Journal of Neurophysiology

Head Stabilization by Vestibulocollic Reflexes During Quadrupedal Locomotion in Monkey

Yongqing Xiang, Sergei B. Yakushin, Mikhail Kunin, Theodore Raphan, Bernard Cohen


Little is known about the three-dimensional characteristics of vestibulocollic reflexes during natural locomotion. Here we determined how well head stability is maintained by the angular and linear vestibulocollic reflexes (aVCR, lVCR) during quadrupedal locomotion in rhesus and cynomolgus monkeys. Animals walked on a treadmill at velocities of 0.4–1.25 m/s. Head rotations were represented by Euler angles (Fick convention). The head oscillated in yaw and roll at stride frequencies (≈1–2 Hz) and pitched at step frequencies (≈2–4 Hz). Head angular accelerations (100–2,500°/s2) were sufficient to have excited the aVOR to stabilize gaze. Pitch and roll head movements were <7°, peak to peak, and the amplitude was unrelated to stride frequency. Yaw movements were larger due to spontaneous voluntary head shifts and were smaller at higher walking velocities. Head translations were small (≤4 cm). Cynomolgus monkeys positioned their heads more forward in pitch than the rhesus monkeys. None of the animals maintained a forward head fixation point, indicating that the lVCR contributed little to compensatory head movements in these experiments. Significantly, aVCR gains in roll and pitch were close to unity and phases were ≈180° over the full frequency range of natural walking, which is in contrast to previous findings using anesthesia or passive trunk rotation with body restraint. We conclude that the behavioral state associated with active body motion is necessary to maintain head stability in pitch and roll over the full range of stride/step frequencies encountered during walking.


Considerable movement is imposed on the head during both bipedal (Crane and Demer 1997; Grossman et al. 1988; Hirasaki et al. 1993, 1999, 2004; Imai et al. 2001) and quadrupedal locomotion (Dunbar 2004; Dunbar et al. 2004; Hirasaki and Kumakura 2004). The body rotates and translates as weight is shifted from one foot to the other during the swing and stance phases (Hirasaki et al. 1999; Imai et al. 2001; Leigh et al. 1992; Moore et al. 2001; Pozzo et al. 1990). Therefore compensatory reflex mechanisms that tend to stabilize the head and eyes in space are of considerable importance in maintaining gaze and heading (Grasso et al. 1998; Mergner and Rosemeier 1998; Patla and Vickers 1997; Patla et al. 1991, 1999; Peterson 1988; Pozzo et al. 1990; Vallis et al. 2001). Of these mechanisms, the angular vestibulocollic reflex (aVCR) is probably the most important reflex that compensates for head-on-body movement to stabilize the head in space with a possible contribution of vision.

Previous studies of head stabilization by the aVCR have largely utilized rotational apparatus in which the body was fixed and the head was free to move (Aerts et al. 2000; Guitton et al. 1986; Peng et al. 1996; Peterson et al. 1989, 2001). These include studies of decerebrate animals (Ezure and Wilson 1983; Peterson et al. 1985; Wilson and Schor 1999; Wilson et al. 1990), alert animals [cats: (Goldberg and Peterson 1986; Guitton et al. 1986); rabbits: (Fuller 1988); mice: (Baker 2005; Takemura and King 2005); squirrel monkey: (Kubo et al. 1981); rhesus monkey: (Cullen and Roy 2004)], and humans (Guitton et al. 1986; Keshner et al. 1992, 1995; Viviani and Berthoz 1975; Zangemeister et al. 1981). Surprisingly, there was little difference in activation of the aVCR between decerebrate and alert animals (Goldberg and Peterson 1986). In a number of studies on humans, rotated with their trunks fixed but their head free, the gain of the yaw aVCR was small (Keshner and Peterson 1995; Keshner et al. 1995) or negligible (Guitton et al. 1986) at oscillation frequencies <1 Hz and gradually became larger at frequencies between 1 and 3 Hz (Keshner and Peterson 1995; Keshner et al. 1995). Mechanical resonance then became dominant at oscillation frequencies >3–4 Hz (Peng et al. 1996) and there were large phase errors. In cat, the measured aVCR gain was small at <1 Hz, became larger between 1 and 2 Hz, and saturated at higher frequencies (Goldberg and Peterson 1986). Very little is known about how well this reflex functions in natural conditions, which was one of the major purposes of this study.

There is reason to believe that vestibular compensatory mechanisms function differently when the body is free to move than when the body is fixed and passively rotated or when animals are anesthetized or decerebrate. In humans, for example, the gain of the angular vestibuloocular reflex (aVOR) is ∼0.5 in normal subjects if they are passively rotated in darkness. The gain of the aVOR is close to unity, however, if one imagines a spatially-fixed target when passively rotated in darkness or if subjects actively rotate their heads in light or darkness (Robinson 1976). Vestibular neurons that code head velocity and project to the neck (McCrea et al. 1999), also respond differently in active and passive conditions. These neurons respond to head velocity when monkeys are rotated passively but no longer code this signal when the head is moved actively (Cullen and Roy 2004; McCrea et al. 1999; Roy and Cullen 2002). This demonstrates that central neural mechanisms associated with active head movements alter the signal processing in the vestibular nuclei and that behavioral state is a critical factor in vestibular function. If correct, this would put into question the limits of gain and phase determined in the previously cited studies on the characteristics of the aVCR that were done with the animals anesthetized, decerebrate, or when subjects were rotated with the body fixed.

Sub-human primates walk both quadrupedally and bipedally (Aerts et al. 2000; Hirasaki and Matano 1996; Mori et al. 2001). Studying vertical head translation and head pitch, Hirasaki et al. and Dunbar and colleagues found that vertical head translations and head pitch are related to the step cycle (Dunbar 2004; Dunbar et al. 2004; Hirasaki and Kumakura 2004; Hirasaki et al. 1999). The amplitudes of vertical head translation in the Japanese macaques were similar during bipedal and quadrupedal locomotion, ranging from 1.7 to 2.0 cm (Hirasaki et al. 1999). Peak vertical accelerations of the head ranged from 0.3 to 0.7g, the angular deviation in pitch was from 4 to 5°, and the peak angular pitch velocities were ∼60–75°/s. Little is known of the angular or translation characteristics of their head movements in three dimensions during natural locomotion, however. A second aim of this study was to elucidate the gain and phase characteristics of the aVCR in three dimensions during natural locomotion.

