On the Evolution of Feathers from an Aerodynamic and Constructional View Point 1

AMER. ZOOL., 40:676 686 (2000) On the Evolution of Feathers from an Aerodynamic and Constructional View Point 1 SAMUEL F. TARSITANO, 2, *ANTHONY P. RUSSELL, FRANCIS HORNE,* CHRISTOPHER PLUMMER,* AND KAREN MILLERCHIP* *Biology Department, Southwest Texas State University, San Marcos, Texas 78666 Department of Biological Sciences, University of Calgary, Alberta, T2N 1N4, Canada SYNOPSIS. The evolution of birds and feathers are examined in terms of the aerodynamic constraints imposed by the arboreal and cursorial models of flight evolution. The cursorial origin of flight is associated with the putative coelurosaurian ancestry of birds. As presently known, coelurosaurs have a center of mass located in the pelvic region and an elongated pubis that is ventrally or anteriorly directed. Both of these characteristics make it difficult to postulate an origin of flight that would involve a gliding phase because the abdomen cannot be flattened into an aerodynamic shape. Moreover, the cursorial model must counteract gravity using the hindlimb and, thus, selection for the power requirement for lift-off would not focus on the forelimb. Therefore, if the hypothesis proposing a coelurosaurian ancestry of birds is to remain viable, it must be via an as yet undiscovered taxon that is compatible with the morphological and aerodynamic constraints imposed by flight evolution. The arboreal model, currently centers around non-dinosaurian taxa and is more parsimonious in that early archosaurs have short pubes that do not preclude an aerodynamic body profile. Moreover, the arboreal proavis uses gravity to create the airflow over the body surfaces and is, thus, energy efficient. Consideration of the initial aerodynamic roles of feathers and feather design are consistent with a precursory gliding phase. Whether avian ancestry lies among coelurosaur theropods or earlier archosaurs, we must remain mindful of the complex aerodynamic dictates of gliding and powered flight and avoid formalistic approaches that coopt sister taxa, with their known body form, as functional ancestors. INTRODUCTION Two current theories purport to explain the origin of avian flight arboreal and cursorial. The arboreal theory usually involves the consideration of a small early archosaur as an ancestor (Marsh, 1880; Heilmann, 1927; Bock, 1965, 1985), whereas the cursorial theory generally implicates a small coelurosaurian dinosaur as an ancestor (Lowe, 1935; Ostrom, 1974, 1976; Caple et al., 1983; Gauthier, 1986). Extensive reviews of the history of these theories and the origin of birds can be found in Heilmann (1927) and Stephan (1974). In the context of these theories, much has been written about 1 From the Symposium Evolutionary Origin of Feathers presented at the Annual Meeting of the Society for Integrative and Comparative Biology, 6 10 January 1999, at Denver, Colorado. 2 E-mail: st04@swt.edu the flight capabilities of Archaeopteryx, as it already had a well-developed array of modern flight feathers (de Beer, 1954; Feduccia and Tordoff, 1979). The wing morphology and the asymmetrical nature of the feather vanes indicate that Archaeopteryx was already a flier (de Beer, 1954; George and Berger, 1966; Parkes, 1966; Feduccia and Tordoff, 1979; Feduccia, 1985; Norberg, 1985). Thus, Archaeopteryx is far removed from the time in avian evolution when the first attempts at flight were undertaken. However, it is axiomatic that morphological preadaptations existed among the earliest members of what was to become the proavian lineage to accommodate behavioral changes that would lead to the development of flight (Von Wahlert, 1965). Taking this into account, we can examine the two opposing views of the origin of flight to see if they have equal probability from a functional morphological standpoint. 676

EVOLUTION OF FEATHERS 677 The arboreal model hypothesizes that the proavis was a small early archosaur that leaped from branch to branch, or from a branch to a tree trunk, and that this behavior was key in the origin of the first flying archosaur (Heilmann, 1927). The cursorial theory maintains that the proavis was a bipedal coelurosaurian dinosaur which, by leaping to catch insects in its mouth, became adapted to flight (Ostrom, 1976; Gauthier and Padian, 1985; Padian and Chiappe, 1998). Peter Griffiths (personal communication) advocates an arboreal coelurosaur and, thus departs from most supporters of the coelurosaurian ancestry of birds who claim a cursorial habit (Ostrom, 1976; Gauthier and Padian, 1985; Padian and Chiappe, 1998). We herein endeavor to divorce, in the first instance, considerations of taxonomic ancestry in our contemplation of flight origins and deal with aspects of aerodynamics. Only secondarily is it necessary to consider morphology as demonstrated by known fossil taxa as potential ancestors. This is so since regardless of the taxonomic starting point, each morphology must deal with the effects of drag, lift and thrust. Definitions and discussions of these topics that include laminar and turbulent flow and vortex formation are to be found in White (1974); Nachtigall (1977); Ellington (1980, 1984, 1991); Vennard and Street (1982); Tucker (1990), and Vogel (1988). The primary characters of birds are feathers without which birds cannot fly (in air). Therefore it is necessary to examine the conditions under which, and the purposes for which, feathers may have evolved, and how these might relate to selective factors implicating them in flight. Feathers make their first appearance in the fossil record with Archaeopteryx in the Upper