December in moscow, and the temperature drops under 15 degrees below zero. The radiators in the bar have grown cold, so I sit in a thick coat and gloves, drinking vodka while I ponder the fossil birds. The year is 2001, and the now late Evgeny N. Kurochkin of the Russian Academy of Sciences and I have just spent hours at the paleontology museum as part of our effort to survey all the avian fossils ever collected in Mongolia by joint Soviet-Mongolian expeditions. Among the remains is a wing unearthed in the Gobi Desert in 1987. Compared with the spectacularly preserved dinosaur skeletons in the museum's collections, this tiny wing—its delicate bones jumbled and crushed—is decidedly unglamorous. But it offers a strong hint that a widely held view of bird evolution is wrong.

More than 10,000 species of birds populate the earth today. Some are adapted to living far out on the open ocean, others eke out a living in arid deserts, and still others dwell atop snow-capped mountains. Indeed, of all the classes of land vertebrates, the one comprising birds is easily the most diverse. Evolutionary biologists long assumed that the ancestors of today's birds owed their success to the asteroid impact that wiped out the dinosaurs and many other land vertebrates around 66 million years ago. Their reasoning was simple: although birds had evolved before that catastrophe, anatomically modern varieties appeared in the fossil record only after that event. The dawning of ducks, cuckoos, hummingbirds and other modern forms—which together make up the neornithine (“new birds”) lineage—seemed to be a classic case of an evolutionary radiation in response to the clearing out of ecological niches by an extinction event. In this case, the niches were those occupied by dinosaurs, the flying reptiles known as pterosaurs and archaic birds.

Over the past decade or so, however, mounting evidence from the fossil record—including that crushed wing—and from analyses of the DNA of living birds has revealed that neornithine birds probably diversified earlier than 66 million years ago. The findings have upended the traditional view of bird evolution—and sparked important new questions about how these animals soared to evolutionary heights.

Early Birds
Birds are one of just three groups of vertebrates ever to have evolved active, flapping flight. The other two are the ill-fated pterosaurs and the bats, which appeared much later and share the skies with birds to this day. For years paleontologists debated the origin of the earliest birds. One side argued that they evolved from small, meat-eating dinosaurs called theropods; the other contended that they evolved from earlier reptiles. But the discoveries over the past two decades of birdlike dinosaurs, including many with downy coats, have convinced most scientists that birds evolved from theropod dinosaurs.

Connecting the dots between ancestral avians and modern birds has proved far trickier, however. Consider Archaeopteryx, the 145-million-year-old creature from Germany that is the oldest known bird. Archaeopteryx preserves the earliest definitive evidence for wings with asymmetric feathers capable of generating the lift required for flight—one defining characteristic of the group. Yet it more closely resembles small-bodied dinosaurs such as Velociraptor, Deinonychus, Anchiornis and Troodon than modern birds. Like those dinosaurs, early birds such as Archaeopteryx and the more recently discovered Jeholornis from China and Rahonavis from Madagascar possessed long, bony tails, and some had sharp teeth, among other primitive traits. Neornithines, in contrast, lack those characteristics and exhibit a suite of advanced ones. These features include fully fused toe bones and fingerless wings, which reduce the weight of the skeleton, allowing more efficient flight, and highly flexible wrists and wings, which enhance maneuverability in the air. How and when the neornithines acquired these traits were impossible to determine, however, thanks to an absence of fossils documenting the transition.

This is not to say the fossil record lacked avian remains intermediate in age between the first birds and the postextinction neornithines. Clearly by more than 100 million years ago, in the Early Cretaceous, birds representing a wide range of flight adaptations and ecological specializations had evolved. Some flew on wings that were broad and wide; others had wings that were long and thin. Some lived in forests eating insects and fruit; others made their home along lakeshores or in the water and subsisted on fish. This incredible diversity persisted through the latest stages of the Cretaceous, 66 million years ago. In fact, along with my Dutch colleagues at the Maastricht Natural History Museum, I have described remains of toothed birds found just below the geologic horizon that marks the end-Cretaceous extinction event. But all the Cretaceous birds complete enough to classify belonged to lineages more ancient than neornithines, and these lineages did not survive the catastrophe—which is why, until recently, the available evidence implied that the simplest explanation for the rise of modern birds was that they originated and radiated after the extinction event.

Molecular Clues
By the 1990s, while paleontologists were still looking for ancestral neornithines in the Cretaceous and coming up empty-handed, another method of reconstructing the evolutionary history of organisms—one that did not involve the fossil record—was gaining traction. Molecular biologists were sequencing the DNA of living organisms and comparing those sequences to estimate when two groups split from each other. They can make such estimates because certain parts of the genome mutate at a more or less constant rate, constituting the “ticking” of the so-called molecular clock.

Molecular biologists had long questioned the classical, fossil-based view of modern bird evolution. So they tackled the problem using their clock technique to estimate the divergence dates for major lineages of modern birds. Among the most significant splits is the one that occurred between the large, mostly flightless paleognaths (ostriches and emu and their kin) and the Galloanserae (which includes chickens and other members of the Galliformes group, as well as ducks and other members of the Anseriformes group). The DNA studies concluded that these two lineages—the most primitive of the living neornithines—split from each other deep in the Cretaceous. And researchers obtained similarly ancient divergence dates for other lineages.

