The Origin of Feathers: a Display Hypothesis

These extracts from our original paper (Cowen and Lipps, 1982) and from three successive editions of History of Life (1990, 1995, 2000) show how the idea has been updated, being tested at each iteration by new data. Cowen holds the copyright on the last three, and the first is out of print, being posted here under the fair practice principle.

The 1982 version is obviously outdated, though the basic idea is unchanged. I present here only an extract from it that deals with feathers and flight. The preamble was a defense of scenarios as testable hypotheses, which is still true but so unfashionable now that it will simply annoy many people and perhaps cause them to discount the very real and viable hypothesis we present.

The 1995 version is the fullest, since space constraints required the 2000 version to be shorter. Nevertheless, by 2000 the hypothesis had been tested by the discovery of feathered theropods in China, and it emerged from that test stronger than ever.

The 2005 version is in the newest (4th) edition, but is not radically altered because it doesn't have to be. But see a recent comment from August 2004.

The textbook extracts were, of course, written for undergraduates. The language is comparatively informal, and there are no references, but the presentation is clear. I would be happy to discuss, reference, or expand on any particular point: E-mail me at rcowen@ucdavis.edu

The versions are laid out here in reverse chronological order. They should be referenced as

  1. Cowen, R., and J. H. Lipps. 2000. The origin of feathers and the origin of flight in birds. In Cowen, R., History of Life, 3rd edition, Chapters 13 and 14. Malden, Mass.: Blackwell Science.
  2. Cowen, R., and J. H. Lipps. 1995. The origin of feathers and the origin of flight in birds. In Cowen, R., History of Life, 2nd edition, Chapter 13. Cambridge, Mass.: Blackwell Scientific Publications.
  3. Cowen, R., and J. H. Lipps. 1990. The origin of feathers and the origin of flight in birds. In Cowen, R., History of Life, 1st edition, Chapter 12. Cambridge, Mass.: Blackwell Scientific Publications.
  4. Cowen, R., and Lipps, J. H. 1982. An adaptive scenario for the origin of birds and of flight in birds. Proceedings of the 3rd North American Paleontological Convention, Montréal, vol. 1, 109-112.

EXTRACTS FROM HISTORY OF LIFE, 3rd edition (2000)
From Chapter 13:

The Origin of Feathers

The proteins that make feathers in living birds are completely unlike the proteins that make reptilian scales today. Feathers originate in a skin layer deep under the outer layer that forms scales. It is very unlikely that feathers evolved from reptilian scales, even though that thought is deeply embedded in the minds of too many paleontologists. Feathers probably arose as new structures under and between reptile scales, not as modified scales. Many birds have scales on their lower legs and feet where feathers are not developed, and penguins have such short feathers on parts of their wings that the skin there is scaly for all practical purposes. So there is no real anatomical problem in imagining the evolution of feathers on a scaly reptilian skin. But feathers evolved in theropods as completely new structures, and any reasonable explanation of their origin has to take this into account.

Obviously, feathers did not evolve for flight. They evolved for some other function and were later modified for flight.

Feathers may have evolved to aid thermoregulation. The feathered Chinese theropods all have down, probably as insulation to keep their bodies at an even temperature. It would not matter whether they used their feathers to conserve heat in cold periods, or to keep heat out in hot periods, or both. In either case, insulation would have been useful.

The thermoregulatory theory for the origin of feathers is probably the most widely accepted one today, but it does have problems. Why feathers? Feathers are more complex to grow, more difficult to maintain in good condition, more liable to damage, and more difficult to replace than fur. Every other creature that has evolved a thermoregulatory coat, from bats to bees and from caterpillars to pterosaurs, has some kind of furry cover. There is no apparent reason for evolving feathers rather than fur even for heat shielding.

Furthermore, thermoregulation cannot account for the length or the distribution of the long feathers on Protarchaeopteryx or Caudipteryx. Short feathers (down) can provide good thermoregulation, but thermoregulation does not require long feathers, and it would not help thermoregulation very much to evolve long strong feathers on the arms and tail. So it is difficult to suggest that feathers evolved for thermoregulation alone. It would be better to think of another equally simple explanation.

I naturally prefer an idea that I developed years ago, with my colleague Jere Lipps. In living birds, feathers are for flying, for insulation, but also for camouflage and/or display. Lipps and I suggested that feathers evolved for display. The display may have been between females or between males for dominance in mating systems (sexual selection), or between individuals for territory or food (social selection), or directed toward members of other species in defense.

Living reptiles and birds often display for one or all of these reasons, using color, motion, and posture as visual signals to an opponent. Display is often used to increase apparent body size; the smaller the animal, the more effectively a slight addition to its outline would increase its apparent size. Lipps and I therefore proposed that erectile, colored feathers would give such a selective advantage to a small displaying theropod that it would encourage a rapid transition from a scaly skin to a feathery coat. Display would have been advantageous as soon as any short feathers appeared, and it would have been most effective on movable appendages, such as forearms and tail. (Display on the legs would not be so visible or effective.) Forearm display by a small theropod would also have drawn particular attention to the powerful weapons the theropod carried there, its front claws (Figure 13.1). The Chinese theropod Caudipteryx carried long feathers on its middle finger, between the two outside claws, and it could fold that middle finger away, with the feathers out of harm's way, during a strike.

The display hypothesis explains more features of the feathered theropods and the first bird Archaeopteryx than other hypotheses, with fewer assumptions. It explains completely the feather pattern: the evolution of long strong feathers on arms and tail.

Once they evolved, feathers could quickly have been co-opted for thermoregulation, and the down coat on the Chinese theropods may show that process. Down can only be for thermoregulation. Although down is not proof of warm blood, it is very strong evidence in favor of it. In living birds, down feathers are associated with the problem of heat loss for hatchlings.

From Chapter 14:

The Origin of Powered Flight in Birds

Since ground-running theropods had feathers, the question of the origin of flight in birds has nothing to do with the appearance of feathers. (Flight evolved in bats and pterosaurs without feathers.) There have been three important hypotheses for the origin of bird flight, and I shall add a fourth.

The Arboreal Hypothesis
The arboreal hypothesis suggests that ancestral birds evolved flight by jumping out of trees. The arboreal theory was the most favored until recently, and it still has supporters. But it must be abandoned in the face of the new theropods from China. With long, erect limbs, a comparatively short trunk, and bipedal locomotion, Archaeopteryx and the feathered theropods are exactly the opposite in body plan of all living mammals and reptiles that jump and glide from tree to tree. There is nothing in the ancestry of birds as we now know it to suggest any arboreal adaptations at all.

Flapping arms or proto-wings (in fact, any feathers at all on wings or tail) increase drag. Aerodynamically, the transition from gliding to flapping is difficult: there is only a narrow theoretical window through which the transition could have been made (page 216). Such a transition would have been especially difficult for Archaeopteryx because it had such a long, bony tail with long feathers on it. This kind of tail adds much more drag than it adds lift.

The Cursorial Hypothesis
Perhaps some adaptations in a ground-dwelling theropod could provide some of the anatomy and behavior necessary for flight, such as lengthening the forearms, especially the hands, placing long, strong feathers in those areas, and evolving powerful arm movements. An early version of the cursorial hypothesis suggested that a fast-running reptile might evolve long scales on the arms. In this theory, the scales generated lift as the arms were actively flapped on the run (Figure 14.17). The animal could now take long leaps, perhaps encouraging the scales to evolve into feathers and the leaps to evolve into powered flapping flight.