There is also relatively little information about the characteristics of the linear VCR (lVCR). It has been posited that the lVCR pitches the head to counter vertical body movement to stabilize the head in the pitch plane (Hirasaki and Kumakura 2004; Hirasaki et al. 2000; Imai et al. 2001; Moore et al. 2001) to produce a stable head fixation point (HFP) (Berthoz and Pozzo 1994; Hirasaki and Kumakura 2004; Pozzo et al. 1990, 1991a). (The HFP is defined as the point where the head naso-occipital axes coincide in space for different head positions over locomotion cycles.) The lVCR is also believed to counter lateral body movements to maintain a stable head fixation position in yaw (D. M. Moore et al. 1991; S. T. Moore et al. 1991). Hirasaki found that Japanese macaques and gibbons had a HFP from 150 to 300 mm that was larger in the gibbon (Hirasaki et al. 1999). He concluded that both quadru- and bipedal locomotion produce essentially similar vertical translations and pitch rotations of the head and that the magnitude and speed of these movements in the pitch plane were sufficient to activate the compensatory lVCR. In the vervet monkey, however, the head frequently pitched up as the trunk moved up (Dunbar 2004; Dunbar et al. 2004). This would put the HFP in the back of the head where it could not stabilize gaze. The final purpose of this study, therefore, was to determine whether the animals we studied had a HFP that could be used to stabilize gaze and whether the phases of head linear acceleration and head angular velocity were appropriate for activation of the lVCR. Preliminary results on head movement have been presented in abstract form (Xiang et al. 2006, 2007b).


Seven monkeys [Macaca mulatta (Rhesus): Rh426, Rh488, Rh652, and Rh002 and M. fasicularis (Cynomolgus): Cy091, Cy101, and Cy115] were used in this study. The animals were of approximately the same size, weight, and age. The kinematics of the limb movements of Rh426, Rh488, Cy091, and Cy101 have been reported previously (Xiang et al. 2007a). The experiments conformed to the Guide for the Care and Use of Laboratory Animal, and were approved by the Institutional Animal Care and Use Committee of the Mount Sinai School of Medicine.

Surgical preparation

Monkeys were prepared under anesthesia and sterile surgical conditions with a head mount that held light-emitting diodes (LEDs) that determined head position in space (Fig. 1A). Details of implantation of the head mount have been given in previous publications (Sirota et al. 1988; Yakushin et al. 2000). The cap or head mount, which weighed ≈150 g, was attached to the skull and served as a rigid body to detect the angular and linear position of the head. The cap added ≈15% to the inertia of the head. This would not significantly affect the dynamics of the head, which are predominantly determined by the elasticity and viscosity of the soft tissues, including the muscles (Goldberg and Peterson 1986; Keshner et al. 1999; Peng et al. 1999; Zangemeister et al. 1981). The cap made it possible to determine the spatial coordinates of the animal's head precisely in three dimensions while it was walking.

FIG. 1.

A: coordinate frames for eye, head, and body movements in space of a monkey viewed from the right side. Spatial coordinates were determined from 4 LEDs embedded in the plate on the wall behind the treadmill. The lower left LED was the origin of the spatial frame. Head coordinates (XH, YH, ZH) were determined from 11 LEDs embedded in the cap on the head. Body coordinates (XB, YB, ZB) were taken from 4 LEDs inserted in 2 plates astride the animal's thoracic spine. Positive rotations for yaw, pitch and roll, shown by the circular arrows, were determined according to the right-hand rule. RFL, LFL, right and left forelimbs; RHL, LHL, right and left hindlimbs. B and C: range of stride frequency (B), and stride length (C) in 2 rhesus monkeys (R002 and R652).


The monkeys were acclimated for several weeks to walk on a treadmill (Key) that was 47 cm wide and 127 cm long. In initial experiments, the monkeys wore a light cloth body suit that held the rigid bodies and LEDs that measured body movement as in a previous study (Xiang et al. 2007a). In later experiments, a 20-cm thermal Neoprene wrap was placed around the chest to hold the rigid bodies. Small 2.5-cm rigid bodies were placed on tape wrapped around the right wrist and ankle to measure step and stride frequencies. During training, animals were intermittently rewarded with sugar pellets, raisins or small biscuits.

Each animal walked about 60 min/day, 3 day/wk during training. Slow and fast treadmill speeds were interspersed to prevent tiring. Gaze was unconstrained, and the monkeys were free to look and orient their head and bodies as they wished. A loose chain attached to a light collar around the neck confined the animals to the treadmill surface, but they otherwise walked without restraint. After the monkeys had attained a steady gait, i.e., when they walked at the velocity of the treadmill and were making regular reciprocal limb movements without bounding or walking sideways, ∼90 s of data were collected at each treadmill velocity.

Coordinate frames

The inertial space-fixed coordinate frame (x, y, z) defined the origin of all other coordinate frames and the x-z (vertical), x-y (horizontal), and y-z (roll) planes (Fig. 1A). The z axis was along the spatial vertical (positive upward), the positive x axis was along the direction of forward walking and orthogonal to the z axis, and the y axis was orthogonal to the x-z plane, positive to the monkey's left.

The head-fixed coordinate frame (XH, YH, ZH) was defined such that the roll axis, XH, was parallel to the naso-occipital axis (positive forward), the pitch axis, YH, was parallel to the inter-aural axis (positive left), and the yaw axis, ZH, was normal to the XH-YH plane (positive upward). The body-fixed coordinate frame (XB, YB, ZB) was defined with XB parallel to the long axis of the trunk (positive forward), YB parallel to the transverse axis (positive to the monkey's left) and ZB normal to the XB--YB plane (positive upward). The directions of positive rotation of the body part or head in the spatial coordinate frame followed the right hand rule. Thus yaw rotation around the ZH axis to the left, pitch down around the YH axis in the XH--ZH plane, and clockwise rotation from the animal's point of view were positive.

Rigid bodies on the head, the right and left posterior thoracic walls just astride the spine, the right forelimb (RFL) and the right hind limb (RHL) defined the positions of the head, body and limbs in space. The rigid body on the head was the cap on the head implant (Fig. 1A). The two rigid bodies on the posterior chest wall defined the center of the body at the mid-thoracic level. Although the scapulae can move independently over the posterior thoracic wall with movement of the forelimbs, the rigid bodies on the posterior trunk were fixed to a 20-cm-wide firm elastic band that encircled the thorax. This band moved with the trunk and not with the scapulae, ensuring that the rigid bodies on the back accurately reflected trunk movement. Head rotation relative to space was given in Euler angles in a Fick sequence (yaw, pitch, roll). In some experiments, only one LED was used on the wrist and ankle to measure their linear movements.