Jurassic. Clearly, they must have originated earlier than this, because the feathers of Archaeopteryx already reflect the morphology typical of those of modern birds. The evolution of feathers from reptilian scales has been advocated to be in direct association with the origin of flight (Marsh, 1880; Heilmann, 1927; Parkes, 1966), although the exact mechanisms or stages of feather formation are wanting in such hypotheses. An alternative hypothesis posits that feathers evolved first as sun screens to aid in thermoregulation, and were only later modified for flight (Regal, 1975, 1985). Dyck (1985) proposed that feathers were evolved for water repellence, this being based upon the grooved surface the barbs bestow on a feather s dorsal surface that aids in the ease with which water rolls off the birds body. Moving in fluids: General considerations If we assume that feathers were at least exapted in part for flight, then we can ask what selective forces may have operated to promote the elongation of scales (if feathers are transformed from scales) or protofeathers (if feathers are de novo structures). Protofeathers occur as a result of a developmental change during the formation of scales (Lucas and Stettenheim, 1974; Zeltinger and Sawyer, 1991). This developmental alteration would be manifested in its initial stages by the elongation of the dermal papilla above the skin level and/or the formation of barbs in the structure. In their early stages of evolution it is likely the barbs would be parallel to one another before a rachis is formed. This condition is present in developing contour and flight feathers (Davies, 1889). The arboreal or elevation model Consideration of the physical phenomena of airflow with respect to morphological stages of feather formation may provide insight into the probable selective forces promoting their evolution. Of what benefit would elongated scales or protofeathers be to an arboreal or cliff dwelling proavis? Jumping between trees or branches, falling, would be assisted if morphology provided mechanisms that would either result in slowing the proavis before impact (increasing the drag on the body), or aiding in lift by maintaining a turbulent boundary layer over the body surfaces that prevents separation at high angles of attack. Separation of the boundary layer from the body surface would decrease lift and might be avoided with small-scale turbulence generation. Animals that resort to this sort of behavior (jumping between trees and the ground) and that lack specialized structures to slow the rate of descent flatten themselves and

678 S. F. TARSITANO ET AL. FIG. 1. A diagrammatic representation of stages in the evolution of a wing. The upper row (A C) represents turbulence generating protofeathers used also as steering and braking devices. The bottom row (D F) represents the transition from large scale turbulence generation to lift generation with the elongation and curvature of the protofeathers. At this point feathers would also act as turbulators to create small scale turbulence allowing the flow to remain on the wing and thus reduce pressure drag. Modified after Tarsitano, 1985. spread their limbs in order to use the body as an airfoil for lift generation (Oliver, 1951; Heyer and Pongapipatana, 1970; Russell, 1979). The angle of attack and degree of flattening of the body contour will determine the ratio of lift vs. drag generated while in the air. Templin (1977) has pointed to the necessity of slowing down to about six meters per second before impact in order to avoid injury. While this minimum speed to avoid injury would vary depending upon the organism, it is still an important point to consider. We know from feather development (Davies, 1889) that presumptive feathers are oriented in a posterior direction and are not elevated more than forty degrees above the horizontal. Thus, in the earliest stages of the evolution of flight in the trees down scenario, when the proavis has not achieved modern feather morphology, what would be the selective advantage to the elongation of scales or protofeathers? Would the protofeathers function in order to effect transition to a turbulent flow (Fig. 1a, b) or would they act as turbulators (or vortex generators) to reattach a separated flow due to high angles of attack the body experienced? We offer the hypothesis that the protofeathers are not themselves producing lift, but rather are producing turbulence. Whether this turbulence is used to enhance lift on the body by allowing the boundary layer to remain attached to the body surfaces, including the limbs, or whether the protofeathers were creating enough turbulence that would lead to separation of the boundary layer from the body surfaces is unknown. Theoretically projections from a surface can lead to turbulence (White, 1974; Vennard and Smith, 1982; Cummings and Bragg, 1996). A criteron for determining whether protofeathers or any other projection will cause transition is that the non-dimensional roughness height, ku/n, be larger than 150, where k is the roughness height, U is the free stream velocity, and n is the kinematic viscosity. For protofeathers that are 4 mm in length and exposed to a flow of 6 m/second, a value of 1,600 is reached, well above the limit to induce turbulence. In experiments we conducted using protofeather models placed in a water tunnel at 10 cm/sec, we found that 4 mm projections inclined in a posterior direction at an angle of 39 degrees above the skin caused turbulence (Fig. 2). This is significant because the turbulence we observed did not cause separation from the model surface, indicating the possibility that protofeathers could have had an aerodynamic function associated with lift. Disturbances in the boundary layer can either lead to an increase in lift or a loss of lift depending upon whether the turbulence enables the boundary layer to adhere to the surface of the body (White, 1974; Vennard and Street, 1982). If such turbulence is large enough, the boundary layer may separate from the body surface and, thus, increase the pressure drag on the body or induce a stall depending upon the position of the separation along the body or wing (Mal-

