1 A general feature of powered locomotion based on an oscillating wing is an asymmetry in how the wing meets the environment during the downstroke compared with the upstroke parts of the wingbeat cycle7. The downstroke in birds, when primary lift and propulsion are achieved, is characterized by an outstretched wing. The more complicated upstroke involves rapid withdrawal of the wing towards the body to reduce its surface area, elevation and subsequent extension in a way that minimally retards lift and thrust gained in the previous downstroke. The distinct musculoskeletal configuration of the m. supracoracoideus (SC) was not present in the Jurassic bird Archaeopteryx8,9, nor is there firm evidence for its presence in recently described Late Jurassic or Early Cretaceous species10,11,12,13,14,15. The evolution of this condition is thought to have been important for flapping powered flight in improving the function of wing elevation16,17, but the significance of its highly derived organization has not been appreciated.

We measured isometric forces of rotation (torque) about the longitudinal axis of the humerus, the extent of unrestrained humeral elevation and rotation, and the muscle's contractile properties, by direct nerve stimulation of the SC in situ. Stimulation of the SC at joint angles of elevation/depression and protraction/retraction coincident with the downstroke–upstroke transition and mid-upstroke2 imparted a substantial mean isometric torque about the longitudinal axis of the humerus (Fig. 2). A mean isometric value of 4.9N (n = 2) was recorded at the downstroke–upstroke transition. In the humeral excursion experiments, we measured in situ humeral rotations of up to 70° and maximum elevations of 50°. In three length–force experiments, we measured a mean (±s.d.) maximal tetanic force of 6.5 (±1.2)N, approximately 10 times body weight. The length of the SC coincident with the downstroke–upstroke transition corresponds to a position on the ascending limb of the active length–force curve of 3.0N (Fig. 3). In lateral view, the right humerus rotates anticlockwise about its longitudinal axis and elevates a total of 50° during upstroke (−10° to 40°); these movements correspond to muscle shortening of 3.0mm and a coincident decrease in the SC's potential for active force production (Fig. 3). The ascending arm of the active length–force curve occurs over a short distance (5mm) and is steep owing to the short fascicle lengths associated with the SC's bipinnate architecture. These data reveal that, although the SC is capable of elevation of the humerus, a far more important role is to provide a high-velocity rotation about its longitudinal axis.

Figure 2: Rotational force of the m.
figure 2

supracoracoideus (SC) measured at the deltopectoral crest. We measured in situ isometric rotational force (torque) produced by the SC in two experiments at two positions of the humerus relevant to the wingbeat cycle; downstroke–upstroke transition (a) and mid-upstroke (b). Force measured at the downstroke–upstroke transition (a) was consistently higher than that at mid-upstroke (b). The moment arm of the SC for rotation is short (2mm). The resulting mechanical advantage for rotation of the humerus by the SC is relatively low, but when input forces are high, the arrangement is favourable for the production of high-velocity rotation at the distal wing. Time bar, 100ms; force bar, 2.5N.

Figure 3: Functional correlates of length–force in European starlings (Sturnus vulgaris).
figure 3

Top, active and passive length–force characteristics of the supracoracoideus muscle derived by direct nerve stimulation from four in situ experiments. Bottom, anterior views of the right shoulder at the downstroke–upstroke transition (A), mid-upstroke (B) and late upstroke (C). The humerus is depressed 10° below the horizontal at the downstroke–upstroke transition, which corresponds to a position on the ascending arm of the active length–force curve (A) where the muscle is capable of producing 85% of total force. As the wing is elevated/rotated to the mid-upstroke (B) and late-upstroke (C) positions, the capacity for force production by the SC declines. Scale bar, 1.0cm.

Further support for this conclusion can be drawn from the mechanical organization of the SC, its electrical activity patterns reported during slow and fast flight, and the highly derived morphology of the avian shoulder. The bipinnate structure of the SC, with its relatively short but numerous fascicles characteristic of all birds we examined, is an architecture for production of high force. The SC's moment arm for humeral rotation about the longitudinal axis is short; we estimate its maximum in the starling to be 2.0mm. When a large torque (as provided by the SC) is applied across this short moment, the resultant velocity of humeral rotation and its attached distal wing element is magnified.