Measurement of head, body, and limb movements

Movements of the head, body, right wrist, and right ankle were detected by tracking the position of the infrared LEDs with a video-based motion detection system (Optotrak 3020, Northern Digital). Thin wires connected the LEDs in the rigid bodies to an external strobe unit fixed above the monkey. This unit pulsed the LEDs, and their activation was sensed by the high-resolution, three-camera Optotrak sensors. A central control unit and computer recorded information about the LED positions. The Optotrak software determined the positions of individual LEDs embedded in the rigid bodies (Fig. 1A). Programs were developed to determine the angular and linear positions of the rigid bodies, which represented the head, body, and limbs, relative to space as well as the head movement relative to the body. The accuracy of measurement of horizontal and vertical translations was 0.3 mm. In depth the accuracy was 0.45 mm. The angular accuracy of the device for rotation around any axis was ∼0.1° at this distance from the sensor.

The walking animals were viewed from the side and diagonally from behind by the Optotrak in different experiments. The side view provided information about movements of the head and limbs in spatial coordinates in relation to swing and stance phases (Xiang et al. 2007a). The diagonal view, 45° from the back, as animals walked away from the Optotrak afforded measurement of the center of the trunk and the head in spatial coordinates. Translations and rotations of the rigid bodies relative to space and relative to the center of the thorax were computed during postprocessing of the raw positional data. Rotations of the trunk were also calculated in the same terms as the Fick sequence of the Euler angles of head rotation. In this paper, “head movements” always refer to head movements in space. When head movements relative to the body are being discussed, the notation “head movement re body” will be given.

Simultaneous digital video was also recorded while the animals walked on the treadmill. The synchronous video helped identify sections of data when the animals were walking with a stable gait.

Calibration of head and body coordinate frames

To calibrate the head coordinate frame, the animals were lightly anesthetized and placed in a stereotaxic frame while wearing the cap with embedded LEDs that was used to determine head rotation. LEDs were also placed at the level of the ear bars and in the center of the head at the level of the inferior orbital ridges. The animals and the stereotaxic apparatus were then rotated slowly until each LED was recognized by the Optotrak. The calibrating LED's were then removed, and virtual markers were established that defined the head coordinate frame.

To determine the trajectory of the body, four markers were embedded on two rigid bodies, which formed an oblong rectangle when they were affixed astride the thoracic spine. The rigid bodies were held in place by a Velcro patch to a light 20-cm-wide neoprene tube, which was wrapped around the monkey's chest.

Data processing

Optotrak data, sampled at 90∼100 frame/s, were converted into digital files. The sampling frequency was sufficient to establish the three-dimensional (3-D) angular and linear positions of the rigid bodies on the trunk and limbs and to capture the full spectrum of the body and limb movements, which only extended to ∼10 Hz in our data (Kunin et al. 2007; Xiang et al. 2007a). The 3-D reconstruction was done by the Optotrak software directly from the sampled data. A cubic spline interpolation algorithm (Matlab 7.0, Mathworks) was used to restore occasional dropped frames because of obstruction of view of one or more LEDs from the Optotrak cameras. A five-point average window then acted as a low-pass filter to smooth the data. Approximately 10–15 cycles of gait were used for each condition at each walking velocity to characterize the frequencies, amplitudes, and phases of the translations and rotations of the head and body in pitch, yaw, and roll. Because the animals were not trained to fixate, voluntary yaw head movements could not be prevented, and no measures were implemented to remove them from the records. However, we attempted to use sections of the record in which the animals had regular walking and relatively few spontaneous yaw head movements, as seen in the synchronous video, whenever possible. To calculate peak-to-peak amplitude of head movements, we assumed that the spontaneous head turning would essentially be randomized over >10 cycles, and, further, utilization of cycle averaging would reduce its effect on the more regular head yaw movement. This was not a concern in pitch and roll because there were few, if any, spontaneous head movements in these directions.

The head and body in space were considered as rigid bodies that rotate and translate. The Optotrak software determined these rotations and translations. Translational velocity and acceleration were determined by finding the slope of the linear regression of 11 data points. Rotation matrices of the head and body were determined from the Optotrak data and the axis and angle of the rotation were determined using the Euler-Rodriguez formula (Goldstein 1965; Raphan 1998). Fourier analysis was used to obtain the frequency range over which movements of the head, body, and ankle/wrist were compared.

The angular velocity of the trunk rigid body was computed re space and expressed in the spatial coordinate system. The angular velocity of the head rigid body was computed both re space (expressed in the spatial coordinate system) and re the trunk (expressed in the body coordinate system). The components of the angular velocity vector along the x, y, and z axes were denoted as roll, pitch, and yaw velocities respectively.

Angular velocity vectors were estimated as follows (Kunin et al. 2007): The angular velocity vector corresponding to the kth sample of the trajectory was determined from a window of samples kn,…, k–1, k+1,…, k+n. In each window, n incremental rotation matrices (n = 5 in our implementation) were computed for each pair of samples symmetric with respect to k Math

These incremental matrices were fit by a matrix RΔ (corresponding to the average rotation of the rigid body from one sample to the next within the given window) by minimizing Math where Ω(R) denotes the axis-angle representation of a rotation matrix R. The axis-angle of RΔ divided by the duration of the sampling interval was taken as the estimator of the angular velocity vector (see Kunin et al. 2007 for a complete description of the methodology and the significance of the fits).

To determine the gain and phase of compensatory head movement relative to body perturbation, a FFT was computed on stretches of raw data to produce the power spectrum and phase of the frequency responses of angular velocities of the body re space and those of the head re body. Gains of the compensatory head rotations at each walking velocity are represented as the ratio of the power of head re body velocity divided by that of body re space velocity at the stride frequency for yaw and roll and at the step frequency for pitch. Phases were calculated as the difference of phase angles of head re body velocity and body re space velocity.

Statistical analysis

Parameters measured in this study included angular pitch, yaw and roll and linear displacement in x, y, and z axes of the head and body at a number of walking speeds. Average values are reported as means ± SD. They were compared with measurements of step and stride frequency and x-, y-, and z-axis movements of the wrist and ankle that have been reported in these same animals in our previous publication (Xiang et al. 2007a). Statistical comparisons of pairs of data were made with a Student's t-test. We also determined the relationship between head angular movements and linear accelerations as well as the motion of the body and the HFP over the stride cycle for each walking speed (Hirasaki et al. 1999; Imai et al. 2001; Moore et al. 1999). At each speed, each parameter was compared across monkeys using a one-way ANOVA. The null hypothesis was that all average values of the parameters were the same, tested using a 95% significance level.


General characteristics of walking

Monkeys walked with either a diagonal sequence with diagonal couplets (diagonal gait) or with a lateral sequence with lateral couplets (lateral gait). The cynomolgus monkeys almost always walked with a diagonal gait, whereas the rhesus walked with either a lateral or diagonal gait. Typical stride frequencies in two rhesus monkeys (Rh002 and Rh652) ranged between 0.4 and 1.5 Hz and were linearly related to walking velocities between 0.45 and 1.45 m/s (Fig. 1, B and C). Four other animals from our previous study (Fig. 2 of Xiang et al. 2007a), which are included in this paper, had mean stride lengths of 39 cm at 0.45 m/s and 56 cm at 0.89 m/s. Swing and stance phase durations were 0.33 and 0.54 s at 0.45 m/s and 0.3 and 0.34 s at 0.89 m/s. Although there were differences in the walking of the individual monkeys, these kinematics are typical for young adult macaque monkeys.

