Guineafowl with a twist: asymmetric limb control in steady bipedal locomotion

Analyses of avian bipedalism typically focus on steady locomotion. Birds are predominantly studied moving forward over level ground at relatively constant speeds in both kinematic (Cracraft, 1971; Jacobson and Hollyday, 1982; Muir et al., 1996; Gatesy and Biewener, 1991; Gatesy, 1999a; Abourachid, 2000,, 2001; Reilly, 2000; Verstappen et al., 2000; Rubenson et al., 2007; Nyakatura et al., 2012; Provini et al., 2012; Stoessel and Fischer, 2012) and kinetic (Clark and Alexander, 1975; Alexander et al., 1979; Roberts et al., 1998; Hancock et al., 2007; Goetz et al., 2008; Nudds et al., 2010; Rubenson et al., 2011; Andrada et al., 2013, 2014) analyses. Compared with the uniformity of locomotion on a treadmill or straight trackway, the inherent variability of unsteady behaviors is much more difficult to characterize. Consequently, examinations of birds accelerating (Roberts and Scales, 2002, 2004), maneuvering/turning (Jindrich et al., 2007; Kambic et al., 2014), and running over uneven terrain (Daley and Biewener, 2006; Daley et al., 2009; Birn-Jeffery et al., 2014; Gordon et al., 2015) remain relatively uncommon.

Steady locomotion is typically assumed to be symmetrical in striding bipeds. Although out of phase, right and left limbs are thought to mirror each other to a large degree. Most researchers would likely concur that absolutely symmetrical operation of the limbs is rare, as there are many potential sources of functional imbalance. These sources include deviations from bilateral symmetry in the skeleton and soft tissues (e.g. Van Valen, 1962), shifting of the center of mass due to uneven visceral loads or from changes in head and tail position, left–right differences in step length or duty factor when one leg differentially accelerates or decelerates the body, and irregular cycles of yaw, pitch and roll among strides (Gatesy, 1999a; Rubenson et al., 2007; Jindrich et al., 2007; Abourachid et al., 2011; Stoessel and Fischer, 2012). However, these factors are not usually considered significant enough to warrant special attention or to override the general assumption of symmetry. For instance, many studies measure the motion, muscle activity or mechanics of a single limb under the premise that these data are representative of both limbs. Asymmetry in these cases will be a component of the calculated standard deviations, undifferentiated from measurement error, stride-to-stride variation in speed, and individual variation.

While collecting data for a 3D kinematic study of avian walking and running, we were surprised to find major deviations in the orientation of the pelvis relative to the direction of movement. Our preliminary results revealed that birds could hold their position on a treadmill while maintaining a net yaw. The implications of deviating the body's longitudinal axis away from the direction of travel led us to a new set of questions about symmetry and differential limb control. How asymmetrical is avian bipedalism during steady locomotion? When symmetry is broken, how are the many degrees of freedom (DoF) within joints, among joints and among limbs coordinated to move the two legs differently?

Until recently, such questions were difficult to answer. With the development of XROMM (X-ray reconstruction of moving morphology; Brainerd et al., 2010; Gatesy et al., 2010) we now have the ability to measure six degree of freedom skeletal kinematics at high resolution. Our 3D analysis of maneuvering locomotion in helmeted guineafowl (Numida meleagris) revealed the critical role of femoral and tibiotarsal long-axis rotation (LAR; Kambic et al., 2014). Hip and knee LAR were critical for achieving the non-planar poses required to transversely shift and reorient the body. We hypothesized that a closer look inside steady, yet asymmetrical, forward locomotion might reveal comparable non-sagittal foot motions governed by similar joint control strategies.

All procedures conducted with animals were approved by the Institutional Animal Care and Use Committee at Brown University. Six helmeted guineafowl, Numida meleagris (Linnaeus 1758), individuals (1.4±0.19 kg) were acquired from a local source and housed as a group with food and water available ad libitum.

The methods used for analyzing X-ray videos, animating bone models and applying joint coordinate systems in this paper are identical to those described in Kambic et al. (2014). Briefly, birds were surgically implanted with multiple conical carbide markers in the pelvis, femur, tibiotarsus and tarsometatarsus. The first two birds were implanted only in the right limb, the third in the right limb and left femur. The final three individuals were bilaterally implanted with markers in both limbs. After sufficient recovery time, individuals were recorded in the W. M. Keck Foundation XROMM facility located at Brown University. This system uses two EMD Technologies 425 model EPS 45-80 X-ray generators connected to Varian model G-1086 X-ray tubes suspended on ceiling-mounted telescoping cranes. The X-rays are captured by Dunlee model TH9447QXH590 image intensifiers (40.64 cm diameter) mounted on mobile-arm bases. The image intensifiers are backed with Phantom v10 high-speed digital video cameras (Vision Research, Wayne, NJ, USA). We recorded at 1760×1760 pixel resolution. Additional Phantom v9.1 cameras were used to capture light video at 1600×1200 pixel resolution. All videos were recorded at 250 frames s⁻¹ and 1/2000 s shutter speed.