Feathers did not evolve from scales, but in any case the idea does not work mechanically. Any lift generated by a flapping arm decreases the ground traction given by the feet, and acceleration is lost. A racing car is held down on the track by its airfoils for good traction, and an aircraft cannot be driven through its wheels on the takeoff run. A running theropod that flapped its arms would increase drag: the faster the run, the greater the drag. Only a very small amount of thrust would have been generated by the arms in the early stages of the process. A new version of this theory was published in 1999, but I suspect that the assumptions in that paper were not correct.

The Running Raptor
More recent versions of the cursorial hypothesis are much better: they are mechanically sounder and include behavior that involved strong, synchronized arm strokes and the evolution of strong pectoral muscles.

Gerald Caple and his colleagues suggested that a protobird hunted by running fast and leaping after flying or jumping insects it disturbed. To catch an elusive dodging prey while its feet were off the ground, a protobird would have to be able to adjust its body attitude in the air and then regain a stable position for landing. Such adjustments could be made aerodynamically by generating a small amount of lift or drag, and that would be best added at the tips of the arms. If the right arm movements were added, the effect would be greater still. The protobird would now be well on the way to flapping takeoff, and the flights would be gradually prolonged until complete control had been reached.

But this proposed activity would consume a lot of energy. No predator today, bird or otherwise, runs at high speed to flush out insects it can leap after. Furthermore, effective attitude control for a leaping animal requires a critical airspeed that is high in the early stages of the process, when the proto-wings are just beginning to generate lift. The required speed might have been 10 meters per second, over 25 mph, far too much for any reasonable proto-bird.

The Display and Fighting Hypothesis
Jere Lipps and I suggested that display was involved in the evolution of flight as well as feathers. Theropods had long, strong display feathers on arms and tail (Chapter 13). Successful display was increased by lengthening the arms, especially the hand, and by actively waving them, perhaps flapping them rapidly and vigorously. Flapping in display would have encouraged the evolution of powerful pectoral muscles, and the supracoracoideus system.

Display can be very effective, and not just for sexual ends. Frigate birds and bald eagles often try to rob other birds of food instead of catching prey themselves. Because the penalty for wing injury is high, many birds can be intimidated by display into giving up their catch rather than fighting to defend it.

But a threat display must not be exposed as an empty bluff. Fighting is the last resort. Living birds often fight on the ground, even those that fly well. The wings no longer have claws but are still used as weapons in forward and downward smashes (steamer ducks are particularly deadly at this). Beaks and feet can be used as weapons too, and are most effective when used in a downward or forward strike.

A strong wing flap, directed forward and downward, is also the power stroke that gives lift to a bird in takeoff. Lipps and I suggest that strong wing flapping is a simple extension of display flapping, encouraged by fighting behavior. Powerful flapping used to deliver forearm smashes could have lifted the bird off the ground, allowing it also to rake its opponent from above with its hind claws. The more rapidly the wings could be lifted for another blow, the more effective the fighting. This would rapidly encourage an effective wing-lifting motion that minimized air resistance, so the wing action would then be almost identical to a takeoff stroke.

A variant of our idea has also been proposed by Kevin Padian, who prefers to think of the wing stroke evolving from the arm strike used by a theropod in predation. It is not clear how this could have led easily to whole-body takeoff, however. Caudipteryx was able to fold its feathers away while making its fighting stroke.

Archaeopteryx fits our display-and-fighting hypothesis well. It was well adapted for display. Like any small theropod, it was well equipped for fighting with its teeth and the strong claws on hands and feet. Archaeopteryx did not have long primary feathers on its fingers (Figure 13.1), probably because they would have hidden the claws in display and would most likely have broken in a fight.

Archaeopteryx could not fly well; I suspect that it hardly flew at all. It may have been able to glide a short distance, but it could not have sustained flapping flight. It did not have the supracoracoideus system (Figure 14.15). This muscle passes through the shoulder joint, and as well as raising the wing, it twists it. On the upstroke, the twist arranges the wing and feathers so that they slip easily through the air, with little drag. At the top of the upstroke, the wing is in exactly the right position to give a powerful downbeat. Without the supracoracoideus (which is easily identified because it leaves a strong trace on the shoulder joint), a bird cannot fly by wing flapping. In fact, it cannot even take off and land, because the greatest power from the wings is required during slow flight.

In small flying birds today, the wishbone acts as a spring that repositions the shoulder joints after the stresses of each wing stroke. It is needed to give the rapid flaps necessary for flight (a starling flies with 14 complete wing beats per second). The wishbone also helps to pump air in and out of the lungs, and to recover some of the muscular energy put into the downstroke. In Archaeopteryx and in theropods with wishbones, the bone is U-shaped and strong and solid; it could not have acted as an effective spring. Furthermore, Archaeopteryx did not have the long primary feathers on the wing tips, or the breastbone anchoring the muscles that are needed for routine takeoff and landing. It could not have raised its arms high above its body for an effective downstroke. In fact, Archaeopteryx evolved structures that were active deterrents to flight. Its tail was long and bony, with long feathers. Among living birds with display feathers, this sort of tail is aerodynamically the worst of all possible tail styles, adding a lot of drag and little lift.

Archaeopteryx, then, was a fierce little fast-running, displaying bird, which probably spent its life scurrying around the Solnhofen shore, hunting for small prey such as crustaceans, reptiles, and mammals. In hunting style, Archaeopteryx was probably much like the roadrunner of the dry country of the American Southwest, but its ecological setting was like that of a steamer duck: on a shoreline with year-round food supply. Archaeopteryx did not compete in the air with the pterosaurs that are also found in the Solnhofen Limestone.

From Fight to Flight
Display and fighting in birds takes a lot of energy, whether it is for territory, dominance, or food, but it provides an enormous payoff in survival and selection. Sexual display in most living birds must be done correctly, or no mating takes place. New behaviors can be evolved rapidly, and they are often evolutionarily cheap, because they usually don't require important morphological changes in their early stages.

The display hypothesis suggests that the earliest birds gained flight behavior, anatomy, and experience at low ground speed and low height: ideal preflight training. The selective payoff for successful mastery of the flight motions gave significant advantages, even before flight itself was possible. From that point, the many advantages of flight were added to those of social or sexual competition. I do not think it is a coincidence that the males of the Early Cretaceous Chinese birds Confuciusornis and Changchengornis had extravagantly long (display) feathers on the tail!

Once liftoff was achieved, flapping flight quickly followed. In more advanced birds than Archaeopteryx, the supracoracoideus tendon system evolved in the shoulder, while the wishbone evolved into a spring. The breastbone evolved as the anchor for the flight muscles. The forearms became longer, lighter, and more fragile in bone structure, becoming specialized as wings, and losing the finger claws. Meanwhile, the feet and beak became the dominant fighting weapons, as in most living birds.