FIG. 2.

Superimposed tracings of the forward-back (x axis) positions of the right forelimb (A), and of head yaw (B), head pitch (C), and head roll (D) of monkey Cy091 walking at 0.89 m/s, all represented in spatial coordinates. The vertical line in A-D shows the end of the swing and the beginning of the stance phases. Positive values, shown by the vertical arrows are to the left (B), down (C), and clockwise (CW) from the animal's point of view (D). E-H: Fourier analysis of data from a sequence when Cy091 walked at 0.89 m/s. E: forward-backward movement of right fore limb (RFL X); F: head yaw; G: head pitch; and H: head roll. The peaks from the power spectrum confirmed that yaw and roll head rotations were related to stride frequency (1.8 Hz), and that pitch was related to the step frequency (3.6 Hz; step frequency = 2 × stride frequency).

Temporal and spectral properties of head and limb movements

Head rotation in yaw (Fig. 2B ), pitch (C), and roll (D) were related to the swing and stance phases of the front paws in Cy091 (A). While walking at 0.89 m/s with a diagonal gait, the head yawed to the left (Fig. 2B), pitched down (C), and rolled clockwise to the right from the animal's point of view (D) during the right forepaw swing (A). The average yaw component had an approximately fixed temporal relation to the swing phase (Fig. 2B), but there was more variation in yaw than in the other components due to spontaneous voluntary yaw head movements that were not synchronized with the gait. The cycle-averaged peak-to-peak amplitudes of the oscillations of the component head movements were ∼10° in yaw, 8° in pitch, and 8° in roll.

There was also a relationship between the frequency of forward (x axis) forelimb movement and head oscillations (Fig. 2, EH). The power spectrum was small >6 Hz and is not shown in Fig. 2. The x-axis forelimb movement had maximal power at 1.8 Hz, the stride frequency in this sequence (Fig. 2E). Head yaw (Fig. 2F) and head roll (H) also had their maximum frequencies at 1.8 Hz. Consistent with the irregularity in the head yaw in Fig. 2B, there was also broad low frequency content in its power spectrum (F). Most of this was due to spontaneous head movement as the animal looked from side to side while walking. The peak activity in head pitch was at the next harmonic, 3.6 Hz, the step frequency (Fig. 2G). In sum, head roll and yaw were most closely associated with stride frequency, and head pitch was most closely associated with step frequency. This was true regardless of whether the animals walked with a lateral or diagonal gait.

Angular velocities and accelerations of the head

Yaw angular velocities and accelerations generally exceeded 50°/s and 500°/s2 (Fig. 3, A and D). Pitch velocities and accelerations were more coherent, reaching >50°/s and 1,000°/s2 (Fig. 3, B and E). There was a consistent difference between the velocity and acceleration of the up- and downward head movements in this animal; the downward velocities and accelerations were larger (Fig. 3, B and E). The up- and downward velocities and accelerations were of the same magnitude in the other animals, however. Velocities and accelerations in torsion were smaller, only reaching ∼25°/s and 400°/s2 (Fig. 3, C and F). These angular accelerations and velocities were more than adequate to activate the aVOR in all dimensions.

FIG. 3.

Head yaw, pitch, and roll velocities (AC), angular accelerations (DF), and head x, y, and z axis translational accelerations (GI) averaged (solid line) from 17 cycles of CY091 walking at 0.89 m/s. The dashed vertical lines show the separation between the swing and stance phases and the solid horizontal lines show 0 velocity and acceleration. Irregularities in yaw velocities (A) and accelerations (D) were due to interposed spontaneous head turns. Head pitch velocities (B) and pitch accelerations (E) were larger in the downward direction. Note that head z acceleration (H) is in synchrony with head pitch velocity (B), and that head yaw velocity (A) lags head y acceleration (G) by ≈90°, the opposite of compensation.

The linear head accelerations were also coherent and reached ≈2 m/s2 along the y axis (Fig. 3G) and 2.5 m/s2 along the x axis (Fig. 3I). The Z-head accelerations, associated with the steps of the right and left forelimbs, were unequal, however, reaching 1.5 and 4 m/s2 (Fig. 3H). This imbalance was not present in other animals.

Head pitch position

For monkey Cy091, there was a consistent difference between the angular velocity and acceleration of the up- and downward pitch head movements; the downward angular velocities and accelerations were larger (Fig. 3, B and E). We questioned whether this asymmetry was related to the head position held by different monkeys in pitch and whether the torque generated by gravity might have contributed to the downward pitch of the head. Head positions were unrelated to walking speed in each of the animals so that an average pitch head position could be computed across all walking velocities (Fig. 4B). The mean pitch position over all animals was 18.4° from the upright (Fig. 4A, solid line). A striking finding was that the heads of the cynomolgus monkeys were pitched down >20° (Fig. 4A, open symbols), whereas the heads of the rhesus monkeys were held closer to the upright (Fig. 4A, filled symbols). The mean cynomolgus head position over all walking velocities was 30.4°, which was significantly larger than the mean rhesus head position, 7.2° (P < 0.0001). The largest average head pitch was ≈44° in Cy091 (Fig. 4A, dashed line), which probably accounted for the asymmetry in its upward and downward head pitch velocities and accelerations (Fig. 3, B, E) and head-z accelerations (Fig. 3H).

FIG. 4.

A: average head pitch positions of 7 monkeys during walking at treadmill velocities from 0.4 to 0.98 m/s. Zero on the ordinate is upright and 90° is prone. Open symbols are from cynomolgus and filled symbols from rhesus. The dashed line at 44.4° shows the average head position for Cy091. It's data are shown in Figs. 2 and 3. The heavy horizontal line shows the average head position of all monkeys in pitch (18.4°), which was independent of walking velocity. B: mean head pitch positions of the 7 monkeys. The rhesus monkeys had the most upright head positions and the cynomolgus the most prone head positions.

Translation and rotations of the head as a function of walking velocity

We next considered whether head angular and translational movements were a function of walking velocity. One animal (Cy101) had an idiosyncratic walking style with significantly larger head movements than the other six monkeys; it will be considered separately. y-axis translation and yaw movements decreased slightly as the animals walked faster (Fig. 5, A and D), and there was a modest linear relationship for head yaw (R2 = 0.28). A reduction in spontaneous yaw head movements at higher walking velocities could have produced the relationships shown in Fig. 5, A and D, but this could only be determined in trained animals, which was beyond the scope of this study. Oscillations along the z (B) and x axes (C) and around the pitch (E) and roll (F) axes, however, had no relation to walking velocity (R2 < 0.1). Thus there was no effect of walking velocity (or stride frequency and stride length) on head z- or x-axis translations (Fig. 5, B and C) or on head pitch or roll (E and F) while walking on the treadmill.

