Narrow Primary Feather Rachises in Confuciusornis and Archaeopteryx Suggest Poor Flight Ability

See allHide authors and affiliations

Science  14 May 2010:
Vol. 328, Issue 5980, pp. 887-889
DOI: 10.1126/science.1188895

Poor Flight of the Ancients

In order to fly, the feathers of birds must be strong enough to support the bird's weight without breaking or bending. The main part of a feather providing structural support is its central shaft, which stiffens the feather along its length. In modern birds, this is hollow to reduce weight. Nudds and Dyke (p. 887) show that the cross-section of the shaft of the Mesozoic birds Archaeopteryx and Confuciusornis was much smaller than that of modern birds. Calculations imply that even if it was solid, it would have been too weak to support powered flight and barely strong enough to allow gliding. Thus, powered flight probably arose later in the evolution of birds and these early birds were poor fliers.


The fossil birds Archaeopteryx and Confuciusornis had feathered wings resembling those of living birds, but their flight capabilities remain uncertain. Analysis of the rachises of their primary feathers shows that the rachises were much thinner and weaker than those of modern birds, and thus the birds were not capable of flight. Only if the primary feather rachises were solid in cross-section (the strongest structural configuration), and not hollow as in living birds, would flight have been possible. Hence, if Archaeopteryx and Confuciusornis were flapping flyers, they must have had a feather structure that was fundamentally different from that of living birds. Alternatively, if they were only gliders, then the flapping wing stroke must have appeared after the divergence of Confuciusornis, likely within the enantiornithine or ornithurine radiations.

The earliest birds, including Archaeopteryx [Late Jurassic; 145 million years ago (Ma)] and Confuciusornis (Early Cretaceous; 120 Ma), have feathered wings that appear capable of generating some aerodynamic force (1). The likely magnitude of these forces and allied flight capabilities, however, remain controversial (210). Some studies suggest that Archaeopteryx was capable of flapping flight (2, 3, 5, 9, 10); others have concluded the opposite (4, 68). The flight capabilities of Confuciusornis (11) have received less attention, despite hundreds of fossils being known (1215), many with complete primary feathers (12) (Fig. 1A). Primary feathers are fundamental to bird flight, contributing 40 to 51% of the total wingspan (16), and they must be capable of sustaining lift forces without breaking (17). Sustained flight, in which a horizontal position is maintained in the air, requires that the flight surfaces of the bird support a lift force equal to body weight [mass × acceleration of gravity (g)]. Much higher forces, far in excess of body weight, are generated in extreme maneuvers and accelerations (18). Previous work indicates that feathers fail by buckling (17). Accordingly, we used Euler-Bernoulli beam theory to determine the load-bearing capabilities of the primary feathers of Archaeopteryx and Confuciusornis and infer their likely flight capabilities (19).

Fig. 1

(A) One of the specimens of Confuciusornis sanctus (BSP 1999 I 15) included in the study. The approximate length of one functional primary feather is marked with a black line. Scale bar is 20 mm. (B) Plots of rachis diameter (bottom) and feather length (top). The regression lines [from (21)] describe the relationships for extant birds (length versus body mass, y = 0.21x 0.30 (0.23 to 0.37) and diameter versus body mass y = 0.004x 0.37 (0.31 to 0.43)). (C) Plot of rachis diameter against feather length of extant birds. The regression line describing the relationship for modern birds is y = 0.03x 1.16 (0.99 to 1.33) (21). Data for Archaeopteryx are represented by dark gray diamonds and for Confuciusornis by light gray diamonds. Extant species are denoted by black circles. The order (left to right) in (B) is L. ridibundus, C. palumbus, G. fulvus, and D. exulans. In (C), the order is C. palumbus, L. ridibundus, D. exulans, and G. fulvus. Dotted lines are 95% confidence limits.

The mean first primary feather length of five Confuciusornis specimens was 207 ± 17 mm, and the average rachis diameter was 1.06 ± 0.12 mm, compared with 129 and 0.75 mm, respectively, for the longest primary of the Munich Archaeopteryx (Table 1), which is consistent with measurements by Elżanowski (20). There was no evidence of flattening during preservation of the fossil feathers, but flattening would have increased apparent diameters and therefore increased our estimates of feather strength. The primary feather lengths of Confuciusornis and Archaeopteryx are close to those predicted for extant birds of similar size (Fig. 1B). In both fossil birds, however, the rachises are narrow for the birds’ body size and feather lengths (Fig. 1, B and C). In contrast to the fossil birds, the feather morphologies of two similar-sized living birds (Larus ridibundus and Columba palumbus) and two larger soaring birds (Diomedea exulans and Gyps fulvus) are comparable to those predicted from the relationships for other extant birds determined by Worcester (21).

Table 1

The morphology, rachis buckling failure moments, and calculated lift forces for the species used in the study.

View this table:

It is a well-established concept that the larger of two cylinders with walls of the same thickness relative to overall radius can withstand a larger moment before buckling (Fig. 2). The significant result here is that the moment required to buckle a Confuciusornis primary feather (0.0036 N·m) is two orders of magnitude less (Fig. 2) than that required (0.3234 N·m) if its diameter matched that of a primary from an extant bird with similar feather length (4.8 mm). Similarly, the primary feathers of Archaeopteryx were one order of magnitude more likely to fail by buckling (critical moment = 0.0013 N·m versus 0.0652 N·m) than are the feathers of a similar-sized extant analog (Fig. 2). Even if the primary feathers are modeled as the strongest possible structure—a solid cross-section of keratin—the critical failure moment (tensile failure) is one order of magnitude below that expected for a similar-sized extant bird for both Confuciusornis (0.0264 N·m) and Archaeopteryx (0.0094 N·m).