EXTRACT FROM HISTORY OF LIFE, 2nd edition (1995), Chapter 13:

The Origin of Feathers

Feathers in living birds originate in a deep skin layer deep under the outer layer that forms scales. Evolutionarily, then, feathers probably arose under and between reptile scales, not as modified scales. Many birds have scales on their lower legs and feet where feathers are not developed, and penguins have such short feathers on parts of their wings that the skin there is scaly for all practical purposes. So there is no real anatomical problem in imagining the evolution of feathers on a reptilian skin. But feathers are completely novel structures, and any reasonable explanation of their origin has to take this into account. They evolved in the first birds to replace scales as the primary skin covering. The problem is to reconstruct why this happened.

Feathers may have evolved directly for flight. If so, they evolved in a reptile that was already launching itself into the air, presumably in a tree-dwelling jumper and perhaps parachuter. In this hypothesis, feathers were first an aid to parachuting and then a way to achieve flapping flight. This is a difficult process to imagine. Why feathers? The scales of gliding reptiles do not project beyond the boundary layer of air around the body, so they do not generate any lift. It's not clear that protofeathers would have been any improvement. The bone-supported skin membranes of parachuting reptiles are a much easier and cheaper way to evolve an airfoil than any conceivable airfoil made of protofeathers. Bats and pterosaurs evolved flapping flight without feathers. Perhaps feathers evolved for some other function and were later modified for flight.

Feathers may have evolved to aid thermoregulation. Small theropods probably had a high metabolic rate and may have been warm-blooded (Chapter 12). Very small theropods would have needed additional insulation to keep their bodies at even temperature. A few small reptiles today use long scales to help trap a layer of air between the environment and the body surface to cut down temperature fluctuations, usually as a heat shield against the sun. It would not matter whether protofeathers were needed to conserve heat in cold periods, or to keep heat out in hot periods, or both. In either case, insulation would have been useful. The skin musculature would have been able to rais and lower protofeathers, allowing free flow of air to the skin when necessary.

This theory for the origin of feathers is probably the most widely accepted one today, but it does have problems. Again, why feathers? Feathers are more complex to grow, more difficult to maintain in good condition, more liable to damage, and more difficult to replace than fur. Every other creature that has evolved a thermoregulatory coat‹from bats to bees and from caterpillars to pterosaurs‹has some kind of furry cover. There is no apparent reason for evolving feathers rather than fur even for heat shielding.

Even within birds, down feathers are much better for retaining heat than the contour feathers that are preadaptive to flight. An adult emperor penguin has very efficient thermoregulatory feathers, but they must also be water-resistant and hydrodynamically efficient. But an emperor penguin chick does not fly, swim, or even walk very much. Its primary need is to survive in the dark on the Antarctic ice cap without a nest, in temperatures that average around -25°C (-13°F), and in winds of 40 meters per second (100 mph). Its first feathers are molted and replaced before it needs them for any other function, so they can be the most efficient feathers evolved for thermoregulation alone. The emperor penguin chick has down feathers. They are nothing like flight feathers, display feathers, or the feathers of Archaeopteryx, and they are developed equally over the body except for the wings and feet, where they are shorter than normal rather than longer.

Thermoregulation cannot account for the length or the distribution of the earliest known feathers, those of Archaeopteryx. Thermoregulation would require feathers developed equally over the whole body, whereas Archaeopteryx had its longest, strongest feathers on the wings and tail. Thermoregulation can be achieved perfectly well with short feathers: it does not require the long feathers of Archaeopteryx.

So it is difficult to suggest that feathers evolved for thermoregulation without also arguing that the feathers of Archaeopteryx had already been evolved for some other function or functions and then modified. And once that argument is made, the hypothesis of thermoregulation becomes untestable on present evidence. It would be better to think of another equally simple explanation of the feather pattern of Archaeopteryx.

I naturally prefer an idea that I developed jointly with my colleague Jere Lipps of the University of California, Berkeley. In living birds, feathers are for flying, for insulation, but also for camouflage and/or display. Lipps and I suggest that feathers evolved first for display. The display may have been between females or between males for dominance in mating systems (sexual selection), or between individuals for territory or food (social selection), or directed toward members of other species in defense of territory or food.

Living reptiles and birds often display for one or all of these reasons, using color, motion, and posture as visual signals to an opponent. Display is often used to increase apparent body size; the smaller the animal, the more effectively a slight addition to its outline would increase its apparent size. Lipps and I therefore proposed that replacing scales with erectile, colored feathers would give such a selective advantage to a small displaying theropod that it would encourage a rapid transition from a scaly skin to a coat of feathers. Display would be most effective on movable appendages, such as forearms and tail. Display on the legs would not be so visible or effective. Forearm display by a small theropod would also have drawn particular attention to the powerful weapons it carried there, its front claws (Figure 13.18).

The display hypothesis explains more features of Archaeopteryx than other hypotheses, with fewer assumptions. It explains completely the feather pattern of Archaeopteryx. It explains why the feather impressions are so faint on the smallest specimen of Archaeopteryx, which may not have reached full adult size or status. This specimen is only about half the size of the others and has no wishbone preserved, possibly because it had not yet ossified. The display hypothesis assumes only that display was important to Archaeopteryx: it assumes nothing special about its habits, habitat, or body temperature.

The Origin of Powered Flight in Birds

Most people feel sure that protofeathers must already have been evolved before birds attempted flight. The question of the origin of flight is thus independent of the origin of feathers, because flight evolved in bats and pterosaurs without feathers. There have been three important hypotheses for the origin of bird flight, and I shall add a fourth.

The Arboreal Hypothesis
The arboreal hypothesis suggests that a ground-running biped first became adapted to life in trees, where it took to leaping from branch to branch, then parachuting. Later it developed flapping flight. Feathers became aerodynamically important at the jumping stage and evolved directly into flight feathers. The arboreal theory is the most favored at the moment, but it does have some difficulties.

Like all little theropods, Archaeopteryx was bipedal, with legs and feet that were well adapted for ground running. Bipediality is a rather poor preadaptation for living in trees, and Archaeopteryx had long, erect hind limbs that were particularly ill-suited to climbing tree trunks (some arboreal theorists suggest that it climbed sloping branches instead!). Archaeopteryx, with long, erect limbs, a comparatively short trunk, and bipedal locomotion, was exactly the opposite in body plan of all living mammals and reptiles that jump and glide from tree to tree.

The claws on the hands of Archaeopteryx were long, thin, and sharp. They look like very effective tearing and slicing weapons, but were far too sharp and pointed to have been useful for climbing either trees or rocks. The claws on its feet have been compared with the claws on the feet of perching birds, but they were also very like the talons of an eagle or a theropod dinosaur, which only shows that they were equally well adapted for clutching branches or prey, or both, and we cannot tell which. Certainly theropod dinosaurs with "perching" claws did not climb trees (Figure 12.5).

Altogether, the arboreal hypothesis is not unreasonable, but it does require a lot of special conditions. It looks vulnerable to a better suggestion that would explain more of the evidence.

The Cursorial Hypothesis
Perhaps some adaptations in a ground-dwelling theropod could provide some of the anatomy and behavior necessary for flight, such as lengthening the forearms, especially the hands, placing long, strong feathers in those areas, and evolving powerful arm movements. An early version of the cursorial hypothesis suggested that a fast-running reptile might evolve long scales on the arms. In this theory, the scales generated lift as the arms were actively flapped on the run (Figure 13.19). The animal could now take long leaps, perhaps encouraging the scales to evolve into feathers and the leaps to evolve into powered flapping flight.