Three individuals were recorded within the acrylic enclosure (29.5 cm wide×100 cm long×48 cm high) of a custom-built hand-crank treadmill at speeds up to 1 m s⁻¹. Bilaterally marked individuals were recorded on a DC5 model Jog-a-Dog motorized treadmill (JOG A DOG, LLC, Ottawa Lake, MI, USA) within a similar enclosure (35 cm wide×106 cm long×48 cm high). The treadmill was oriented between the horizontally arranged X-ray systems, providing two 45 deg oblique views (Fig. 1A,B). Standard light video cameras recorded anterior and lateral perspectives (Fig. 1C,D). Individuals were run for short durations (∼15–30 s) and had several minutes of rest in between trials while the files were evaluated and saved. Trials were deemed suitable for keeping if the sequence included a number of consecutive strides where all markers were visible in both X-ray videos. If individuals appeared to be tiring, data collection would be postponed until the next day. Birds were generally run without interference. To prevent slowing birds from dropping out of the X-ray volume, an operator standing behind the treadmill would lightly contact the tail with a nitrile glove hanging loosely from a dowel. The exception to this protocol was for a single trial where a wedge of foam was placed at the front of the running box to induce a larger yaw. The foam formed a 45 deg angle with the treadmill, and the treadmill speed was lowered such that the individual could maintain a large yaw while progressing forwards. When data collection for an individual was completed, the individual was induced with 5% isoflurane and killed with Beuthanasia.

Bilaterally marked individuals were also recorded running over a trackway (34 cm wide×310 cm long) positioned the same way as the treadmills between the X-ray systems. The sides of the trackway were constructed from opaque siding, with a 70 cm section of acrylic in the middle of the trackway to allow light video recording. This transparent section coincided with the X-ray field of view.

For treadmill data, trials to be analyzed were chosen for their length and consistency in keeping markers in view. The longest trials in which the individual remained within the X-ray volume were preferred. Light video was not used to identify trials for marker tracking, and we made no attempt to infer pelvic yaw before a trial was analyzed. ‘Steady locomotion’ was inferred based on the limitations of the X-ray volume coupled with the consistency of the treadmill speed. Trackway sequences were chosen for analysis without regard for speed or pelvic yaw.

Data were analyzed using the marker-based XROMM workflow (Brainerd et al., 2010; xromm.org). Bone models were segmented using OsiriX (v.4.1.2, Geneva, Switzerland; Rosset et al., 2006) from CT scans derived from a hospital scanner (Philips Medical System, Best, The Netherlands) at 512×512 pixel resolution and 0.625 mm intervals, with one exception in which a Fidex micro-CT scanner (Animage, LLC, Pleasanton, CA, USA) at 512×512 pixel resolution and 0.456 mm intervals was used. Geomagic Studio 2013 (3D Systems, Morrisville, NC, USA) was used to clean bone models. XrayProject, a set of XROMM tools for Matlab (MathWorks, Natick, MA, USA), was used for undistorting X-ray videos, calibrating cameras, and 3D marker tracking. These tools were also used to generate transformation matrices to position and orient the bone models in 3D space (Fig. 1). Bone models were animated in Maya 2010 (Autodesk Inc., San Rafael, CA, USA). The raw data, including X-ray and light videos, and calibration and distortion grid images, were uploaded to the X-ray Motion Analysis Research Portal (xmaportal.org) and will be made publicly available on publication. As a measure of marker tracking precision, the standard deviation of intermarker distances within single bones (Tashman and Anderst, 2003; Brainerd et al., 2010) averaged ±0.222 mm.

Joint angles were calculated according to joint coordinate systems (Grood and Suntay, 1983; Wu et al., 2002) that were set up identically to those described in the appendices of Kambic et al. (2014). The sign conventions were as follows. Pelvic yaw was positive to the left, pelvic pitch was positive when raising the cranial end, and pelvic roll was positive when rolling the right hip higher than the left. At the hip, extension was positive, abduction was positive and external rotation was positive. At the knee, extension was positive, adduction was positive (note that this differs from the hip and ankle), and external rotation was positive. Finally, at the ankle, extension was positive, abduction was positive and external rotation was positive.