FIG. 5.

Range of peak-to-peak translational (AC) and angular (DF) head movements that were cycle-averaged on stride from 6 animals for walking velocities between 0.4 and 0.98 m/s. Head y axis (A) and yaw movements (D) tended to be smaller as the animals walked faster (R2 = 0.28), but mean translational z (B) and x axis movements (C) and angular pitch (E) and roll movements (F) were the same over all walking velocities. This was true for both cynomolgus (open symbols) and rhesus monkeys (filled symbols).

The ranges of the mean angular and linear head movements across all speeds in both species are shown in Table 1. Values for the group of six monkeys are shown top, and those for Cy101 are shown at the bottom. For the six monkeys, head angular peak-to-peak amplitudes generally ranged from 3 to 10° (except for head yaw in Rh002) and head translation from 1 to 5 cm. The largest averaged head movements were in yaw (9.1 ± 3.9°, n = 48, all walking velocities) and the smallest were in roll (5.1 ± 2.2°), with pitch head movements falling in between (6.4 ± 1.9°). Translations of the head were all small. Lateral (y axis) head movements were the largest (4.1 ± 1.4 cm), forward/backward (x axis) head movements were smaller (2.7 ± 1.3 cm), and vertical (z axis) head movements were the smallest (1.8 ± 1.1 cm). Walking of individual animals was not exactly the same as reflected in an ANOVA analysis, in which only the yaw movements did not reach statistical significance. However, the averaged head movements from six monkeys across two species would appear to be typical for head stabilization in most monkeys while walking on a flat surface.

View this table:

Head rotations and translations in space

Monkey Cy101 walked with a “swimming motion,” widely extending its forelimbs (Xiang et al. 2007a). Its stride length was longer, and the stride frequency was less. The amplitude of the translations and rotations of the head were all significantly larger in this animal. These differences denote the fact that there are many possible styles of locomotion, and any general statements characterizing this complex phenomenon are open to exception.

Body movements

Because of the relative stability of the head in space during locomotion, it was of interest to determine how well this stability was maintained relative to the body as a measure of the function of the aVCR. The movements of the chest are primarily dependent on the stepping cycles of the forelimbs. The hindlimbs contribute little to movement of the upper body or the head and neck although the fore- and hindlimbs move synchronously during both diagonal and lateral gaits. For example, the body translates to the left along the y axis, yaws to the left, and rolls counter-clockwise to the left (from the animal's point of view) in association with the stride frequency, i.e., with the swing phases of the right forelimb and the stance phases of the left forelimb, and vice versa. There are also two pitches of the body during each stride cycle, related to the up- and downward movement of the forepaws during the swing phases.

We first considered whether there was any relationship between the parameters of gait and body yaw, pitch, and roll. The magnitude of the yaw and pitch rotational body movements were less than or close to 10°, and there was little or no relationship to the stride length (Fig. 6, A and B). Body roll, on the other hand was larger, ranging from ≈9 to 20°, and increased when the animals walked faster and made longer strides (Fig. 6C). Thus the need for stabilization of the head would be greater in roll than in yaw or pitch.

FIG. 6.

Peak-to-peak body yaw (A), pitch (B), and roll rotations (C) as a function of stride length from the forelimb (○) and hind limb (•) in Rh652 and Rh002. There was no significant association of body yaw (A) or body pitch (B) with stride length, but body roll (C) got larger as stride length increased.

Head movements in relation to body movements

Typical head on body movements over three stride lengths are shown in Fig. 7A for Rh652, walking at 0.98 m/s with a diagonal gait. The head was stabilized in yaw at the onset of this sequence (A), but the animal turned its head to the right [negative (down) according to the right-hand rule] toward the end of the sequence. Oscillations of the body in yaw (dashed lines) were larger and were linked to the stride frequency. Oscillations of the head and body in pitch were at a higher frequency, related to the step cycle (Fig. 7B). The animal moved its head down ∼20° as the sequence progressed. The roll movements of the body were related to the stride cycle and were large, approaching 20°. The head was well stabilized in roll throughout the sequence, however (Fig. 7C).

FIG. 7.

Head-in-space (solid lines) and body-in-space (dashed lines) rotations (AC and GI) and translations (DF) of Rh652 walking at 0.98 m/s with a diagonal gait. Data were represented with a 0 mean. AC: head and body yaw (A), pitch (B), and roll (C) over 3 stride cycles. A: the head was stabilized in yaw at the beginning of the sequence but then moved to the right. B: the head and body pitched at the step frequency (twice the stride frequency). C: there were large body rotations in roll, but the head was well stabilized in space over the 3 cycles. DF: cycle-averaged head and body translations. D: the head and body moved in opposite directions along the y axis and in the same directions along the z (E) and x axes (F). Dotted line, head rotation relative to the body; dashed line, body movement in space; solid line, head movement in space. There was compensation for body movement so that the head was well stabilized in yaw (G) and roll (I). There was also some reduction in head pitch in space (H).

Rotations and translations of the head and body were compared in cycle-averaged data for Rh652 (Fig. 7, DI). Each panel shows the duration of one stride cycle. When monkeys walked with a diagonal gait, the head and body translated laterally along the y axis in opposite directions at the stride cycle frequency (Fig. 7D). When monkeys walked with a lateral gait, the head and body translated together (not shown). The head and body also translated vertically together, along the z axis (Fig. 7E), and forward (F), along the x axis at the step cycle. Angular movements were determined for the body in space (Fig. 7, GI, dashed lines), for the head relative to the body (dotted lines), and for the head in space (solid lines). The head was well stabilized in space in yaw and roll over these cycles (Fig. 7, G and I), and there was less stabilization in pitch (Fig. 7H). As noted previously, it was generally difficult to determine stabilization in yaw because of the many voluntary yaw head movements.

Compensation for body movement

To determine the compensation produced for body movements of different amplitudes, peak-to-peak head rotations in space were compared with body rotation in space (Fig. 8, AC). The diagonal lines separate data points; the top left half represents head rotations that were larger than body rotations, whereas the bottom right half indicates head rotations that were smaller than body rotations. Body yaw and pitch were generally <10°, whereas body roll ranged from 10 to 28°. Head yaw was almost always larger than body yaw in this sample, ranging ≤25° (Fig. 8A). This was likely due to the spontaneous voluntary lateral head and gaze shifts in the horizontal plane. Head pitch (Fig. 8B), on the other hand, was usually equal to or smaller than body pitch, whereas head roll was always <10° and always smaller than body roll (Fig. 8C). Thus compensation for body movement, although present in all angular directions including yaw (see Fig. 7A), was greatest for body roll.