We know now that feathers did not evolve from scales, but in any case this idea does not work mechanically. Any lift generated by a flapping arm decreases the ground traction given by the feet, and acceleration is lost. A racing car is held down on the track by its airfoils for good traction, and an aircraft cannot be driven through its wheels on the takeoff run. A protobird that flapped its arms on the run would increase drag: the faster the running, the greater the drag. Only a very small amount of thrust would have been generated by the arms in the early stages of the process. Running takes a lot of energy, and it's not clear why leaping would have benefited the animal.

Both the arboreal and cursorial hypotheses must face the problem of changing from parachuting flight (from a jump off a branch or a leap into the air from the ground) to true powered, flapping flight. Flapping arms or protowings -- in fact, any feathers at all on wings or tail -- increase drag. Aerodynamically, the transition from gliding to flapping is difficult: there is only a narrow theoretical window through which the transition could have been made. This transition would have been very difficult for Archaeopteryx because it has such a long, bony tail with long feathers on it. Such a tail (an obvious display structure in my opinion) adds much more drag than it adds lift. Therefore the arboreal and cursorial hypotheses are not impossible, but they invite a better idea.

The Running Raptor
More sophisticated recent versions of the cursorial hypothesis are much better: they are mechanically sounder and include behavior that involved strong, synchronized arm strokes and the evolution of strong pectoral muscles.

John Ostrom suggested that the protobird was a fast-running hunter, perhaps using its arms to strike down insects that it disturbed. Such an action could encourage the evolution of the muscles and the joint movements that would approximate a wing-stroke. No living bird catches insects this way, however, probably because a feathered wing generates an air blast, while a well-designed fly-swatter has holes in it to avoid blowing away the prey. Some egrets scare fish into motion by wing flapping, but not at a run, and the wings are not used in strikes.

Gerald Caple and his colleagues at Northern Arizona University suggested instead that a protobird hunted by running fast and leaping after flying or jumping insects it disturbed. To catch an elusive dodging prey while its feet were off the ground, a protobird would have to be able to adjust its body attitude in the air and then regain a stable position for landing. Such adjustments could be made aerodynamically by generating a small amount of lift or drag at appropriate points on the body surface. Calculations suggest that a small amount of lift at the tips of the arms would have a large effect on the body as a whole, and if the right arm movements were added, the effect would be greater still. The protobird would now be well on the way to flapping takeoff, and the flights would be gradually prolonged until complete control had been reached.

But this proposed activity would consume a lot of energy. No predator today, bird or otherwise, runs at high speed to flush out insects it can leap after. Furthermore, effective attitude control for a leaping animal is not possible below a critical airspeed that is highest in the early stages of the process, when the proto-wings are just beginning to generate lift. The required speed might have been 10 meters per second, over 25 mph, far too much for any reasonable protobird. The idea cannot explain the evolution of flight on its own, but it may have components that could apply to later stages in the evolution of flapping flight.

A new and powerful argument for the evolution of flight among fast-running protobirds is related to Carrier's Constraint (Chapter 11). Powered flapping flight demands a sustained high-energy output, so animals that operate it have to solve Carrier's Constraint. Flying insects pump air in and out of their spiracles in synchrony with their wingbeats. In flying vertebrates, the muscles that flap the wing are anchored on the ribcage, and expand and contract the chest cavity with each wingbeat. Fruit bats, vampire bats, and pigeons take exactly one breath per wingbeat, big geese take one breath every three wingbeats, and pheasants and ducks take one breath every five wingbeats. Perhaps, then, a bipedal reptile running rapidly on the ground with erect limbs would already have its breathing synchronized with its running, and it would have a high metabolic rate and the capacity for sustained power output. Such a preadaptation for powered flight would be more likely to occur in a fast, bipedal runner than in a jumping, quadrupedal tree-dweller.

The Display and Fighting Hypothesis
Either the arboreal or the cursorial hypothesis would work, and work much more easily, if a protobird already had long, strong feathers in the right places and already had powerful arm movements. Jere Lipps and I suggested that display was involved in the evolution of flapping flight as well as in the evolution of feathers. Display provided long, strong feathers on arms and tail. Successful display was increased by lengthening the arms, especially the hand, and by actively waving them, perhaps flapping them rapidly and vigorously. Flapping in display would have encouraged the evolution of powerful pectoral muscles.

But a threat display must not be seen as an empty bluff. Fighting is the last resort. Living birds often fight on the ground, even those that fly well. Wings are no longer clawed but are still used as weapons in forward and downward smashes (steamer ducks are particularly deadly at this). Beaks and feet can be used as weapons too and are most effective when used in a downward or forward strike.

A strong wing flap, directed forward and downward, is also the power stroke that gives lift to a bird in takeoff. Lipps and I suggest that strong wing flapping is a simple extension of display flapping, encouraged by fighting behavior. Powerful flapping used to deliver forearm smashes could have lifted the bird off the ground, allowing it also to rake its opponent from above with its hind claws. The more rapidly the wings could be lifted for another blow, the more effective the fighting. This would rapidly encourage an effective wing-lifting motion that minimized air resistance, so the wing action would then be almost identical to a takeoff stroke.

A variant of our idea has also been proposed by Kevin Padian, who prefers to think of the wing stroke evolving from the arm strike used by a theropod in predation. It is not clear how this could have led easily to whole-body takeoff, however.

A few living birds use their wings extensively as weapons. The steamer ducks of the South Atlantic are large, powerful birds with heavy, bright orange, horny knobs on the wings of both sexes. These are used by both sexes in display and fighting. Steamer ducks (especially males) fight a lot among themselves for mates and territory, and they often kill other species of water birds, holding them by the neck and beating them to death with the wing knobs. Some species of steamer ducks are flightless; in other species, the males are often too massive to fly, even though juveniles and females can fly well. Selection has favored fighting ability over flying ability for many steamer ducks. Flight is perhaps less important for them than for many birds, because they live in shoreline habitats where food is plentiful all year round.

Archaeopteryx fits our display-and-fighting hypothesis well. It was well adapted for display. Like any small theropod, it was well equipped for fighting with its teeth and the strong claws on hands and feet. Archaeopteryx did not have long primary feathers on its fingers (Figure 13.8), probably because they would have hidden the claws in display and would most likely have broken in a fight.

Did Archaeopteryx Fly?
Archaeopteryx was at best a poor flier. If it flew at all, it could not have sustained flapping flight for long. In small flying birds today, the wishbone acts as a spring that repositions the shoulder joints after the stresses of each wing stroke. It probably helps to pump air in and out of the lungs, and to recover some of the muscular energy put into the downstroke. In Archaeopteryx and in other theropods with wishbones, the bone is U-shaped and strong and solid; it could not have acted as an effective spring. Archaeopteryx did not have the pulley system of the shoulder that gives a rapid upstroke (Figure 13.16), and it must have had particular difficulty flying at low speed, and in take-off or landing. Its shoulder joint did not allow it to raise its arms very far above the horizontal, so the downstroke could not have been powerful. Furthermore, Archaeopteryx did not have the long primary feathers on the wing tips, or the breastbone anchoring the chief lifting muscles for the wing that are needed for routine takeoff and landing on the ground. In fact, it evolved structures that were active deterrents to flight. Its tail was long and bony, with long feathers. Among living birds with display feathers, this sort of tail is aerodynamically the worst of all possible tail styles, adding a lot of drag and little lift.