Stance and swing phases were initially determined using light videos, with stance being defined as the first frame showing ground contact by digit III, and swing beginning with the first frame where the foot was not in contact with the ground. These sequences were used to guide the creation of rules to infer stance and swing phases from knee joint rotations for the sequences without light video. In the test sequences used to create these rules, the knee method results were typically within 1–2 frames of the light video results. We use the term ‘excursion’ to refer to the overall range (maximum minus minimum) of angles during a stride or phase. A ‘net excursion’ is the difference in angle over a phase (end minus start).

To characterize mediolateral displacements of the foot relative to the pelvis, a transverse distance (Kambic et al., 2014) was calculated that measured the distance from the distal end of the tarsometatarsus to the median sagittal plane. For both limbs, transverse distance was zero at the midline, negative medial to the median plane, and positive lateral to the plane.

None of the six individuals studied showed any signs of lameness, limping or obvious external gait asymmetry following surgery. The subjects moved steadily over a range of treadmill speeds (0.6–1.9 m s⁻¹). Duty factor approached 0.5 at the highest speed, but aerial phases were seldom observed in these walking and ‘grounded running’ gaits (Rubenson et al., 2004; Hancock et al., 2007; Nudds et al., 2011). We were only able to analyze sequences in which the subjects successfully maintained their position within the limited biplanar X-ray volume (ca. 9800 cm³, equivalent to a cube with 21.4 cm sides). Additionally, recording time was limited to a maximum of 10 s per trial. Despite these challenges, full sequences containing as many as 23 consecutive strides were recorded.

Measurements of pelvic yaw confirmed the initial observation that birds routinely deviate substantially from their direction of travel. The direction and magnitude of this deviation can be difficult to detect from a lateral view (Fig. 2A–C). Data from six birds for 10 treadmill sequences composed of over 12,000 frames from 197 steps (116 right, 81 left) were widely distributed (Fig. 2D). For unknown reasons, all individuals had a tendency to yaw to the right (see Discussion). Yaws of 3–5 deg to the right were most common, but values in the teens and twenties were regularly present across the trials analyzed. The broad distribution was not driven solely by the slowest walking sequences (Fig. S1). The bird in the fastest trial we recorded (1.9 m s⁻¹) reached a yaw of greater than 11 deg to the right. Although both overall magnitude and right skewing of yaws may have been an artefact of our two different treadmills, birds moving freely down a trackway also exhibited yaw variation (Fig. 2D).

Within an individual sequence, three-rotational DoF kinematic data from the pelvis and both limbs of a bird moving at 0.86 m s⁻¹ revealed consistent patterns over seven and a half strides (Fig. 3; Movie 1). Variability in pelvic kinematics occurred at two major scales. At the level of the stride, yaw, pitch and roll all showed repeating patterns. During a given stance phase, the pelvis pitched down for the first quarter, pitched up for the middle two quarters, and then pitched down for the last quarter (best seen in Fig. 3, strides 4, 5). The body initially rolled towards the stance limb. Then, about halfway through stance, the body began rolling away from the stance limb (best seen in Fig. 3, strides 2, 4). Finally, the pelvis initially yawed away from the stance limb and then yawed toward it for the remainder of the phase (best seen in Fig. 3, strides 3, 7).

Over the course of the 4 s sequence, intra-stride patterns were repeated, with pitch and roll oscillating about relatively consistent values. In contrast, there were persistent deviations in yaw that exceeded intra-stride variation. The bird began the sequence closely parallel to the tread, but yawed slightly to the left (Fig. 3, stride 1). Soon after, a yaw to the right (negative) accrued until the bird's body was facing more than 20 deg away from its direction of travel (Fig. 3, strides 2–3). The individual then maintained a right yaw over several strides (Fig. 3, strides 4–7).

The hip, knee and ankle (intertarsal) joints underwent substantial flexion/extension (FE). The hips flexed briefly at the beginning of stance and then extended through most of the rest of the phase. Flexion began just prior to toe off and then the hips re-extended just before toe down. The knees flexed considerably throughout the majority of stance, with a variable degree of late-stance extension. Flexion continued past toe off until the knees reversed direction and extended for the rest of swing. The ankles underwent flex–extend–flex sequences during stance, followed by large flex–extend excursions in swing. Abduction/adduction (ABAD) angle changed comparatively little at any of the joints, similar to the minimal ABAD excursions during maneuvering (Kambic et al., 2014).