FIG. 8.

AC: comparison of rotational movements of the head in space (ordinates) and body in space (abscissae) of Rh652 and Rh002 represented as peak-to-peak values. The dotted line in each graph separates ratios in which head rotation was greater than body rotation (>1, above) and when head rotation was smaller than body rotation (<1, below). A: head yaw was generally larger than body yaw in this sample, while head pitch (B) and roll (C) were smaller than the pitch and roll of the body. As body roll became larger (C), head roll movements continued to be small. DI: gains (top) and phases (bottom) for head yaw (D), pitch (E), and roll (F) computed as (head re body velocity)/(body in space velocity). Horizontal lines show unity gain in the top graphs and values that would be 180° out of phase in the bottom graphs. There were no coherent gains or phases for yaw (D), but pitch (E) and roll (F) both had values that were close to 180° out of phase for head and body rotations, with gains close to unity.

Based on the assumption that the reduction in head movement in space were due to the compensatory action of the aVCR, we calculated the gains and phases between the angular head re body velocity and the angular body re space velocity (Fig. 8, DF). Consistent with the idea that most of the yaw head movements were spontaneous and voluntary, gains were 2.1 ± 1.4, and there was no fixed phase relationship for yaw (Fig. 8D). The yaw head phases ranged from 0 to 200°, averaging 121 ± 62°. Gains in pitch had a narrower range and averaged 1.0 ± 0.4 (Fig. 8E) with phases that varied around 193 ± 32° (Fig. 8E). Gains also were close to unity for roll (1.1 ± 0.2), and the phases were 168 ± 12° (Fig. 8F). Thus in this sample from two rhesus monkeys (Rh002 and Rh652) the gains and phases of the pitch and roll movements were consistent with compensatory action of the aVCR. Because of the many spontaneous voluntary yaw head movements, the yaw gains and phases of compensation could not be determined.

The data from the two rhesus monkeys were lumped to obtain average values (Table 2). The combined data were consistent with the previous differences between head and body angular movements shown in Fig. 8. There was a significant reduction in head movements in pitch and roll compared with the body pitch and roll (P < 0.001, paired t-test). Yaw head movements were significantly larger than body movements in these animals. Overall, both head and body translations were small (ranging from 1.4 to 4.1 cm peak to peak), and there was no consistent pattern in head and body translation between the two animals. The only difference in translation was along the y axis (P < 0.001), where body translation was 1.6 cm smaller than head translation. Otherwise, the movements of the head and body were essentially the same along the x (P = 0.448) and z axes (P = 0.559). Thus there is strong stabilization of the head in roll, and some stabilization in pitch and yaw in space compensating for body rotations and translations due to the quadrupedal stepping.

View this table:

Head and body rotations and translations in space

Pitch, yaw, and head Z and Y translation

We also determined whether the lVCR had contributed to compensation in pitch and yaw during locomotion to counter-rotate the head and maintain the direction of the naso-occipital axis at a point, which has been termed the HFP (Hirasaki and Kumakura 2004; Hirasaki et al. 1999; Moore et al. 1999; Pozzo et al. 1991a,b; Raphan et al. 2001). To meet this requirement, head accelerations and velocities should be sufficient to activate the lVCR. Additionally, had the lVCR been active, appropriate phase relations should maintain the HFP in front of the animal, and linear head accelerations should lead angular head velocities by ≈90° in the pitch plane, and lag behind them by ≈90° in the yaw plane.

The relationship between linear head accelerations and angular head velocities was determined first. Head z-axis acceleration ranged from 0.5 to 5 m/s2 and increased as a function of walking velocity, varying widely among the six monkeys (Fig. 9A). Head pitch velocities also increased with the increases in walking speed, ranging from 18 to 100°/s (Fig. 9B). Lateral acceleration of the head along the y axis ranged from 0.5 to 5 m/s2 (Fig. 9D), falling into the same range as the head z-axis acceleration (Fig. 9A). Head yaw velocities (Fig. 9E) ranged from 10 to 120°/s and did not increase with walking velocity (Fig. 9E). There were also positive relationships between head Z acceleration and head pitch velocity (Fig. 9C) and a weaker relationship between head Y acceleration and head yaw velocity (F). Thus as shown in Fig. 9, A and D, the magnitude of the Z and Y accelerations generated during walking would be sufficient for activation of the lVCR. The linear accelerations and angular velocities were larger in the vertical and horizontal planes for the cynomolgus monkeys (○, ◊) than for the rhesus monkeys (•, ⧫, ▪, *) in this group of animals.

FIG. 9.

Ranges of head z (A) and y axis (D) accelerations and head pitch (B) and head yaw velocities (E) in space as a function of walking velocity. - - -, the linear fits. Accelerations and velocities of 2 cynomolgus monkeys (○, ◊) were generally higher than those of the 4 rhesus monkeys (•, ⧫, ▪, *). C and F: relationships between head pitch velocity and head z acceleration (C) and between head yaw velocity and head y acceleration (F). The velocities and accelerations increased together (- - -).

We next determined whether there were appropriate phase relations between the head z-axis acceleration and head pitch, which would signify activation of the lVCR. Most commonly the head moved up and pitched up, which would put the HFP behind the head (Fig. 10A, gray bar). Occasionally, however, there were brief periods in which the head z-axis movement and the head pitch opposed each other (Fig. 10B, gray bar), which would give a fixation point in front of the animal. Cy091 and Rh652 had a HFP in back of the head for ∼50% of time, and the other animals had the HFP in back of the head most of the time. Thus there was no consistent HFP in pitch in front of the animals over the experiments, although each monkey could have brief periods during which the HFP was in front of the head.

FIG. 10.

A and B: individual stride cycles of the right forelimb (RFL X), with related head z translations and head pitch amplitudes as a function of time. A: in the shaded area, the head of Cy101 pitched up as it translated up, putting the position of the head fixation point (HFP) in back of the head. B: the shaded area shows an instance when the head pitch down in Cy091 when it was moving up. This would put the HFP in front of the head. C: phase relationship between head z axis acceleration ( ↑, - - -) and head pitch velocity (↓, —). There was no significant difference between the phases. The head z acceleration was small in the 1st cycle (↑), although the head pitch velocity was unchanged. D: phase relationship between head y acceleration (- - -) and head yaw velocity (—). Y axis acceleration led head yaw velocity by ∼90°.