Archaeopteryx, then, was a fierce little fast-running, displaying bird, which probably spent its life scurrying around the Solnhofen shore, hunting for small prey such as crustaceans, reptiles, and mammals. In hunting style, Archaeopteryx was probably much like the roadrunner of the dry country of the American Southwest, but its ecological setting was like that of a steamer duck -- a shoreline with year-round food supply. Archaeopteryx did not compete in the air with the pterosaurs that are also found in the Solnhofen Limestone.

From Display to Flight
In our theory, display and flighting were simple selective agents that encouraged the evolutionary transition from small dinosaurs to birds. The idea fits with our current knowledge of the biology and behavior of living birds. Display, and fighting if necessary, is very important, even within a species. Bald eagles and frigate birds often try to rob other birds of food instead of catching prey themselves. Because the penalty for wing injury is high, many birds can be intimidated by display into giving up their catch rather than fighting to defend it.

Display and fighting in birds, whether it's for territory, dominance, or food, takes a lot of energy, but only for brief periods or seasons, and it provides an neormous payoff in survival and selection. Sexual display in most living birds must be done correctly, or no mating takes place. New behaviors are quick to evolve, and they are evolutionarily cheap, because they usually don't require important morphological changes in their early stages. Bowerbirds, for example, show distinct behavioral differences in display between closely related species.

The display hypothesis suggests that a protobird gained flight behavior, anatomy, and experience at low ground speed and low height, ideal preflight training. The selective payoff for successful mastery of the flight motions gave significant advantages, even before flight itself was possible. From that point, the many advantages of flight were added to those of social or sexual competition.

Lipps and I envisage Archaeopteryx as a small, fierce predator, capable of liftoff but not true flight. Once liftoff was achieved, flapping flight quickly followed. There is no need to suggest any difficult evolutionary sequence to complete the final transition to full powered flight. In more advanced birds than Archaeopteryx, the pulley system of the shoulder evolved for quick wing upstrokes, while the wishbone evolved into a spring. The breastbone evolved as the anchor for the flight muscles. The forearms became longer, lighter, and more fragile in bone structure, becoming specialized as wings, and losing the finger claws. Meanwhile, the feet and beak became the dominant fighting weapons, as in most living birds today.

The display hypothesis for the origin of flight is particularlty attractive because it demands few of the assumptions required by the arboreal or cursorial hypothesis.

EXTRACT FROM HISTORY OF LIFE, 1st edition (1990), Chapter 12:

The Origin of Feathers

Feathers in living birds originate in a deep skin layer deep under the outer layer that forms scales. Evolutionarily, then, feathers probably arose under and between reptile scales, not as modified scales. Many birds have scales on their lower legs and feet where feathers are not developed, and penguins have such short feathers on parts of their wings that the skin there is scaly for all practical purposes. So there is no real anatomical problem in imagining the evolution of feathers on a reptilian skin. But feathers are completely novel structures, and any reasonable explanation of their origin has to take this into account. They evolved in the first birds to replace scales as the primary skin covering. The problem is to reconstruct why this happened, and why it seems to have happened only once in the entire history of the reptiles.

Feathers may have evolved directly for flight. If so, feathers evolved in a reptile that was already launching itself into the air, presumably in a tree-dwelling jumper and perhaps parachuter. In this hypothesis, feathers were first an aid to parachuting and then a way to achieve flapping flight. This is a difficult process to imagine. One asks, "Why feathers?" The scales of gliding reptiles do not project beyond the boundary layer of air around the body, so they do not generate any lift. It's not clear that protofeathers would have been any improvement. The bone-supported skin membranes of parachuting reptiles are a much easier and cheaper way to evolve an airfoil than any conceivable airfoil made of protofeathers. Bats and pterosaurs evolved flapping flight without feathers. Perhaps feathers evolved for some other function and were later modified for flight.

Feathers may have evolved to aid thermoregulation. Small theropods probably had a high metabolic rate and may have been warm-blooded (Chapter 12). Very small theropods would have needed additional insulation to keep their bodies at even temperature. A few small reptiles today use long scales to help trap a layer of air between the environment and the body surface to cut down temperature fluctuations, usually as a heat shield against the sun. It would not matter whether protofeathers were needed to conserve heat in cold periods, or to keep heat out in hot periods, or both. In either case their insulation would have been useful. The skin musculature would have been able to raise and lower protofeathers, allowing free flow of air to the skin when necessary.

This theory for the origin of feathers is probably the most widely accepted one today, but it does have problems. One asks again, "Why feathers?" Feathers are more complex to grow, more difficult to maintain in good condition, more liable to damage, and more difficult to replace than fur. Every other creature that has evolved a thermoregulatory coat, from bats to bees and from caterpillars to pterosaurs, has some kind of furry cover. There is no apparent reason for evolving feathers rather than fur even for heat shielding.

Even within birds, it's easy to show that at least for retaining heat, down feathers are better than the contour feathers that are preadaptive to flight. An adult emperor penguin has very efficient thermoregulatory feathers, but they must also be water-resistant and hydrodynamically efficient. But an emperor penguin chick does not fly, or swim, or even walk very much. Its primary need is to survive in the dark on the ice cap of Antarctica without a nest, in temperatures that average around -25°C (-13°F), and in winds of 40 meters per second (100 mph). Its first feathers are molted and replaced before it needs them for any other function, so they can be the most efficient feathers that can be evolved for thermoregulation alone. The emperor penguin chick has down feathers. They are nothing like flight feathers, display feathers, or the feathers of Archaeopteryx, and they are developed equally over the body except for the wings and feet, where they are shorter than normal rather than longer.

Thermoregulation cannot account for the length or the distribution of the earliest known feathers, those of Archaeopteryx. Thermoregulation would require feathers developed equally well over the whole body, whereas Archaeopteryx had its longest, strongest feathers on the wings and tail. Thermoregulation can be achieved perfectly well with short feathers: it does not require the long feathers of Archaeopteryx.

Thus it is difficult to suggest that feathers evolved for thermoregulation without also arguing that the feathers of Archaeopteryx had already been evolved and then modified (for some other function or functions). And once that argument is made, the hypothesis of thermoregulation becomes untestable on present evidence. It would be better to think of an alternative explanation that did describe the feather pattern of Archaeopteryx with an equally simple idea

I naturally prefer an idea that was developed jointly by me and my colleague Jere H. Lipps of the University of California, Berkeley. In living birds, feathers are for flying, for insulation, but also for camouflage and/or display. Lipps and I suggest that feathers evolved first for display. It doesn't matter whether the display was between females or between males for dominance in mating systems (sexual selection), or between individuals for territory or food (social selection), or directed toward members of other species in defense of territory or food.

Living reptiles and birds often display for one or all of these reasons, using color, motion, and posture as visual signals to an opponent. Display is often structured to increase apparent body size; the smaller an animal, the more effectively a slight addition to its outline would increase its apparent size. Lipps and I therefore propose that elongating, or, more likely, replacing scales with erectile, colored feathers would give such a selective advantage to a small displaying theropod that it would encourage a rapid transition from a scaly skin to a coat of feathers. Display would be most effective on movable appendages, such as forearms and tail. Display on the legs would not be so visible or effective. Forearm display by a small theropod would also have drawn particular attention to the powerful weapons it carried there, its front claws (Figure 12.20).