LAR decreased progressively from hips to knees to ankles. At the hips, LAR excursions showed high variability (N=15, 9.8±8.5 deg, mean±s.d.). Strides earlier in the sequence (Fig. 3, strides 1–2) had less femoral rotation than later strides, in which hip LAR excursions equaled or even exceeded hip FE excursions for several steps (Fig. 3, strides 3–7). At the knees, LAR excursions were more consistent (N=14, 8.6±4.1 deg, mean±s.d.), but differed slightly between the right and left sides. The right knee internally rotated through most of stance, whereas the left knee first externally and then internally rotated. Both knees underwent net external rotation in swing. Some LAR occurred at the ankles, but excursions during both stance and swing were small compared with those for the knees and hips and were less variable (N=8, mean 3.9±3.8 deg, mean±s.d.).

We compared limb kinematics from low-yaw and high-yaw sequences to determine whether persistent pelvic yaws of large absolute value were associated with specific joint rotations. Only hip LAR underwent large-scale changes in pattern and magnitude that echoed those of yaw.

The relationship between hip LAR and pelvic yaw is more easily observed if the two are plotted together in isolation (Fig. 4). Data from the sequence plotted in Fig. 3 (Fig. 4A) show that the hips used a biphasic LAR pattern initially, when the bird's pelvis and body deviated only a few degrees to the left from its direction of travel. During stance, hip LAR curves were W-shaped as both limbs underwent small internal–external–internal–external excursions (Fig. 4A, strides 1–2). In swing, an external–internal sequence gave each hip LAR curve an upside-down U shape (Fig. 4A, strides 1–2), but there was little or no net excursion during either phase.

Later in the sequence (Fig. 4A, strides 3–7), when the bird yawed well to the right, the biphasic hip LAR patterns shifted to more monophasic waves. During this portion of the sequence, the hips rotated through large net excursions, but in opposite directions. The result of these changes is that stance and swing together created monophasic U shapes (Fig. 4A, strides 3–7). LAR traces from the right and left hips were synchronized in time because the hips rotated in opposite directions at equivalent points in their stride cycles. A 0.92 m s⁻¹ sequence in which the bird maintained a right yaw (Fig. 4B) shows the monophasic hip LAR pattern very clearly. Here, the right and left hips were even more synchronized in time, with traces that were frequently superimposed.

Even in a faster (1.59 m s⁻¹), relatively low-yaw trial (Fig. 4C), body orientation and hip LAR were correlated. Throughout the final two-thirds of the sequence, when the bird maintained a low yaw, the hips exhibited a biphasic LAR pattern with little net LAR in either stance or swing (Fig. 4C, strides 5–9). The first third of the sequence showed LAR asymmetry at yaws of only ca. 10 deg (Fig. 4C, strides 1–4). Despite being biphasic, the hips underwent significant net LAR excursions during stance and swing. Patterns of low-yaw, biphasic LAR and high-yaw, monophasic LAR were consistent across the hips of all individuals.

The relationship between hip LAR and yaw can be summarized by plotting net LAR excursion against average yaw during each stance phase for all treadmill sequences (Fig. 5). Larger net stance hip LAR excursions were correlated with greater deviation from the direction of travel. For example, during stride 2 of Fig. 4B, the right hip underwent 19 deg of net internal rotation while at an average right yaw of 12 deg (Fig. 5, red circle). Right and left hips responded to yaw similarly in net LAR excursion magnitude, but in opposite directions. Yaws to the right co-occurred with net internal rotation at the right hip during stance, but net external rotation at the left hip during stance, and vice versa.

Here, we report that guineafowl can substantially vary their yaw during steady locomotion. In sequences in which birds maintained position within a limited X-ray volume, yawing occurred over a range of speeds. Multiple individuals in multiple runs surpassed 20 deg of yaw compared with their direction of travel. This yaw offset was overlaid on the normal, rhythmic, intra-stride yaw pattern. Guineafowl varied pelvic yaw both on a treadmill and when moving over a trackway. Marker-based XROMM analysis revealed measurable LAR at the hip and knee regardless of yaw (Figs 3, 4). Previously, we found that LAR at these joints was crucial for sidesteps, yaw maneuvers and slow turns (Kambic et al., 2014). The results presented here demonstrate that the role of LAR is not limited to maneuvering; LAR is an essential component of avian steady locomotion as well.