The phase relations were also not consistent with this postulate. In Cy091, the head pitch velocity (Fig. 10C, —) did not lag appropriately for it to have been generated by the head Z acceleration (Fig. 10C, - - -). Furthermore, the head pitch velocity in Cy091 had two identical peaks during the steps in each stride cycle, whereas the head Z acceleration was imbalanced in this animal at the time of this recording, due to a limp. If the lVCR was responding to the stimulus from the head Z acceleration, it should not have generated the same amount of head pitch. Another animal (Cy115) had only small head movements along the z axis (0.4 cm), and there was no sinusoidal variation in head position, making it impossible to adequately characterize the z-axis head movement. Despite this, the head pitch amplitudes were similar in this as in the other six animals (Table 1). Therefore there was no strong evidence for activation of the lVCR in pitch. Similarly, if the lVCR was activated to establish a HFP in the yaw plane, the head yaw velocity should have led the head Y acceleration by 90°. Instead, in Cy091, the head yaw velocity (Fig. 10D, —) lagged the head Y acceleration (- - -) by ≈90°. Thus neither was there evidence for activation of the horizontal lVCR in this animal.


The main result of this study is that the head is held relatively stable in three dimensions in space over a wide range of walking velocities during active quadrupedal locomotion of rhesus and cynomolgus monkeys on a treadmill. On average, pitch and roll head movements were less than the corresponding component of body rotation with gains close to unity and phases close to 180°, and translations along the x, y, and z axes were ≤4 cm. Presumably, the angular stabilization of the head was due to compensatory action of the aVCR, responding to the head movement in space generated by body movement in space. Because of spontaneous shifts of the head in yaw, we were unable to determine the gains and phases of yaw compensation, but there were clear examples of stabilization of the head in yaw, as well (e.g., Fig. 7A). Thus while humans had a resonant peak for head rotation when the body was passively rotated in darkness (Keshner 2003), alert monkeys maintained their head position in space during locomotion over a wide range of frequencies. The aVOR could then adequately stabilize gaze for the residual angular head velocities and accelerations (Fig. 3) at all walking velocities.

The magnitude of the head stabilization was greater in roll than in pitch. This is consistent with previous studies in decerebrate and anesthetized animals, which have shown that vestibular-induced neck reflexes are more active in roll than in pitch (Lindsay et al. 1976; Magnus 1924; McCouch et al. 1951; Schor and Miller 1981; Schor et al. 1986; Wilson et al. 1986). The functional reasons why the gain of the aVCR might be larger in roll than in pitch could be related to the lower gain of the aVOR for roll and to the need to maintain gaze aligned with the spatial horizontal during a forward heading. The gain of the roll aVOR is maximally ≈0.6 in both monkeys and humans (Crawford and Vilis 1991; Henn et al. 1992), which is considerably less than the close-to-unity gains of the yaw and pitch aVOR. Therefore the aVOR can easily compensate for disparity between head and eye movement in the vertical and horizontal directions but not in roll. Additionally, ocular pursuit has unity gains and compensatory phases in monkeys and humans at frequencies ≤1 Hz in the cardinal directions, whereas ocular pursuit is virtually nonexistent in roll (Ghasia and Angelaki 2005). Otolith-ocular orienting reflexes also have small gains in roll in monkeys and humans, there being only ≈4° of ocular counter-roll for 60–90° of head tilt (see (Cohen et al. 2001) for review). Thus the head needs to be better stabilized in space in roll than in pitch and/or yaw to help stabilize gaze. Such stabilization of the head could also be useful for orienting the body in space relative to the trajectory when moving around corners (Hicheur et al. 2005; Imai et al. 2001), which is probably produced largely through activation of the aVCR. Supporting this, the majority of vestibulocollic neurons encountered in single-unit studies of the vestibular nuclei were excited by side-down roll, and neurons responding to head pitch were less common (Schor 1974; Schor et al. 1984, 1985). Thus the neurophysiological basis exists for the generation of larger compensatory head re body movements in roll than in pitch.

What can be implied about the organization of the descending signals and the pattern of cervical muscle activations that generate the aVCR during quadrupedal locomotion? When the semicircular canals are stimulated, the direction of head and body movement is always in the summed plane of the activated canals, regardless of position of the head on the neck (Suzuki and Cohen 1964). Therefore in response to the accelerations of the head produced by body movement, a signal pattern must emanate from the semicircular canals that is transformed in the cervical spinal cord to generate the compensatory head rotation (Shinoda et al. 1993, 1994, 1997; Sugiuchi et al. 1992; Wilson et al. 1995). During quadrupedal locomotion, the cervical column was probably held close to the upright as it is in most mammalian species, with an additional inflection point at C7, the cervico-throracic junction (Graf et al. 1995a,b; Vidal et al. 1986, 1988). With the head z axis upright and close to the spatial vertical, as in the rhesus monkey, head yaw in space would predominantly come from the rotation of C1 on the dens (C2), whereas pitch in space would be due to rotation of the occipital condyles on C1 with an additional contribution of the cervical vertebrae during locomotion (Kunin et al. 2007). With the head pitched forward toward the spatial horizontal, however, as in the cynomolgus monkeys, compensatory yaw head movements in space must have involved a mixture of yaw and roll rotations relative to the body, and the reverse would be true for head roll in space. Therefore the exact muscle groups that would be engaged by each yaw, pitch, or roll rotation would vary, depending on how the head was positioned on the neck.

Frequency characteristics of head stabilization during locomotion and its relation to the aVCR

Controlled studies of human head stabilization, when the body was rotated at frequencies of 0.1–1 Hz with mental arithmetic in darkness showed that the gain of head rotation re body rotation was small, <0.15 (Guitton et al. 1986). The conclusion was that the contribution of the CCR and aVCR to head stabilization in space were negligible. Considering the physiological findings of Schor et al. (Schor and Miller 1981; Schor et al. 1986, 1988), it is likely that the aVCR is active across a wide frequency range, but that the gain is low at frequencies <1 Hz (Peng et al. 1996, 1999). It is possible that the lower gains were due to an interaction of the cervico-collic reflex (CCR) and the aVCR at the lower frequencies. The CCR would tend to hold the head stable relative to the body, thereby reducing the apparent gain of the aVCR. Regardless, 1 Hz was below the frequency range of locomotion in the current study.