EVOLUTION OF FEATHERS 679 FIG. 2. Flow visualization of flat plate and protofeather models in water tunnel operating at 10 cm/sec. A. laminar flow over a smooth plate. B D. Turbulence generated by protofeathers 4 mm in length and inclined at 39 40 degree angle. This implies that protofeathers could indeed affect the flow characteristics over the body of a protoavis. kiel and Mayle, 1996; Huang et al., 1996). It is possible that the turbulence generated by projections (in our case protofeathers) could theoretically prevent boundary layer separation at different angles of attack (Abdel-Rahman and Chakroun, 1997). Projections into the boundary layer have been shown to behave in this manner (Carlson and Lumley, 1996; Debisschop and Nieuwstadt, 1996; Abdel-Rahman and Chakroun, 1997), although at different Reynolds numbers than might be experienced by evolving protobirds (Sundaram et al., 1996; Kerho and Bragg, 1997). Along these lines, we suggest that the early stages of flight had to do with lift generation whereby differential airflow over a cambered body surface generated a vertical force on the animal s body (Peters and Gutmann, 1984). This lifting force centered initially more on the body than the forelimbs and was used to extend a glide between trees or branches. The first step in the evolution of bird flight was a change in behavior that resulted in the proavis leaping between branches (Von Wahlert, 1965; Russell, 1979). The protofeathers were selected for maintaining the boundary layer on the body surfaces at varying angles of attack. Any hypothesis concerning feather evolution must account for the elongation of not only the flight and contour feathers, but also the feathers (termed coverts) which overlap the remiges anteriorly along the wing. This feather elongation and distribution may be explained in the following way. Because the boundary layer will increase in thickness across the chord of the protowing, it is more likely (Vogel, 1983), that the more posteriorly placed protofeathers would elongate more than those placed at the leading edge of the limb (Fig. 1e). We suggest this hypothesis in light of the necessity to increase the roughness past the subviscous layer of the boundary layer (White, 1974; Vennard and Street, 1982; Kerho and Bragg, 1997) in order to initiate transition from a laminar to turbulent flow. Are elongated scales found in the integument of archosaurs? Unfortunately, the integumentary form in probable arboreal early archosaurs, such as Megalancosaurus, is unknown. However, many crocodilians, such as alligators and gharials, have elon-

680 S. F. TARSITANO ET AL. gated scales on the trailing edges of their hindlimbs and a tuberculate scale pattern on the forelimb. The presence of scale elongation in crocodilians reinforces the possibility that archosaurs leading to birds might have possessed similar modifications. Thus, there might have been a developmental constraint as to how, and in which direction, scale elongation could have proceeded. As elongation of protofeathers proceeded, an airfoil with tapered trailing edge would be developed and thus a higher lift to drag ratio could evolve. The protofeather elongation on the limb would lead to the upper surface of the protowing being cambered. With changes in the angle of attack of the wing, changes in lift could be accomplished with no additional or specialized mechanisms, utilizing only the pre-existing deltoid and pectoral musculature. Once the forelimb was a wing (i.e., an organ capable of producing more lift than drag), selection would favor enhancement of the pectoral adductor musculature to resist the lifting forces on the wings. In summary, our observations of smooth limbs in water tunnel tests showed mainly laminar flow over the limb s surface. Roughened surfaces with protofeathers created turbulence whereas the smooth surface did not (Fig. 2). Both smooth and rough surface models were tested at a Reynolds number of 90,000. At this point in time, we suggest that protofeathers could have added directly to lift on the body by preventing the separation of flow and thereby actually narrowing the wake behind the animal. The limbs might have been used for steering using turbulence generation. Testing of small chord, model limbs with protofeathers is necessary to understand the flow patterns and lift/drag ratios. To state at this time that protofeathers would add to lift generation operating in the manner of modern feathers is premature given the difference between mathematical predictions of drag coefficients and their actual measurements on models (Vennard and Street, 1982). Garner et al. (1999) who postulated that flight evolved as a mechanism to support pouncing onto prey from an elevated position support the viewpoint that limb protofeathers were used to generate turbulence. Their scenario suggested that early stages of flight were based upon drag developed by the protowings that later were exapted for lift generation as postulated by Tarsitano (1985). This change might have been exaptive because the protofeathers would impart eventually a curvature to