Maintaining a non-zero pelvic yaw on a treadmill has important consequences for the relative motion of the body and feet. Low and high yaw steps (strides 1 and 4 from the sequence plotted in Fig. 3, Fig. 4A) are compared in Fig. 6. In lab space, a stance foot stays fixed on the tread, which it follows regardless of pelvic orientation (Fig. 6A,C). If pelvic yaw is small, the foot's trajectory closely parallels the body's longitudinal axis (Fig. 6A). The more the pelvis is yawed away from the direction of travel, the more the foot's path is skewed relative to the body (Fig. 6C). This difference in trajectory is emphasized when viewed from a pelvic reference frame, as if looking down the body's median plane from above (Fig. 6B,D). At low yaw, the transverse distance from the distal tarsometatarsus to the median plane is small and changes little over the course of stance (Fig. 6E). In contrast, at a larger yaw, the foot takes a diagonal path relative to the body midline, moving from medial to lateral for the right leg and a right yaw (Fig. 6E). The transverse distance from the median plane starts small and increases over 5 cm through stance (Fig. 6E).

Hip LAR patterns closely match transverse distance patterns. During the low yaw step, hip LAR forms a W with similar beginning and ending values (Fig. 6F, i and ii). During the high yaw step, the hip begins more externally rotated, internally rotates through stance, and ends at a much lower angle (Fig. 6F, iii and iv). The similarity of the transverse distance and hip LAR traces suggests that hip LAR is used to modulate transverse distance during forward locomotion. Pooled data from several steady treadmill trials confirm this kinematic relationship (Fig. 7, filled circles). Larger net hip LAR excursions are associated with foot paths that deviate further away from or towards the median plane. Even more extreme transverse distance and hip LAR excursions (Fig. 7, open squares) were achieved when an angled foam barrier at the front of the treadmill induced a bird to walk slowly at yaws of 35–65 deg.

Yaws of sufficient magnitude correlate with a monophasic hip LAR pattern, in which the femora undergo substantial net LAR excursion during stance and swing (Fig. 4B, Fig. 5). At high yaws, the right and left hips internally and externally rotate together, generating similar curves when plotted versus time. Perhaps counter-intuitively, these shared monophasic patterns are evidence of kinematic asymmetry. Right and left hips internally and externally rotate simultaneously, but the two limbs are 180 deg out of phase with respect to the stride cycle. One hip internally rotates in stance while the other internally rotates in swing. When both hips externally rotate, the side that was in stance has switched to swing and the side that was in swing has switched to stance. In order to act symmetrically, both limbs would need to undergo similar LAR excursions during the same phase of the stride, thereby alternating in time. Plots of hip, knee and ankle FE (Fig. 3), and hip LAR at low yaws (Fig. 4C, Fig. 5) demonstrate this kinematic symmetry. In contrast, at high yaws the limbs both retract in stance and protract in swing, but undergo hip LAR in opposite directions.

When walking with a yawed pelvis, the pattern of femoral LAR resembles the coordination reported for guineafowl executing sidestep maneuvers (Kambic et al., 2014). During a sidestep, birds sequentially spread and converge the feet transversely (Fig. 8A; Movie 2). Mediolateral motion of the limbs is similarly required for forward movement while yawed, but the amount of mediolateral skewing required depends on pelvic orientation. When the individual is induced to walk slowly at extreme yaws around 45 deg, mediolateral and craniocaudal foot excursions are close to equal (Fig. 8B). At progressively lower yaws, less and less skewing of the foot trajectories is required (Fig. 8C,D). Therefore, forward progression with substantial yaw (Fig. 8B,C) can be seen as intermediate between the two extremes of a crab-like sidestep (Fig. 8A) and symmetrical striding with little yaw (Fig. 8D).

The sidestep model of yawed forward progression explains the synchronized, monophasic pattern of hip LAR characteristic of high yaw locomotion (Fig. 4B; Fig. 7, red circle). Net hip LAR produces mediolateral excursions that spread and converge the feet to generate appropriately skewed trajectories. For example, when yawed to the right (Fig. 8C), the right limb moves from medial to lateral in stance by net internal hip LAR. Simultaneously, the left limb moves from medial to lateral in swing, again using net internal hip LAR to widen the transverse distance between the feet. Both femora then externally rotate together to return to their narrower configuration. Rather than alternating, the striding limbs exhibit hip LAR angles that are coordinated in time. Right–left asymmetry is actually increased by making hip LAR patterns more identical.