The display hypothesis explains more features of Archaeopteryx than its competitors, with fewer assumptions. It explains completely the feather pattern of Archaeopteryx. It explains why the feather impressions are so faint on the smallest specimen of Archaeopteryx, which may not have reached full adult size or status. The "Eichstatt specimen" is only about half the size of the "Solnhofen specimen" and has no wishbone preserved, possibly because it had not yet ossified. The display hypothesis assumes only that display was important to Archaeopteryx: it assumes nothing special about its habits, its habitat, or its body temperature.

The Origin of Flapping Flight in Birds

Most people feel sure that protofeathers must already have been evolved before birds attempted flight. The question of the origin of flight is thus independent of the origin of feathers, since flight evolved in bats and pterosaurs without feathers. There have been three important hypotheses for the origin of bird flight, and I shall add a fourth.

The Arboreal Theory
The arboreal theory suggests that a ground-running biped first became adapted to life in trees, where it took to leaping from branch to branch, then parachuting. Later it developed flapping flight. Feathers became aerodynamically important at the jumping stage and evolved directly into flight feathers. The arboreal theory is the most favored at the moment, but it does have some difficulties.

Archaeopteryx had hind limbs and feet that look well adapted for ground running, but poorly adapted for perching. Bipedality is a poor preadaptation for living in trees, and Archaeopteryx had long, erect hind limbs that were particularly ill-suited to climbing tree trunks (some arboreal theorists suggest that it climbed sloping branches instead!). Archaeopteryx, with its long, erect limbs, its comparatively short trunk, and its bipedal locomotion, is exactly the opposite in body plan of all living mammals and reptiles that jump and glide from tree to tree.

Archaeopteryx was at best a poor flier, and could not have sustained flapping flight for long, if it flew at all. In small flying birds today the wishbone acts as a spring. It repositions the shoulder joints after the stresses of each wing stroke, it probably helps to pump air in and out of the lungs, and it probably recovers some of the muscular energy put into the downstroke. In Archaeopteryx, and in those theropods that have a wishbone, the bone is U-shaped and strong and solid, and could not have acted as an effective spring. Archaeopteryx did not have the pulley system of the shoulder that gives a rapid upstroke (Figure 12.18), and it must have had particular difficulty flying at low speed, or in take-off or landing. If it launched from branches, it must have climbed upward rather than flapping up into the trees. It would have had great difficulty landing accurately on a branch, so probably had to land on the ground. And there are no large pieces of wood fossilized in the Solnhofen rocks, showing that there were no large trees anywhere near the shoreline. Thus arboreal theorists often talk about broken terrain in which a protobird would clamber around and over large rocks and gullies.

The claws on the hands of Archaeopteryx are very long, very thin, and very sharp. They look like very effective tearing and slicing weapons, but are far too sharp and pointed to have been useful for climbing either trees or rocks. In contrast, the claws on the feet are shorter and longer.

Altogether, the arboreal hypothesis is not unreasonable, but it does require a lot of special conditions. It looks vulnerable to a better suggestion that would explain more of the evidence.

The Cursorial Theory
Perhaps some adaptations in a ground-dwelling theropod could provide some of the anatomy and behavior necessary for flight, such as lengthening the forearms, especially the hands, placing long, strong feathers in those areas, and evolving powerful arm movements. An early version of the cursorial hypothesis suggested that a fast-running reptile might evolve long scales on the arms. In this theory, the scales generated lift as the arms were actively flapped on the run (Figure 12.21). The animal could now take long leaps, perhaps encouraging the scales to evolve into feathers and the leaps to evolve into powered flapping flight.

We know now that feathers did not evolve from scales, but in any case this idea does not work mechanically. Any lift generated by a flapping arm decreases the ground traction given by the feet, and acceleration is lost. A racing car is held down on the track by its airfoils for good traction, and an aircraft cannot be driven through its wheels on the takeoff run. A protobird that flapped its arms on the run would increase drag: the faster the running, the greater the drag. Only a very small amount of thrust would have been generated by the arms in the early stages of the process. Running takes a lot of energy, and it's not clear why leaping would have benefited the animal.

Both the arboreal and cursorial theories described so far must face the problem of changing from a gliding flight (from a jump off a branch or a leap into the air from the ground) to true powered, flapping flight. Flapping arms or protowings increases drag. Aerodynamically, the transition from gliding to flapping is difficult: there is only a narrow theoretical window through which the transition could have been made. The transition is not impossible, but it means once again that the theories outlined so far are vulnerable to a better idea.

The Running Raptor
More sophisticated recent versions of the cursorial hypothesis are much better: they are mechanically sounder and include behavior that involved strong, synchronized arm strokes and the evolution of strong pectoral muscles.

John Ostrom of Yale University suggested that the protobird was a fast-running hunter, perhaps using its arms to strike down insects that it disturbed. Such an action could encourage the evolution of the muscles and the joint movements that would approximate a wing-stroke. However, no living bird catches insects this way, probably because a feathered wing generates an air blast, while a well-designed fly-swatter has holes in it to avoid blowing away the prey. Some egrets scare fish into motion by wing flapping, but not at a run, and the wings are not used in strikes.

Gerald Caple and his colleagues at Northern Arizona University suggested instead that a protobird hunted by running fast, leaping after flying or jumping insects it disturbed. To catch an elusive dodging prey while its feet were off the ground, a protobird would have to be able to adjust its body attitude in the air and then regain a stable position for landing. Such adjustments could be made aerodynamically by generating a small amount of lift or drag at appropriate points on the body surface. Calculations suggest that a small amount of lift at the tips of the arms would have a large effect on the body as a whole, and if the right arm movements were added, the effect would be greater still. The proto-bird would now be well on the way to flapping takeoff, and the flights would be gradually prolonged until complete control had been reached.

But this proposed activity is very energy-consuming. No insect predator today runs at high speed trying to flush out insects it can leap after. Furthermore, effective attitude control for a leaping animal is not possible below a critical airspeed that is highest in the early stages of the process, when the proto-wings are just beginning to generate lift. The required speed might have been 10 meters per second, over 25 mph, far too much for any reasonable proto-bird. The idea cannot explain the evolution of flight on its own, but it may have components that are useful during a second stage in the evolution of flapping flight.

A new and powerful argument for the evolution of flight among fast-running protobirds is related to Carrier's Constraint (see Chapters 9 and 10). Powered flapping flight demands a sustained high-energy output, and so animals that operate it have to solve Carrier's Constraint. So flying insects pump air in and out of their spiracles in synchrony with their wingbeats. In flying vertebrates, the muscles that flap the wing are anchored on the ribcage, and expand and contract the chest cavity with each wingbeat. Fruit bats, vampire bats, and pigeons take exactly one breath per wingbeat, big geese take one breath every three wingbeats, and pheasants and ducks take one breath every five wingbeats. Perhaps, then, a bipedal reptile running rapidly on the ground with erect limbs would already have its breathing synchronized with its running, and it would have a high metabolic rate and the capacity for sustained power output. Such a preadaptation for powered flight would be more likely to occur in a fast, bipedal runner than in a jumping, quadrupedal tree-dweller.