the protowing. The curvature would allow differential airflow over its surfaces producing lift and a tapered posterior edge, to reducing pressure drag of the limb by streamlining (Tarsitano, 1985). The cursorial model Turning to the cursorial model of the origin of avian flight, we are faced with quite a different set of circumstances. The cursorial proavis, taken to be a coelurosaur (Ostrom, 1976; Gauthier and Padian, 1985; Padian and Chiappe, 1998) must run or jump in order to create an air current for flight. How would elongating the scales (or forming protofeathers) in the earliest stages of a cursorial model benefit such an ancestor? For an animal running to gain speed, exposure of as much surface area as possible would be counterproductive and unlikely to favor elongated scales or protofeathers. In the early stages of such a scenario, most lift would be generated by the body surface. The large hindlimbs oscillating during locomotion would cause largescale turbulence and separation of flow from the body surfaces, leading to increased drag. The faster the animal ran, the greater the production of drag because parasite drag increases with the square of forward velocity (Pennycuick, 1972; Nachtigall, 1977; Vogel, 1983; Rayner, 1985). Therefore, by running, a cursorial proavis would expend energy only to induce detrimental drag (Tarsitano, 1985). Our water tunnel tests on dinosaur models indicated that this scenario is close to the truth. These models, patterned after Compsognathus, created large-scale turbulence where the water flow separated from the body at the pelvic region. This did not happen with the quadrupedal early archosaur model that had laterally oriented limbs and a flat abdomen. In contrast to an arboreal model, a cursorial proavis must work against gravity (Tucker, 1938). Leaping into the air from

EVOLUTION OF FEATHERS 681 the ground would reduce forward velocity and maneuverability (Rayner, 1985), making it more difficult to effect a lift-off in the nascent stages of the development of the protowing which could not generate more lift than drag. It seems clear that during running, a bipedal cursorial coelurosaur would benefit most by presenting the least amount of surface area to the air, in order to minimize profile drag (pressure and skin friction drag) and to prevent airflow separation from the body. Thus, the elongation of scales in such forms as presently known seems unlikely (Tarsitano, 1985). Other important factors in the evolution of flight are size, volume distribution, and surface area. Both the cursorial and arboreal models of flight require a small archosaur, but the proposed coelurosaurian proavis is relatively large: the smallest Compsognathus corallestris (nearly complete specimen) adult is over one meter in length (Taquet, 1985). This size is far too large to fit either the model of Caple et al. (1983) or its revised version (Balda et al., 1985) that requires at least 50% of the body weight to be compensated for by generated lift. Neither the size factor, nor the body proportions (a width of 3 cm and a length of 15 cm), as advocated in the Caple et al. (1983) model, approximate any known adult theropod including the newly discovered Sinosauropteryx. For example, in the model proposed by Caple et al. (1983), the hindlimbs are extremely small and the tail is short. However, a prominent feature of all coelurosaurs is large hindlimbs and long tail. The tail of the model coelurosaur is too short to be used as a counterbalance organ (see Charig, 1972). A model based on a hypothetical morphology is of limited use in terms of ancestor descendant or sister group scenarios, but is instructive from a functional standpoint. The large hindlimbs of coelurosaurs (where the forelimbs are less than 70% of the hindlimb length) demonstrate their cursorial nature (Charig, 1972) as does their posteriorly placed center of mass. This contrasts with the quadrupedal stance and gait of early archosaurs. The cursorial theory requires a well-developed hindlimb and pelvic girdle because the hindlimb must provide the induced velocity for lift-off. As Rayner (1985) pointed out, as soon as the proavis leaped from the ground it would lose velocity, and gravity would act against its forward and upward motion. Thus, in order to gain height to make gliding effective, the proavis would have to leap, but to what end? Leaping to catch a flying insect is energetically costly, especially at the level of flight capability of the proavis, where the insect, with powered flight capability, would have an advantage in maneuverability. The cursorial theory proposed by Caple et al. (1983) is not energetically efficient, because gravity must be overcome by the power output of the hindlimb. The only alternative is for the cursorial proavis to run downhill, taking advantage of the negative energy requirements of this type of behavior (Goldspink, 1977). In fact, we offer this suggestion as an addendum to the cursorial theory because, energetically, it provides a less expensive strategy to produce the induced velocity. In the arboreal proavis there is minimal energy expenditure to initiate flight, because the organism works with gravity and drag, whereas the cursorial proavis must work against gravity and drag (Norberg, 1985; Tarsitano, 1985). The arboreal starting point, requires only that a small archosaur, perhaps the size of a gliding gecko (some 10 15 cm long) takes the first leap to enter a new adaptive zone (Tarsitano, 1985). In order to create any airflow about the body surfaces, the arboreal proavis need only fall (see parallel arguments about the origin of insect flight, Kukulova- Peck, 1983). Thus gravity would have worked with the arboreal proavis, and little energy would have had to be expended in order to enter a new adaptive zone (Norberg, 1985; Tarsitano, 1985). In contrast a running coelurosaur would have a good deal of baggage to carry, in the form of the mass of the pelvis and hindlimbs (the hindlimbs once airborne are no longer able to produce power but are still a major source of drag) that would have to be to overcome by the relatively small forelimbs. In addition, there has never been any evidence for feathered forelimbs in any dinosaur and thus the hypothesis that lift pro-