Fig. 2

Graph of moment arm (N·m) against rachis diameter (m) [from equation 4 in (17)]. Dark gray diamonds represent Archaeopteryx, light gray diamonds represent Confuciusornis, and the black circles indicate the data for the four extant species (from left to right: L. ridibundus, C. palumbus, D. exulans, and G. fulvus). Black-bordered diamonds are predicted moments if the rachis diameter relative to feather length was the same as that predicted from extant birds. The line (y = 0.006x3) describes the theoretical critical buckling moment as a function of thin-walled cylinder diameter.

The lift on the primary feathers (Lprimaries) as a percentage of lift on the entire wing was calculated as 42.9% for Confuciusornis and 30.9% for Archaeopteryx. For comparison, the lift is 39.2% for C. palumbus, 29.8% for L. ridibundus, 11.1% for D. exulans, and 25.5% for G. fulvus. For lift generation equal to body weight Lprimaries forces must therefore be 1.05 N (Confuciusornis), 0.42 N (Archaeopteryx), 0.94 N (C. palumbus), 0.42 N (L. ridibundus), 4.46 N (D. exulans), and 9.30 N (G. fulvus). For the two fossil birds, these forces exceed what the primaries could withstand, if they were of comparable structure to that of modern bird feathers (Fig. 3). The estimated buckling failure moment of their rachises was only 39% (Confuciusornis) and 55% (Archaeopteryx) of that generated upon the feather by a lift force equal to body weight (Fig. 3). The estimated rachis safety factors (that is, buckling failure force/lift force equal to body weight) of 13.6, 10.1, 9.9, and 6.1 for L. ridibundus, C. palumbus, D. exulans, and G. fulvus, respectively, were comparable to the previously estimated range for C. livia during routine flight [values of 9 to 17 (17)], confirming that these species are capable of flapping flight (Fig. 3) and, hence, validating our methodology. With feathers structurally similar to those of modern birds, Confuciusornis and Archaeopteryx could only have parachuted with their wings held dorsally, reducing the forces acting on the primaries while providing drag to reduce descent speed. A parachuting arrangement seems incongruent with their overall wing morphologies. Even when they are modeled as solid keratin beams, however, the primary feathers of the fossil birds only provide safety factors of 2.9 (Confuciusornis) and 4 (Archaeopteryx), which is well below those of extant birds (Fig. 3).

Fig. 3

Bar chart showing the critical buckling failure moment (solid bars) and bending tensile failure moment (dotted bordered open bars) of the primary feathers as a percentage of the moment that would be exerted upon the feather by a lift force equal to body mass times gravity. Dashed line represents a safety factor of 1.

At each stage of the calculations, the parameter estimates were selected to provide the likely upper bounds of lift-supporting capabilities. For example, the force acting upon the distal part of the wing was divided equally between 10 feathers. Therefore, we likely underestimated the load on the leading primaries, because not all primaries bear distal lift forces equally. Primary feathers 1 to 4 generally constitute the wing tip, whereas primary feathers numbered 5 and above commonly do not extend right to the tip of the wing, and their long axes are progressively oriented more dorsally. In addition, the chordwise lift distribution is generally at least twice as high cranially than caudally. The estimates of body mass are crucial to the calculations of lift force. For their feathers (if morphologically similar to those of modern birds) to be strong enough to sustain a lift force equal to body weight, Confuciusornis and Archaeopteryx would have had to weigh 0.215 and 0.188 kg, respectively, and their body masses are unlikely to be this low. Bird primary feathers have longitudinal furrows, and these are seen in Archaeopteryx (5). Because of the intractable nature of incorporating a furrow into the cross-section, however, the shaft was modeled without a furrow. Previous work (17) showed that the compressive stress required to buckle pigeon primaries, estimated using Euler-Bernoulli beam theory, was 21% higher than actually required. Therefore, the theoretical estimates here are probably high, and the feathers are actually structurally weaker.

Because even rudimentary forelimb movements can generate useful thrust (1, 22, 23), some thrust generation by these fossil birds cannot be discounted, but the vigorous flapping flight of modern birds is highly unlikely. Flight toward the gliding end of the ability spectrum is consistent with Archaeopteryx’s known musculature (4), shoulder anatomy (6), feather vane anatomy, and feather shaft curvature (2). It is also congruent with the enlarged and excavated deltopectoral crest of Confuciusornis (12). Our results suggest that superficial resemblance of Mesozoic bird wing feathers to those of living birds should not necessarily be taken as an indication of the ability to withstand similar aerodynamic forces. Unless Mesozoic primary feathers were structurally different from those of extant birds, further refinements were required after the appearance of primary feathers for the evolution of a powered flapping flight stroke. Consequently, the origins of the modern wing-beat may have occurred among taxa that diverged subsequently to Confuciusornis, likely within the diverse Cretaceous enantiornithine or ornithurine radiations.

Supporting Online Material

Materials and Methods

References and Notes

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank J. Gardiner, G. Kaiser, J. Sigwart, and three anonymous referees. Collections access was kindly provided by O. Rauhut [Bayerische Staatssammlung fuer Palaeontologie und Geologie (BSP)], R. Brocke [Senckenberg Forschungsinstitut und Naturmuseum (SMF)], and J. Oelkers-Schaefer (SMF).

Stay Connected to Science

Navigate This Article