Aerodynamic models predict that active muscle contractile force is far more important for wing elevation in birds during slow than in fast flapping flight18. Passive aerodynamic lift is sufficient at fast flight speeds, particularly in birds with high-aspect ratio wings. In four species of birds with diverse flight styles for which electromyograms of the SC during flight have been reported, however, the muscle remains electrically active over a range of flight speeds2,4,19,20. Passive aerodynamic lift may not be implemented at the flight speeds attained in these experimental situations. The SC may act to adjust the planform of the wing's leading edge at all speeds. Electromyographic activity does not necessarily correlate with useful mechanical force21 and its presence may simply reflect outflow modulation of a central neural control program22. Or, as we believe, the activity of the SC at even high flight speeds is necessary for its most important role, that of humeral rotation.

The avian shoulder joint is structurally derived and functionally complex. The glenoid, best described as a hemisellar (half saddle) joint, is concavoconvex in configuration, faces dorsolaterally, and articulates with a bulbous humeral head. Jenkins23 reviewed the evolution of this joint in a comparative study and provided a new interpretation of its functional morphology based on a cineradiographic analysis of the wingbeat cycle. The articulation of the humeral head on the dorsally facing surface of the glenoid (the Labrum cavitatis glenoidalis) at the upstroke–downstroke transition allows full abduction of the wing into the parasagittal plane. Our observations further reveal that this extensive abduction of the wing, characteristic of so many flying birds, is accomplished not by simple elevation of the humerus, but primarily by rotation about its longitudinal axis.

In a classic study of the functional organization of the wing24, Sy described humeral axial rotation as a mechanism for the execution of wing upstroke in pigeons and generalized its importance for other relatively small birds with powered flight. To appreciate the functional significance of this rotation, it is necessary to consider the orientation of the humerus to the body axis. When viewed from above (in dorsal view), the angle formed by the long axis of the humerus and the bird's longitudinal axis when the wing is in the downstroke–upstroke transition position is not 90° as might be expected: the humerus is retracted back towards the body to form an acute angle with the long axis of the back. At the downstroke–upstroke transition in starlings, for example, this angle is at its maximum of 55–60° during early upstroke, and decreases to its minimum of 25–30° (that is, the humerus is retracted). Simple elevation of a retracted humerus orients the wing's ventral surface posterolaterally instead of in the more functional lateral position. Simultaneous rotation and elevation of the humerus in early upstroke is necessary so that subsequent extension of the hand and forearm during late upstroke will orient the fully outstretched wing in the parasagittal plane; that is, the wing's ventral surface will face laterally, the position appropriate for the beginning of the subsequent downstroke. Humeral rotation/retraction as described here is key to the execution of high-amplitude, high-frequency wingbeats in the European starling, a generalized member of the order Passeriformes, and may also be important in other birds, including those with longer wings, as well as for other flying vertebrates. For example, in a study of the shoulder of the Cretaceous pteryodactyloid pterosaur Santanadactylus brasilensis, which has an estimated body mass of 3.9–7.3kg and a total wingspan of 4.7m, humeral rotation was proposed to be the key to wing elevation in this species25.

In the seven preserved specimens of the Jurassic bird Archaeopteryx, there is no evidence of a derived SC with a dorsally inserting tendon, as indicated by the lack of a triosseal canal or an acrocoracoid8,9,26,27. The lack in Archaeopteryx of this organization, and therefore of rapid humeral rotation, may have restricted its ability to execute rapidly the high-amplitude wing movements associated with slow flight, take-off and landing in modern species of similar size. The early fossil record of birds since the Jurassic demonstrates a number of advanced characteristics that signal evolutionary improvements towards powered flight. For example, the glenoid of Archaeopteryx and the glenoid of the younger Lower Cretaceous Sinornis both face laterally8,9,23, which presumably limited extensive abduction of the wing into the parasagittal planeinboth species. In the recently described series of Enantiornithineandnon-Enantiornithine Mesozoic birds (that is, Eoalulavis, Neuquenornis, Cathayornis, Concornis, Ambiortis and Iberomesornis)10,12,15,28,29,30, the presence of an elongated coracoid, furcula and scapular acromion suggests that a derived SC may have been present, although an acrocoracoid or triosseal canal has not been identified de facto.