682 S. F. TARSITANO ET AL. duction of any significance from such limbs is dubious. The forelimb/hindlimb ratios of coelurosaurs are unlike those of Archaeopteryx. Ostrom (1976), in fact, reported the forelimbs of Deinonychus were about 66% of the hindlimb length. In Archaeopteryx, the forelimb length exceeds that of the hind-limb and thus the forelimbs are over 100% of the hindlimbs. Those coelurosaurs with the largest forelimbs are so large and heavy as to preclude the evolution of flight. The smaller theropods, such as Compsognathus longipes, have forelimb/hindlimb ratios of under 40% (Ostrom, 1978). These limb ratios demonstrate how Archaeopteryx differs from its theropod cousins. The thecodont, Megalancosaurus (Calzavara et al., 1981), has elongated forelimbs; thus a comparison of relative limb lengths in this form with those of Archaeopteryx might well be instructive. Even if provided with protofeathers, the relatively short forelimbs of known coelurosaurs would not have been able to provide the surface area during the early stages of wing evolution necessary to lift the mass of the animal. This is the stumbling block of any cursorial theory. It is most probable that a proavis did not start to fly with the wing span and aspect ratio typical of Archaeopteryx, but with a moderately-developed forelimb possessed of slightly elongated protofeathers. In addition, the vertically oriented pubes in Deinonychus and other Asian coelurosaurs would further counteract lift by forming a V shaped ventral surface rather than a flattened or concave surface for gliding. Lizards that leap to launch a glide flatten the abdomen (Russell, 1979), to make effective use of the body as an airfoil. In contrast to early archosaurs with short pubes (Ewer, 1965), morphologically-known coelurosaurs would have been incapable of flattening the abdomen and thus would gain little lift from their body form. Furthermore, a quadruped or semibiped has a body shape that would be much more concordant with the body shape and center of mass dictated by the requirements of the development of flight. This is so because the center of mass would be more anterior in such forms than the center of mass in an archosaurian biped. Therefore, any lift required from a feathered tail would be less than in a biped. For these reasons, the only tetrapods that have demonstrably given rise to flight are quadrupeds. The major requirements for early gliding without help from elongated scales or protofeathers is small size, a flattened ventrum, and body proportions that do not have any massive areas to cause unbalanced rotations and pressure drag. These conditions are not met by any known coelurosaurs. The primitive, antero-ventrally directed pubes in coelurosaurs might have precluded their flattening the belly for gliding. Thus, the major source of lift the body is inconsistent with the demands of flight and alone may have precluded the evolution of flight in any known coelurosaurs, if our arguments hold that a gliding stage is a necessary intermediate. For these reasons coelurosaurs possessing a different morphology than that presently known should be sought. Regardless of whether they begin flight from a cursorial or arboreal mode their morphology should be more compatible with the aerodynamic constraints of body shape and weight distribution. A wind-tunnel test of a Deinonychus model would help to determine drag coefficients of the body shape and at what point elongated scales or protofeathers would provide enough lift to support a glide. Those who have suggested that there was no gliding stage in the evolution of avian flight (e.g., Gauthier and Padian, 1985) have neglected the morphological and physiological constraints on active flight (see George and Berger, 1966; Tucker, 1973; Greenwalt, 1975 and Clark, 1977 for excellent discussions). Furthermore, a terrestrial takeoff requires the most complex wing stroke (Brown, 1951), and thus is unlikely to evolve before gliding. In this way, the balance system proposed by Peters and Gutmann (1985) can be developed as the evolution of flight proceeded and there is no need for any morphological prerequisites for flight save the ability to flatten the abdomen (Russell, 1979). To summarize, the reasons for arguing against a cursorially launched flight using known coelurosaur body morphology are:

EVOLUTION OF FEATHERS 683 1) extremely high energetic requirements of a ground-based take-off; 2) the placement of the center of mass in the pelvic region instead of in the thorax; 3) a body shape that precludes the production of a body gliding surface; and 4) hindlimb and pelvic girdle morphology that creates large-scale turbulence. Paleontologists (Padian and Chiappe, 1998) who favor the cursorial theory have not effectively dealt with the aforementioned criticisms. A groundlaunched flight is only possible after the evolution of a feathered wing and wing stroke had been evolved (Burgers and Chiappe, 1999). The evolution of feathers A problem common to both the arboreal and cursorial theories of the origin of avian flight lies in the evolution of the structure of the feather itself. In the wings of modern birds, or even Archaeopteryx, overlapping remiges act to brace the feathers against the lifting force of the wing. Strengthening of the feathers vane is facilitated by the formation of hooklets on the ends of the barbules, which attach to the barbules of an adjacent barb (Davies, 1889; Lucas and Stettenheim, 1972; Dyck, 1985). Pennycuick (1972) presented a scenario to account for this morphology. According to his explanation, the lifting forces acting on the wing would tend to curl the edges of protofeathers upward, increasing the induced drag on the wing (due to vortex formation; see Rayner, 1985 and Norberg, 1996 for an explanation). If the edges of the protofeather developed corrugations, these would help to eliminate the bending moments at the boundaries of the protofeather. Inward extensions of these corrugations toward the midline of the protofeather would result in the production of barbs. Our examination of feather development has cast doubt upon this scenario because barbs do not form in this way, but rather emanate as parallel rods from the circular collar (Davies, 1889; Lucas and Stettenheim, 1972). The growth of the rachis gives the barbs their angle on the vane (Lillie and Juhn, 1938). Our studies indicate that the barbs are not uniform in morphology or thickness along their lengths and that this morphological detail may help contribute to the angle they make with the rachis. Corrugations can only occur at the vane edges to form barbs in the flat plate scenario used by Regal (1975, 1985), only if the keratin strands making up these protobarbs are already directed outward toward the vane boundaries. The development of barbs from the circular collar precludes the corrugation hypothesis. The method of feather formation and vane morphology can be used to corroborate the hypothesis that feathers were selected for a flight function. Regal (1975, 1985) illustrated the protofeather with the protobarbs oriented parallel to the rachis and extending outward as horizontal plates. Feather development and vane construction shows that the barbs are oriented at 90 degrees to the long axis of the rachis (Lucas and Stettenheim, 1972) and not horizontally placed, as in the model posited by Regal (1985). Horizontally oriented barbs will fulfill a sun screen function, but such an arrangement would be wholly incapable of preventing the buckling of either the protofeather or modern feathers that are subjected to aerodynamic forces. Barbs are actually oriented with their widths facing in a dorso ventral plane to resist these bending forces. Therefore, if Regal (1985) were correct, the developmental plan for barb formation would have to have completely reoriented itself, so that the barbs could function in an aerodynamic mode. If feathers were evolved for insulation and then exapted for flight it is difficult to understand why such a complex structure was developed for the initial purpose. Hair-like structures would have been simpler to build. Thus the developmental complexity of feathers contradicts Regal s (1985) interpretation on both biomechanical and embryological grounds. CONCLUSIONS While the systematic relationships of birds have been extensively studied, concomitant arguments about the origin of avian flight have generally been constrained by forcing transitions from a putative relative of known body form, rather than arguing from the standpoint of aerodynamic principles. In addition, few proponents of var-

684 S. F. TARSITANO ET AL. ious theories have tested their flight models experimentally. At the present time, we know more about theropod evolution than of small arboreal early archosaurs. Since coelurosaurs are well represented in the fossil record, we should be able to find theropods whose morphology fits the aerodynamic constraints of evolving flight in birds. Instead, we discover that theropods differ from the aerodynamic form that would fit the model that Caple et al. (1983) postulated. Coelurosaurs are too large and are anatomically specialized for cursorial locomotion. Accordingly, paleontologists, saddled with a cursorial archosaur, have had to disregard the drawbacks of a highly specialized pelvis and hindlimb and propose theories of flight that contain unsound energetic and ecological hypotheses in order to get a theropod derivative off the ground (Ostrom, 1976; Padian, 1982; Caple et al. 1983). Coelurosaurs, as presently known, are morphologically unsuited for flight of any kind. In order to better fit the aerodynamic model a coelurosaur should be found that have the following features: 1) forelimbs as long or longer than the hindlimbs; 2) small enough size that wing loading would be within the realm of gliding; 3) either short pubes or an opisthopubic condition that oriented the pubes nearly parallel with a horizontal ischium to make body gliding possible; 4) a center of mass in front of the pelvis; and 5) a well-developed and opposable hallux (an arboreal character not found in known coelurosaurs) and coracoids that were oriented in two planes, indicating an elongation over the known coelurosaur condition (Tarsitano and Hecht, 1980). Aerodynamic evaluation is consistent in its indication of a lightly built, small archosaur that could use a favorable surface to volume ratio and gravity to cover relatively long distances in its environment at less energetic cost (Tucker, 1973; Rayner, 1985; Tarsitano, 1985). Testing of models in wind tunnels is necessary to support the morphological starting points of avian flight and the