Our finding that guineafowl routinely yaw their bodies compared with their direction of travel may be less significant if this behavior is not ‘natural’. We do not know whether or how often birds yaw while walking and running in the wild. In the lab, one concern with marker-based XROMM is surgery-induced asymmetry. Yet our subjects yawed in similar ways whether one or both limbs had been implanted. Based on head orientation, birds appear to have been looking toward the same side they were yawing most of the time. Head deviation may provide evidence that the body is yawed, although it is not yet clear whether head motion precedes or follows body rotation. Further analysis of the visual saccades that transition between ‘fixed’ head orientations relative to the body and direction of travel is warranted (Dunbar, 2004).

The consistent treadmill bias to the right is suggestive that one or more features of the lab – personnel, lighting, equipment, reflection, etc. – were drawing the individuals' attention, although data from a trackway in the same location and orientation do not show the same pattern. It is possible that the researcher operating the treadmill from behind influenced asymmetry, although both hand-cranked and motorized systems yielded similar results. Regardless of the cause of the right bias, yawing behavior was performed at will by all individuals on two different treadmills as well as on a trackway. If solely a lab-based phenomenon, orienting the body away from the direction of travel may be common in other studies of avian kinematics as well.

For 2D analyses of motion from lateral perspective recordings, errors in FE joint angles will be minimal when yaws are small. However, knowing that this behavior occurs should advise caution. If the behavior is steady and cyclical, it is common to measure kinematics and physiological parameters (muscle activity, muscle strain, bone strain, etc.) for a single limb and assume that the other limb is behaving similarly at equivalent points in the stride. Indeed, our initial XROMM markers were implanted in only one limb. Given our results, restraint is warranted in assuming symmetrical limb motion if yaw is not measured. Yaw variation should also be considered when summarizing angular data as averages. Assuming symmetry may mask explainable inter-stride variation, as numerous researchers working on human locomotion have found (see Hausdorff, 2007, for a review).

Despite these cautions, there are potential benefits to individuals routinely orienting away from their direction of travel. The patterns of kinematic variation with yaw are intriguing from a motor control standpoint. Birds graded smoothly between biphasic and monophasic hip LAR patterns, and blended elements of each depending on the circumstances of the stride. The modulation of LAR, combined with the similarity of yawed locomotion patterns to sidestep patterns, may be evidence that cranial–caudal limb motion and mediolateral limb motion are controlled by separate sets of muscle activation patterns, termed muscle synergies (e.g. d'Avella et al., 2003; Ting and MacPherson, 2005). Many researchers have found evidence of joint coordination in steady locomotion (e.g. Ogihara et al., 2014), and we identified similar integration of hip and knee LAR during maneuvering (Kambic et al., 2014). However, in this study we did not find a strong correlation between hip and knee LAR patterns – a result that should be investigated further.

We seek to learn more about the muscles responsible for controlling femoral LAR (Hutchinson and Gatesy, 2000) as well as the contribution of muscles having both FE and LAR moments about the hip. Based on their stance phase activity pattern, muscles such as the iliotrochantericus caudalis and medius appear to be ideally situated to induce internal hip LAR (Gatesy, 1994,, 1999b). Yet, Hutchinson et al. (2015) found that several of the most significant hip extensors have large internal rotation (called medial rotation in their paper) moment arms at relevant joint angles as well. Determining how these muscles (and active antagonists) interact with inertial loads and passive forces to modulate LAR during stance will require a more sophisticated, fully 3D approach. Likewise, the control of LAR in swing awaits further experimental and dynamic modeling work.

The resolution of our XROMM approach has revealed how the outwardly appearing ‘steady’ locomotion of striding birds can be asymmetrical in previously under-appreciated dimensions. Joints such as the avian hip and knee operate in 3D and cannot be reduced to hinges. However, our understanding of the kinematic mechanisms for controlling yaw during forward locomotion also remains incomplete. Although we have identified patterns of LAR excursion associated with yaw deviations, the coordination among joints and degrees of freedom used to change yaw await further analysis.

We thank Elizabeth Brainerd, David Baier and the Brown Morphology Group for their work on XROMM. Ariel Camp, Peter Falkingham, Erika Giblin and Angela Horner assisted with data collection. John Hutchinson, Danny Miranda and Mike Rainbow provided advice on joint coordinate systems. We extend thanks to Kia Huffman for help with the XMA Portal. Farish Jenkins, Jr and William Amaral originally designed conical markers that were adapted for this study, with additional fabrication advice from Amy Davidson. Comments and suggestions from two anonymous reviewers significantly improved the final manuscript.