The findings implied that, contrary to conventional paleontological wisdom, neornithines lived alongside dinosaurs. It is funny to think of a robin perched on the back of a Velociraptor or a duck paddling alongside a Spinosaurus. But the molecular evidence for the contemporaneity of modern birds and dinosaurs was so compelling that even the paleontologists—who have typically viewed with skepticism those DNA findings that conflict with the fossil record—began to embrace it. Still, those of us who study ancient skeletons urgently wanted fossil confirmation of this new view of bird evolution.

Ducks in a Row
At last, after the new millennium, paleontologists' luck began to change for the better, starting with the tiny Mongolian wing that Evgeny and I focused on in Moscow. Back when Evgeny first saw the fossil in 1987, he told me that he thought it looked like a member of the presbyornithids, a group of now extinct ducklike birds related to modern ducks and geese. But at 70 million years old, it was a Cretaceous bird, and everyone knew—or thought they did—that there was no definitive evidence for presbyornithids in the Cretaceous. Yet our comparisons in the museum that cold winter in 2001 demonstrated conclusively that the wing—with its straight carpometacarpus (the bone formed by the fusion of the hand bones) and details of canals, ridges and muscle scars—did indeed belong to a presbyornithid, which, moreover, was the oldest unequivocal representative of any neornithine group. Our finding fit the predictions of the molecular biologists perfectly. In a 2002 paper that formally described the animal, we gave it the name Teviornis.

Before long, Teviornis was joined by a second confirmed early neornithine, Vegavis, from Antarctica's Vega Island. Vegavis had been found in the 1990s only to languish in relative anonymity for years before its true significance came to light. In 2005 Julia A. Clarke, now at the University of Texas at Austin, and her colleagues published a paper showing that Vegavis was another bird from the Cretaceous that exhibits a number of features found in modern ducks, particularly in its broad shoulder girdle, pelvis, wing bones and lower legs. At 68 million to 66 million years old, Vegavis is a little younger than Teviornis but still clearly predates the mass extinction. And it is a much more complete fossil, preserving the better part of a skeleton.

For most paleontologists, Vegavis clinched the case for Cretaceous neornithines. Thus enlightened, researchers have begun reexamining fossil collections from this time period, looking for additional examples of early modern birds. One investigator, Sylvia Hope of the California Academy of Sciences in San Francisco, had been arguing for years that bird species she has identified from fossils found in New Jersey and Wyoming that date to between 100 million and 80 million years ago are modern. But the finds—mostly single bones—had been considered by other researchers as too scrappy to identify conclusively. The revelations about Vegavis and Teviornis suggest that she was right all along. Comparisons of Hope's bones with more complete remains should prove illuminating in this regard.

Flying the Coop
Rooting modern birds in the Cretaceous neatly aligned the fossil record with the DNA-based divergence dates. But it raised a vexing new question: Why were modern birds able to survive the asteroid impact and its attendant ecological changes when their more primitive avian cousins and their fellow fliers, the pterosaurs, were not? To my mind, this constitutes the single biggest remaining mystery of bird evolution. The answer is still very much up for grabs, and I am devoting much of my research at the moment to trying to get at it.

With only a couple of confirmed Cretaceous neornithines on record, there is not much in the way of fossil clues to go on. Insights have come from studies of living birds, however. Using a huge data set of measurements of living birds, my colleagues in the U.K. and I have shown, for example, that the wing-bone proportions of primitive modern birds, including Teviornis and Vegavis, are no different from those of the extinct enantiornithines. Comparing the fossil wing-bone proportions with those of living birds allows us to infer some aspects of wing shape and hence gain information about the aerodynamic capabilities of fossil birds. But so far as we can tell, the wing shapes of the two groups of fossil birds do not differ; in other words, we do not think that early neornithines were any better at flying than were the enantiornithines (although both these groups were most likely better in the air than earlier theropodlike birds such as Archaeopteryx).

If flight ability did not give the neornithines an advantage over their Cretaceous counterparts, what did? A number of paleontologists, including me, have posited that differences in foraging habits might have conferred a competitive edge. In support of that theory, I have shown in a series of papers published over the past few years that modern birds preserved in the immediate aftermath of the mass extinction, in rocks 60 million years old and younger, probably lived mostly in wet environments: coastlines, lakes, the edges of rivers and the deep ocean, for example. Many of the birds that inhabit such environments today—ducks among them—are typically generalists, able to subsist on a wide variety of foods. And ducklike birds are currently the one confirmed lineage of modern birds we have found in the Cretaceous. The groups of Cretaceous birds that did not survive the disaster, in contrast, have been collected from rocks that were formed in many different kinds of environments—including seashores, inland areas, deserts and forests. This ecological diversity may indicate that the archaic birds had evolved specializations for feeding in each of these niches. Perhaps, then, the secret of early modern birds' success was simply the fact that they were less specialized than the other groups.

Such flexibility might have enabled the neornithines to adapt more easily to the changing conditions that followed the asteroid impact. It is an appealing idea, but these are early days. Only with the discovery of more fossils—whether in the ground or in museum drawers—will we be able to determine how modern birds eluded elimination and took wing.