aerodynamic function of protofeathers. Wind tunnel testing should concentrate on the lift/drag ratios of both coelurosaurian and early archosaur body shapes. Such tests should be applied to protofeather models to determine how the protofeathers affect the charactistics of the boundary layer at various sizes and distributions on the limb and body surfaces. Moreover testing of protofeathers on limb and body models at different angles of attack are needed to determine how such protofeathers would benefit an archosaur in a gliding mode. ACKNOWLEDGMENTS We would like to thank Paul Maderson and Dominique Homberger for inviting us to participate in this symposium. In addition, we thank them for reviewing the manuscript over the course of its development. Thanks to Linda Treub, Larry Martin, Ulla Norberg, Philip Burgers and Carl Haux for their review and suggestions for this manuscript. Special thanks are also due to David Bogard of the University of Texas at Austin for allowing us use of the Turbulence and Turbine Cooling Research Laboratory and for his contributions to the aerodynamic explanations in this paper, and to Larry Martin, Max Hecht and Alan Feduccia for many discussions on the evolution of birds. We thank the National Science Foundation for travel support to attend this symposium. REFERENCES Abdel-Rahman, A. A. and W. M. Chakroun. 1997. Surface roughness effects on flow over airfoils. Wind Engineering 21(3):125 137. Balda, R. P., G. Caple and W. R. Willis. 1985. Comparison of the gliding to flapping sequence with the flapping to gliding sequence. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 267 277. Freunde des Jura-Museums, Eichstätt. Bock, W. J. 1965. The role of adaptive mechanisms in the origin of higher levels of organization. Syst. Zool. 14:272 287. Bock, W. J. 1985. The arboreal origin of flight. In M. K. Hecht, J. H. Ostrom, G. Viohl and P. Wellnhofer (eds.), The beginnings of birds, pp. 199 207. Freunde des Jura Museums, Eichstätt. Bragg, M. B. and G. M. Gregorek. 1987. Experimental study of airfoil performance with vortex generators. J. Aircraft. 24:13 22. Brown, R. H. L. 1951. Flapping flight. Ibis 93:333 359. Burgers, P. and L. M. Chiappe. 1999. The wing of Archaeopteryx as a primary thrust generator. Nature 399:60 62. Calzavara, M., G. Muscio, and R. Wild. 1981. Mega-

EVOLUTION OF FEATHERS 685 lancosaurus preonensis, n. g., n. sp., a new reptile from the Norian of Friuli, Italy. Atti Mus. Fruil. Storia. Nat. 2:49 64. Caple, G., R. P. Balda, and W. R. Willis. 1983. The physics of leaping animals and the evolution of preflight. Am. Nat. 121:455 476. Carlson, H. A. and J. L. Lumley. 1996. Flow over an obstacle emerging from the wall of a channel. AIAA Journal 34:924 931. Charig, A. J. 1972. The evolution of the archosaur pelvis and hindlimb: An explanation in functional terms. In K. A. Joysey and T. S. Kemp (eds.), Studies in vertebrate evolution, pp. 121 155. Oliver and Boyd, Edinburgh. Clark, B. D. 1977. Energetics of hovering flight and the origin of bats. In M. K. Hecht, P. C. Goody, and B. M. Hecht (eds.), Major patterns in vertebrate evolution, pp. 423 425. Plenum Press, New York and London. Cummings, M. J. and M. B. Bragg. 1996. Boundarylayer transition due to isolated three-dimensional roughness on the airfoil leading edge. AIAA Journal 347:1949 1953. Davies, H. R. 1889. Die Entwicklung der Feder und ihre Beziehungen zu anderen Integumentgebilden. Gegenbaurs Morph. Jahrb. 15:560 649. de Beer, G. R. 1954. Archaeopteryx lithographica: A study based upon the British Museum specimen. British Museum (Natural History), London. Debisschop, J. R. and F. T. M. Nieuwstadt. 1996. Turbulent boundary layer in an adverse pressure gradient: Effectiveness of riblets. AIAA Journal 34: 932 937. Dyck, J. 1985. The evolution of feathers. Zool. Scr. 14:137 154. Ellington, C. P. 1980. Vortices and hovering flight. In W. Nachtigall (ed.), Instationäre Effekete an schwingenden Tierflügel, pp. 64 101. Franz Steiner, Wiesbaden. Ellington, C. P. 1984. The aerodynamics of hovering insect flight. Phil. Trans. Roy. Soc. Lond. B. 305: 1 181. Ellington, C. P. 1991. Aerodynamics and the origin of insect flight. Adv. Insect Physiol. 23:171 210. Ewer, R. F. 1965. The anatomy of the thecodont reptile, Euparkeria capensis, Broom. Phil. Trans. Roy. Soc. Lond. (B) 248:379 435. Feduccia, A. and H. B. Tordoff. 1979. Feathers of Archaeopteryx: Asymmetric vanes indicate aerodynamic function. Science 203:1021. Feduccia, A. 1985. On why the dinosaur did not evolve feathers. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 75 79. Freunde des Jura-Museums, Eichstätt. Garner, J. P., G. K. Taylor, and A. L. R. Thomas. 1999. On the origins of birds: The sequence of character acquistition in the evolution of avian flight. Proc. R. Soc. Lond. 266:1259 1266. Gauthier, J. 1986. Saurischian monophyly and the origin of birds. In K. Padian (ed.), The origin of birds and the evolution of flight, pp. 1 55. Memoirs of the California Academy of Sciences, San Francisco. Gauthier, J. and K. Padian 1985. Phylogenetic, functional and aerodynamic analyses of the origin of birds and their flight. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 185 197. Freunde des Jura- Museums, Eichstätt. George, J. C. and A. J. Berger 1966. Avian myology. Academic Press, New York and London. Goldspink, G. 1977. Mechanics and energetics of muscle in animals of different sizes, with particular reference to the muscle fibre composition of vertebrate muscle. In T. J. Pedley (ed.), Scale effects in animal locomotion, pp. 37 55. Academic Press, New York. Greenwalt, C. H. 1975. The flight of birds. Trans. Amer. Phil. Soc. New Series. 