The Display and Fighting Hypothesis
Either the arboreal or the cursorial hypothesis would work, and work much more easily, if a protobird already had long, strong feathers in the right places and already had powerful arm movements. Jere Lipps and I suggested that display was involved in the evolution of flapping flight as well as in the evolution of feathers. Display provided long, strong feathers on arms and tail. Successful display was increased by lengthening the arms, especially the hand, and by actively waving them, perhaps flapping them rapidly and vigorously. Flapping in display would have encouraged the evolution of powerful pectoral muscles.

But there is another implication of display. A threat display must not be seen as an empty bluff. A last resort is to fighting. Living birds often fight on the ground, even those that fly well. Wings are no longer clawed but are still used as weapons in forward and downward smashes (steamer ducks are particularly deadly at this). Beaks and feet can be used as weapons too, and are most effective when used in a downward or forward strike.

A strong wing flap, directed forward and downward, is also the power stroke that gives lift to a bird in takeoff. Lipps and I suggest that strong wing flapping, in a simple escalation of display flapping and forearm smashes, could have lifted the bird off the ground, allowing it to rake its opponent from above with its hind claws. The more rapidly the wings could be lifted for another blow, the more effective the fighting. This would rapidly encourage an effective wing-lifting motion that minimized air resistance, so the wing action would then be almost identical to a takeoff stroke.

Lipps and I envisage the protobird as a small, fierce predator, capable of liftoff but not true flight. Once liftoff was achieved, we suggest that flapping flight quickly followed. The pulley system of the shoulder evolved for quick wing upstrokes, accompanied by the evolution of the wishbone into a spring. The breastbone evolved as the anchor for the flight muscles. The forearms became longer, lighter, and more fragile in bone structure, and became specialized as wings, losing the finger claws. Meanwhile the feet and beak became the dominant fighting weapons. Today, for example, the northern jacana, a storklike bird, has large bony spurs on its wings. The spurs look like effective weapons; they are accentuated by color, being yellow against a maroon background, and they are used in threat display. The female in this species is much the larger sex, and it is females that display and attract males as mates. But the bony spurs of the jacana are only for display; when it comes to fighting, she uses the claws on her feet.

A few living birds use their wings extensively as weapons. The steamer ducks of the South Atlantic are large heavy birds with powerful heads and necks. They have heavy, bright orange horny knobs on the wings of both sexes. These are used by both sexes in display and fighting. Steamer ducks (especially males) fight a lot among themselves for mates and territory, and they often kill other species of water birds, holding them by the neck and beating them to death, using the wing knobs to cause internal hemorrhages. Some species of steamer ducks are flightless. In other species, the males are often too massive to fly, even though juveniles and females can fly well. Selection has favored fighting ability over flying ability for many steamer ducks. Flight is perhaps less important for them than for many birds, because they live in shoreline habitats where food is plentiful all year round for birds that are able to hold a territory. The African black duck has much the same accent on fighting ability, in a different habitat on a different continent, but also where food can be found year-round without flight.

Archaeopteryx fits our hypothesis well. It was well adapted for display, as we have seen, and like any small theropod, it was well adapted for fighting with its teeth and the strong claws on hands and feet. Archaeopteryx does not have long primary feathers on its fingers (Figure 12.20), probably because they would hide the claws in display and would most likely break in a fight.

Archaeopteryx does not have the shoulder structure, the long primary feathers on the wing tips, or the breastbone anchoring the chief lifting muscles for the wing that are needed for routine takeoff and landing on the ground. It must at best have been a poor flier. But we need not suggest any difficult evolutionary sequence to complete the final transition to full powered flight. Archaeopteryx fits our hypothesis as a fierce little fast-running, displaying bird, which probably spent its life scurrying around the Solnhofen shore, opportunistically hunting for small prey such as crustaceans, reptiles, and mammals. In hunting style, Archaeopteryx was probably much like the roadrunner of the dry country of the American Southwest, but its ecological setting was like that of a steamer duck, on a shoreline with year-round food supply. Archaeopteryx did not compete in the air with the pterosaurs that are also found in the Solnhofen Limestone.

In our theory, display and flighting were simple selective agents that encouraged the evolutionary transition from small dinosaurs to birds. The idea fits with our current knowledge of the biology and behavior of living birds. Display, and fighting if necessary, is very important, even within a species. Bald eagles, for example, often try to rob other individuals for food instead of catching prey themselves. Because the penalty for wing injury is high, many birds can be intimidated by display into giving up their catch rather than fighting to defend it. Display and fighting in birds, whether it's for territory, dominance, or food, takes a lot of energy, but only for brief periods or seasons, and it provides an enormous payoff in survival and selection. Sexual display in most living birds must be done correctly, or no mating takes place. New behaviors are not only fast to evolve but evolutionarily cheap, because they usually don't require important morphological changes in their early stages. Bowerbirds, for example, show distinct behavioral differences in display between closely related species.

The display hypothesis suggests that a protobird gained flight behavior, flight anatomy, and flight experience at very low ground speed and very low height: ideal preflight training. The selective payoff for successful mastery of the flight motions would have been very significant even before flight itself was possible. From that point, the many advantages of flight were added to those of social or sexual competition.

The display hypothesis for the origin of flight is particularly attractive because it demands few of the additional assumptions required by the arboreal or cursorial hypotheses. Once display and fighting had selected for protobirds with large strong feathers and coordinated powerful wingbeats, it is simple to envisage the final evolution of powered flight in either a ground-running or a tree-dwelling bird.

EXTRACTS FROM Cowen, R., and Lipps, J. H. 1982.

We present here a scenario. It is a hypothesis which relates the evolution of feathers, of birds, and of flight in birds, to one selective theme of display and intraspecific threat and combat. This simple theme, we believe, is the key to understanding the dramatic evolutionary transition between vertebrate classes and between terrestrially- and aerially-dominant ways of life...

The Origin of Feathers

Current Hypothesis for the Origin of Feathers
Ostrom (1974) and Regal (1975) argued that feathers arose from an archosaurian scale as insulating, thermoregulatory features. Ostrom believed that feathers were an insulating coat to keep a small warm-blooded animal warm; Regal thought that feathers were originally insulating devices to keep a small animal cool in high radiation.

The only good evidence is the feathers of Archaeopteryx and the possible feather fossil Praeornis. The body covering of Archaeopteryx is faint and ambiguous, but probably consists of feathers. However, the thermoregulatory hypothesis alone cannot explain the length and strength of the large feathers that form the "wing" on the forelimb and the large vane on the tail of Archaeopteryx. Some other selective factor must have operated to give rise to feathers strong and large enough to have become pre-adaptive to flight.

The Display Hypothesis
Display is a vital function of the feathers in very many living birds. Display translates directly into reproductive success via the intraspecific communication used in sexual selection, and it is important in repelling predators. We propose, as a working hypothesis, that display was important in the evolution of feathers from scales.

Feathers are scales which have been elongated and raised as large planar structures. A sophisticated dermal musculature allows them to be raised, lowered, and waved. Birds go to great lengths to maintain plumage in good condition. All these attributes are interpreted as related to display (and some other functions). We suggest that the enlargement and the planar structure, the motility and the maintenance of plumage all arose as part of a display function during the first appearance of proto-feathers. In particular, if the archosaurian ancestor already had skin or scale display (as many living reptiles do), we expect that selection, acting on the new, improved display devices, induced the very rapid evolution of feathers (and of birds).