Our reinterpretation here of the primary action of the SC, that of high-velocity humeral rotation, explains a new aspect of the evolutionary significance of its highly derived morphology in birds. The selective advantage of high-velocity rotation of the humerus to position the wing appropriately for downstroke was undoubtedly an important advancement to powered flight which contributed to the extensive adaptive radiation witnessed in modern birds.


Forces of rotation and elevation. Under deep anaesthesia (ketamine 60mg per kg, xylazine 6mg per kg, with supplemental ketamine given as needed), we bisected the latissimus dorsi and rhomboideus muscles to expose the brachial plexus and isolate the nerve to the SC. A trachea tube was inserted to provide unidirectional ventilation (80% O2, 20% N2) after opening the posterior air sacs. All other components of the brachial plexus were severed to prevent stimulation of adjacent muscles, and the nerve to the SC was mounted on silver bipolar electrodes. We stabilized the bird by clamping the sternal keel, proximal coracoid and vertebral margin of the scapula to a heavy frame and body temperature was maintained at 39°C with warmed avian ringers and a heat lamp. A silver wire (0.38mm diameter) was threaded through a small hole drilled in the deltopectoral crest and attached to a Grass T4 force transducer. The humerus was stabilized by inserting a needle (23 gauge) into its distal shaft, thus allowing ‘free’ rotation about the long axis of the humerus while restricting elevation. At wing positions corresponding to the downstroke–upstroke transition and mid-upstroke, a supramaximal stimulus (0.2-ms pulse, 60Hz, 500-ms train) was delivered to the nerve of the SC. Substantial torque was produced by the SC around the longitudinal axis of the humerus in both positions. After removing the needle from the shaft of the humerus, we measured forces of humeral elevation by securing a piece of 5-0 surgical silk (compliance, 0.45μmN−1cm−1) to the humerus at midshaft. This configuration detected the elevational component of force but allowed the humerus to rotate on its long axis. Forces of elevation were consistently lower than forces of rotation.

Excursion of the humerus. We measured the total in situ rotational excursions of the humerus during tetanus of the SC for two starlings. We measured humeral rotation by placing a 23-gauge pin guided by a rack and pinion through a small hole drilled in the distal head of the humerus. This pin served as a pivot for rotation while restricting the elevational component of movement. We placed a 26-gauge pin perpendicular to the long axis of the humerus, which served as a dial with which to measure the degree of rotation. We removed the 23-gauge pin and allowed the humerus to move during stimulation to measure humeral elevation. We stimulated the nerve tetanically (60Hz; 500-ms train duration) and measured humeral rotation with a protractor.

Active and passive length–force. We measured the muscle's passive and active properties at a series of lengths (Fig. 3, top; abscissa) within the muscle's normal in vivo excursion. We established the active length–force curves from maximal twitch responses (single stimulus, 0.2-ms pulse) and measured maximum whole-muscle tetanic force (60Hz, 500-ms train) at the end of each experiment at the length coincident with maximum twitch force. After isolating the nerve to the SC (as already described), we deflected the deltoid and propatagialis muscles from the dorsal aspect of the humerus, cut the bone around the tendon's site of attachment, and removed the bone chip together with the tendon. With the bird stabilized to a heavy frame by clamping the sternal keel, proximal coracoid and vertebral margin of the scapula, the bone/tendon structure was secured to the force transducer with a silk tie (compliance, 0.45μmN−1cm−1). A force transducer (Grass FT4) was mounted on an adjustable rack and pinion (millimetre calibration), which allowed us to change the muscle's length in millimetre increments. We generated passive length–force curves by lengthening the muscle in 1.0-mm increments over its physiological working range. We returned the muscle to the length of ‘zero’ passive force after each measurement before pulling it to a new length. The active length–force curve was generated by following the same length-change protocol and delivering a single supramaximal stimulus (0.2ms) to the nerve at each length. We correlated absolute muscle length to wing angle of elevation and depression (Fig. 3, bottom; abscissa) by manipulating the contralateral wing in the bird after killing (sodium pentobarbital, 100mgkg−1). We dissected the origin of the SC from its attachment on the coracoclavicular membrane and sternum and attached it to a rack and pinion. This configuration allowed us to manipulate wing elevation, depression and rotation, and to correlate these with changes in tendon excursion. We manipulated the humerus in 10° increments above and below the horizontal.