65: Part 4, 1 67. Heilmann, G. 1927. The origin of birds. Witherby, London. Heyer, W. R. and S. Pongapipatana. 1970. Gliding speeds of Ptychozoan lionatum (Reptilia: Gekkonidae) and Chrysopelea ornata (Reptilia: Colubridae). Herpetologica 26:317 319. Huang, R. F., W. W. Shy, and S. W. Lin. 1996. Influence of surface flow on Aerodynamic loads of a cantilever wing. AIAA J. 34:527 532. Kerho, M. F. and M. B. Bragg. 1997. Airfoil boundary layer development and transition with large leading-edge roughness. AIAA J. 35:75 85. Kukulova-Peck, J. 1983. The origin of the insect wing and wing articulation from the arthropodan leg. Can. J. Zool. 6:1618 1699. Lillie, F. R. and M. Juhn. 1938. The physiology of the feather. II General principles of development with special reference to the after feather. Physiol. Zool. 11:434 448. Lowe, P. R. 1935. On the relationships of the Struthiones to the dinosaurs. Ibis 13:398 432. Lucas, A. M. and P. R. Stettenheim 1972. Avian anatomy. Integument. Agricultural Handbook 362. U.S. Government Printing Office, Washington, D.C. Marsh, O. 1880. Odontornithes: A monograph on the extinct toothed birds of North America. Report of the geological exploration of the fortieth parallel, Vol. 7. Professional papers of the Engineer Department, U. S. Army, no. 18. Malkiel, E. and R. E. Mayle. 1996. Transition in a separation bubble. J. Turbomachinery 118:752 759. Nachtigall, W. 1977. Die Bedeutung der Reynolds Zahl in der Schwimmphysiologie und Flugbiophysik. Fortschr. Zool. 168:40 118. Norberg, R. Å. 1985. Function of vane asymmetry and shaft curvature in bird flight feathers: Inference on flight ability of Archaeopteryx. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 303 313. Freunde des Jura-Museums, Eichstätt. Norberg, U. M. 1985. Evolution of flight in birds: Aerodynamic, mechanical and ecological aspects. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 293 302. Freunde des Jura-Museums, Eichstätt. Norberg, U. M. 1996. Energetics of flight. In C. Carey

686 S. F. TARSITANO ET AL. (ed.), Avian energetics and nutritional ecology, pp. 199 249. Chapman and Hall, New York. Oliver, J. A. 1951. Gliding in amphibians and reptiles, with a remark on an arboreal adaptation in the lizard Anolis carolinensis carolinensis Voigt. Amer. Nat. 85:171 176. Ostrom, J. H. 1974. Archaeopteryx and the origin of flight. Quart. Rev. Biol. 49:27 47. Ostrom, J. H. 1976. Archaeopteryx and the origin of birds. Biol. J. Linn. Soc. 8:91 182. Ostrom, J. H. 1978. The osteology of Compsognathus longipes Wagner. Zitteliana 4:73 118. Padian, K. 1982. Macroevolution and the origin of major adaptations: Vertebrate flight as a paradigm for the analysis of patterns. Proc. Third North Amer. Paleo. Conv. 2:387 392. Padian, K. and L. S. Chiappe. 1998. The origin of birds and their flight. Sci. Amer. 278:38 47. Parkes, K. C. 1966. Some speculations of the origin of feathers. The Living Bird 5:77 86. Pennycuick, C. J. 1972. Animal flight, Studies in Biology no. 33. The Camelot Press Ltd., Southampton. Peters, D. S. and W. F. Gutmann. 1985. Constructional and functional preconditions for the transition to powered flight in vertebrates. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 233 242. Freunde des Jura-Museums, Eichstätt. Rayner, J. M. V. 1985. Cursorial gliding in protobirds: An expanded version of a discussion contribution. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 289 302. Freunde des Jura-Museums, Eichstätt. Regal, P. 1975. The evolutionary origin of feathers. Quart. Rev. Bio. 50:35 66. Regal, P. 1985. Common Sense and reconstructions of the biology of fossils: Archaeopteryx and feathers. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. Freunde des Jura-Museums, Eichstätt. Russell, A. P. 1979. The origin of parachuting locomotion in gekkonid lizards (Reptilia: Gekkonidae). Zool. J. Linn. Soc. 65:233 249. Stephan, B. 1974. Urvögel. Neue Brahm-Bücherei no. 465. A. Ziemsen Verlag, Wittenberg-Lutherstadt. Sundaram, S., P. R. Viswanath, and S. Rudrakumar. 1996. Viscous drag reduction using riblets on NACA 0012 airfoil to moderate incidence. AIAA J. 34:676 682. Taquet, P. 1985. Two new Jurassic specimens of coelurosaurs (Dinosauria). In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 229 232. Freunde des Jura- Museums, Eichstätt. Tarsitano, S. 1985. The morphological and aerodynamic constraints on the origin of birds and avian flight. In M. K. Hecht, J. H. Ostrom, G. Viohl, and P. Wellnhofer (eds.), The beginnings of birds, pp. 319 332. Freunde des Jura-Museums, Eichstätt. Tarsitano, S. and M. K. Hecht. 1980. A re-consideration of the reptilian relationships of Archaeopteryx. Zool. J. Linn. Soc. 69:149 182. Templin, R. J. 1977. Size limits in flying animals. In M. K. Hecht, P. Goody, and B. M. Hecht (eds.), Major patterns in vertebrate evolution, pp. 411 421. Plenum Press, New York. Tucker, B. W. 1938. Functional evolutionary morphology: The origin of birds. In Sir G. de Beer (ed.), Evolution, pp. 321 336. Clarendon Press, Oxford. Tucker, V. A. 1973. Bird metabolism during flight: Evaluation of a theory. J. Exp. Biol. 58:689 709. Vennard J. K. and R. L. Street 1982. Elementary fluid mechanics. John Wiley and Sons, New York. Vogel, S. 1983. Life in moving fluids. Princeton University Press, Princeton. Vogel, S. 1988. Life s devices. Princeton University Press, Princeton. von Wahlert, G. 1965. The role of ecological factors in the origin of higher levels of organization. Syst. Zool. 14:288 300. White, F. M. 1974. Viscous fluid flow, McGraw-Hill, New York. Zeltinger, J. and R. H. Sawyer, 1991. Avian scale development. Dev. Biol. 149:55 65.