Testing the Display Hypothesis
Our hypothesis and the thermoregulatory hypothesis cannot be tested directly in the fossil record, but we can search for morphological correlates.

1. If feathers arose for display, the should be most strongly developed on areas that are prominent and easily displayed. They should be movable for maximum effect, to increase apparent size and to draw attention to agility. If display arose in a small bipedal archosaur, it would reasonably have been best developed on the arms and tail, appendages that are comparatively high on the organism, and less intimately associated with locomotion and feeding than the feet and head. Evidence from Archaeopteryx allows the display hypothesis to pass this test.

2. If display feathers arose, some of them might be similar to display feathers in modern birds, as opposed to down or flight feathers, for example. Praeornis, from the later Jurassic of Kazakhstan (Rautian 1978), may not be a feather, but if it is it can only be a display feather.

3. As Geist (1966) and others have documented for mammals, it is common experience with birds that the visual aspect of adults and juveniles is very different. In birds, this has nothing to do with thermoregulation or with flight, but reflects the important of intraspecific communication accomplished with feathers. If display was important in the early evolution of feathers, Archaeopteryx might show an adult/juvenile difference. This test would provide a sharp distinction between the display hypothesis and the thermoregulatory hypothesis.

The "Eichstatt specimen" of Archaeopteryx is smaller than the other four skeletons, and according to Wellnhofer (1974, 1976) it is not as strongly ossified as they are. Its feather impressions are so weak that they were not seen for 20 years after the fossil was found. Yet the specimen is not a chick: it is about 60% of adult size. The Eichstatt specimen could be a large juvenile which had not yet grown the strong plumage of an adult. It does have plumage but it does not have display feathers.

We agree that thermoregulation was an important function of proto-feathers. But we argue that display was very important, and that this scenario is testable and so far survives testing. The relative merits of the thermoregulatory and display hypotheses are less important than to recognize that they had a synergistic selective effect acting to produce well-developed feathers. Feathers have properties which promote thermoregulation and display without detriment to either function. When the demonstrated effectiveness of sexual selection in driving evolutionary change and the correlation between display and sexual selection in living birds is recognized, the apparently sudden appearance of feathers and of birds in the fossil record is not surprising.

The Evolution of Flight in Birds

Previous Hypotheses for the Origin of Flight in Birds
Protofeathers must already be in existence before flight can be attempted. Thus the question of the origin of flight is independent of the question of the origin of feathers. So far, three important hypotheses for the origin of flight in birds have been proposed: the arboreal, the cursorial, and the cursorial-predator hypotheses. We propose a new alternative: the display and threat hypothesis. It is, of course, consistent with the model we suggest for the origin of strong feathers, and requires no shift in selective mechanism.

The arboreal hypothesis, first suggested by March (1880), elaborated by Heilmann (1926) and Bock (1965, 1969), is also advocated by Feduccia (1980). It proposes that a bipedal ground-runner first became adapted to an arboreal life in which it took to jumping from limb to limb, then to gliding, and finally to flying. In this scenario, contour feathers became aerodynamically important at a jumping stage, and thence evolved directly into flight feathers. Ostrom (1974, 1979) cogently argues against these ideas. He interprets the hindlimb of Archaeopteryx as adapted for ground-running, not for perching. Furthermore, no non-flying tree-dweller today is a biped, which implies that bipedality is a poor preadaptation for arboreal life, and thus for evolution through an arboreal stage. We accept these arguments.

The cursorial hypothesis is traced to Williston (1879) and received its modern form from Nopsca (1907, 1923). Reptilian running bipeds, through elongation of the forelimbs and development of scales on the arms, formed thrust surfaces which provided acceleration when flapped, gradually leading to long leaps and flapping flight. Criticisms of this idea include: 1. flapping tends to decrease ground traction and acceleration, thus counteracting thrust from the legs; 2. elaboration of arm surfaces would at first only increase drag; 3. only a tiny amount of thrust would have been generated by an incipient wing (Ostrom 1979).

Ostrom (1974, 1979) proposed that avian flight arose from a cursorial predator whose feathers were selected for greater length and strength through their use as a sort of "insect net". The sweeping action of the arms necessary to trap insects in the feathers would have led to more and more powerful adduction, to assaults on larger and larger prey, to higher and higher leaps into the air after flying insects, and so to flapping flight. Ostrom believed that Archaeopteryx was just at or just past the threshold of powered flight, as it lacks a sternum and strut-like coracoid as well as some minor features of the skeleton which are associated with flight in living birds. We believe that the hypothesis is inadequate because 1. no living creature operated an insect-catching device of the type suggested; 2. an insect net would have caused drag while the animal was running (Ostrom suggested that the arms were held at the sides until the last moment of the chase); 3. the wing beats required to operate the "net" differ from those required for leaping and flying, thus requiring a rapid functional shift for which we see no selective agent in the proposed model; and 4. the model requires synchronized wing action, yet we believe that a single-wingd motion would have been more plausible for catching insects.

The New Hypothesis: Display and Threat
We suggest that feathers developed synergistically as display and thermoregulatory structures. We propose also that flight itself arose in birds as a direct response to selection related to display and male-to-male threat and fighting. Archaeopteryx can receive full morphological interpretation in this model.

Flying birds have long forelimbs, and wing length is closely related to body size, presumably fof reasons connected with flight. Such channeled selection would not have operated on bipedal proto-birds, yet considerable arm-lengthening must have taken place in the lineage leading to Archaeopteryx. We suggest that this arm-lengthening provided more surface area for display, and that it was pre-adaptive to flight. Archaeopteryx has a long reptilian tail with strong feathers attached. The tail, originally held prominently as an inertial balancing organ, served as a display panel, and its feathers increased in size, strength, erectility, and possible coloration as display was accentuated. Once the strong tail display feathers had evolved, selection for fast running in small bipeds would have acted to produce aerodynamic rather than inertial stabilizers.

Display of males to females is commonly interwoven with male-to-male display and confrontation, with an ultimate resort to fighting. In living birds such fights often take place on the ground even when opponents are capable of flight. A stable bipedal stance is important, as is a position above an opponent. Selection would operate to provide wing beats for lift as well as blows with the clawed hands, leading eventusally to the capability for flight.

Conclusions

The display and threat scenario has a simple theme, and it carries a large number of anatomical, behavioral, ecological, and environmental implications, all derived from an immense data base among living birds. The general correlates of display and threat are also rather widespread among other vertebrates, so that they are reasonable models to apply to the ancestors of birds. Many of the implications of our hypothesis can be tested by examination of Archaeopteryx, and the scenario survives the test. Other implications involve the environment in which flight evolved, and are likewise testable, though they are not analyzed in this paper. The display scenario suggests that birds did not evolve through a long chain of very different selective influences, as Bock (1965, 1969) has proposed. The scenario also explains morphology with other scenarios do not, and so it has power as well as simplicity. And if the biology of Archaeopteryx conforms to the pattern we suggest, the predecessors and descendants of this most famous fossil should be made easier to identify and to interpret biologically.