean that the idea is wrong; I, for one, suspect that species selection acts in nature. But it does mean that Gould's species selection has a very different status than Darwin's organismal selection. Organismal selection made sudden sense of a vast number of puzzling phenomena (why, for instance, some moths look like dead leaves). Species selection remains an explanation in search of something to explain.

The worry is that all this misses what Gould was really up to. Over the years, I've come to an admittedly speculative view of Gould's modus operandi as a scientist. It starts from the idea that there were two figures looking over his shoulders as he wrote. One was Charles Darwin, but the other was Thomas Kuhn. Gould was steeped in Kuhn's 1962 book “The Structure of Scientific Revolutions” and his vision of alternating bouts of workaday “normal” science and revolutionary paradigm shifts. In “The Structure of Evolutionary Theory,” Gould fondly reminisces about his first reading of Kuhn as an impressionable first-year graduate student. Kuhn deeply affected Gould's science. One might even argue that punctuated equilibrium is little more than Kuhn's view of the history of ideas transferred wholesale to the history of life, an idea that is reinforced by the fact that Gould and Eldredge began their 1972 paper with talk of paradigms and Kuhn. Even the title of Gould's magnum opus seems a riff on “The Structure of Scientific Revolutions.”

In fact, Kuhn seems to have transformed not only Gould's image of science but his image of the scientist. The revolutionary scientist, Gould seemed to conclude, is one who knows that big progress follows big jolts. Just as old and hopelessly constrained species can do nothing interesting unless they get periodically shaken, so old and hopelessly conservative paradigms can't give way to new science unless they receive a good swift kick now and then. These assaults are characterized by a great deal of noise, confusion, and hurt feelings. But, more than anything, they're characterized by a protagonist who's willing to stick his neck out farther than others dare–one who's willing to be wrong. Major progress, Gould seemed to believe, demands major risk. And, even if you get things wrong, science may ultimately gain. People ask questions and make discoveries that would otherwise have gone unasked and undiscovered. (Gould quotes the economist Vilfredo Pareto with approval here: “Give me a fruitful error any time, full of seeds, bursting with its own corrections. You can keep your sterile truth for yourself.”) Gould might well then represent something new in the historical strata of science: the first self-consciously revolutionary scientist–the first scientist who set out to create a revolution at least in part because he felt that the field just needed one.

If something like this characterizes Gould's approach, it complicates any attempt to assess his legacy. Gould's attacks on adaptationism may have been extreme, but fanciful Just So stories are now, thankfully, rarer. Key parts of punctuated equilibrium were wrong, but paleontological data are, largely due to Gould, richer than ever. Species selection may not make sudden sense of the fossil record, but a reinvigorated paleontology sits at evolutionary biology's high table. In the end, Gould's career may force us to separate two questions that are usually conflated: was he right, and was he good for science? It may not, after all, be a law of nature that the two have the same answer.