The most relevant data for our findings are from humans whose bodies were firmly fixed while they were rotated in pitch (Keshner et al. 1995) or yaw (Keshner and Peterson 1995) with a sum of sines. The paradigm where the subjects were tested in darkness with mental arithmetic probably was the closest to singling out action of the aVCR. In that condition, the gain of compensatory head movements was low at low frequencies (<1 Hz) but peaked for rotations between 1 and 2 Hz, regardless of the rotation plane. Above that range, however, there were differences between yaw and pitch. In yaw, the gains increased to well above 1.0, indicating strong mechanical resonance, whereas in pitch the gains dropped above 2 Hz with significant phase scatter. Even with vision, there was a sharp fall in gain in pitch for frequencies above 1–2 Hz and a drop in yaw gains between 1 and 2 Hz. Phases above 2 Hz were scattered and none were close to 180°, the appropriate compensatory separation between stimulus and response for the aVCR. Thus in all of the conditions they tested, passive rotation and restriction of the body during rotation was not associated with appropriate gains and phases for head compensation at higher frequencies of rotation. In contrast, the gains and phases in pitch were maintained across all walking velocities with stepping frequencies >3 Hz in our animals. (Yaw was not characterized, as previously noted.) Our results also demonstrate that the aVCR can be active in two directions, whereas it is inactive in a third direction. That is, although the animals made many spontaneous movements in yaw while walking, compensatory roll and pitch continued, despite these voluntary movements.

We considered whether inertia and viscoelastic properties of the head neck system were predominantly responsible for stabilizing the head or whether head stabilization was mainly due to the aVCR. The head of the monkey has a smaller mass (≈1 kg), about five times less than in human (Tinniswood and Gandhi 1999), and its radius (≈5 cm) is less than half of that of the human head. Therefore the moment of inertia is considerably less in the monkey than in the human. As such, inertial effects in conjunction with viscoelastic properties would tend to shift the resonance to higher frequencies. These effects would also be expected to be altered at different head positions. Neither of these effects was observed. Therefore we concluded that it was the aVCR that was stabilizing the head in space. In addition, the smaller mass of the monkey's head would have contributed to a higher frequency of resonance and a broader tuning of the head stabilization system. This would explain why the head was still stabilized at 3 Hz in our animals. Potential limitations of these findings were that a comparison of head and body movements were done only in two rhesus monkeys, the monkeys were walking on a treadmill without visual streaming, and they were not trained to watch a specific target. As a result, it was not possible to separate action of the aVCR in yaw from spontaneous head movements in these animals.

Vision is a critical factor in maintaining balance during locomotion, and the addition of vision improved the gains and phases at low frequencies in alert humans whose bodies were restrained and rotated (Guitton et al. 1986; Keshner and Peterson 1995; Keshner et al. 1995). It is not known whether the absence of visual flow might have altered the mechanics of walking or whether this actually enhanced the action of the vestibular system. However, the spectral content of head movements during locomotion (1–4 Hz) ranged above that of visual response characteristics. Therefore it is unlikely that vision contributed significantly to the animals' ability to maintain gain at higher frequencies.

We also questioned whether the base of support altered the characteristics of gait during treadmill walking in comparison to natural overground quadrupedal locomotion. Dunbar and associates (Dunbar 2004; Dunbar et al. 2004) studied monkeys walking naturally and on a treadmill. They concluded: “head and trunk rotational patterns during treadmill walks were comparable to the patterns found during overground walks. The rotational velocities of these segments during both treadmill walks and gallops were also comparable to the velocities found during natural locomotion.” Therefore base of support is unlikely to have dramatically changed the dynamic characteristics of gait.

Head and body posture and species independent strategies of locomotion

Several species differences emerged related to head orientation during locomotion. The cynomolgus monkeys usually walked with a diagonal gait holding their heads pitched down ∼30°, whereas the rhesus monkeys more frequently used a lateral gait (Xiang et al. 2007a) with their heads held upright. The head and body translated laterally together in the rhesus monkeys walking with a lateral gait, whereas the head moved against the body during the reciprocal gait. It has been postulated that a diagonal gait would be more compatible for monkeys walking on narrow runways, such as branches in forests because their weight would not shift from side to side (Hanna et al. 2006; Schmitt 1999, 2003). The body is also held lower during diagonal than during lateral gait sequences, the limbs are extended further to enable larger gait cycles, and more weight is placed on the hindlimbs than the forelimbs (Hanna et al. 2006; Reynolds 1985; Schmidt 2005; Schmitt and Lamelin 2004; Shapiro and Raichlen 2005; Xiang et al. 2007a). These differences may be related to the different demands placed on the animals by their native habitats and behaviors (Cawthon Lang 2005, 2006).

Despite these differences, there was no difference between species in head angular deviations during locomotion, which ranged from 3 to 10° and head translation ranging from 1 to 4 cm. For the two rhesus monkeys that were tested for head on body movements, the magnitudes of the yaw and pitch angular body movements were generally less than or close to 10°, and there was little or no relationship to stride length. Body roll, on the other hand, increased when the animals made longer strides, emphasizing the dependence of body and head movements on the swing and stance phases of the forelimbs. Presumably, the swing and stance phases alternately rotate the chest through the shoulders as well as the attachment of the scapula to the posterior chest wall. These chest movements are then transmitted to the head. There was a consistent difference between the velocity and acceleration of the up- and downward head movements in one monkey that held its head pitched forward 44° from the upright position. The downward velocities and accelerations were larger in this animal probably due to the contribution of a larger gravitational torque to the downward pitch of the head. There was no significant difference in the upward and downward pitch in the other animals.

The largest average head movements across all walking velocities were in yaw (9.1°) and the smallest were in roll (5.1°) with pitch head movements falling in between (6.4°). Translations of the head were all small. Lateral (y axis) head movements were the largest (4.1 cm), forward/backward (x axis) head movements smaller (2.7 cm), and vertical (z axis) head movements were the smallest (1.8 cm). The x-axis translations would have been much larger, of course; had the animals been walking overground, but the y- and z-axis movements would have been essentially the same. A striking finding was that walking velocity (or stride frequency) had no effect on the amplitude of head movement with the exception that there was a small decrease in the size of yaw head movements as walking velocity increased and an increase in body roll at larger stride lengths. The finding that head compensation did not change when body roll became larger at higher walking velocities, i.e., at larger stride lengths, indicates that the aVCR was particular efficient in compensating for body roll. Consistent with the similarity in the amplitude of the sinusoidal head and body movements at all walking velocities was the finding that the angular velocities and linear accelerations rose as the walking velocity increased, a finding that would occur in quasi-sinusoidal movements as the frequency increased with increases in walking velocity.

Working in natural circumstances and on a linear treadmill, Dunbar and colleagues (Dunbar 2004; Dunbar et al. 2004) studied quadrupedal locomotion in freely moving hanuman langur and bonnet macaques, all of which have long tails that are similar to those of cynomolgus monkeys. Dunbar et al. recorded 10 walk and 10 gallop cycles with a movie camera at 100 fps. Analysis was limited to head and body movements along the z axis and in the pitch plane. On average, the head pitched 15° in space relative to 6° in pitch of the trunk in space. Their trunk-in-space movements were similar to those in this study, but the pitch head movements in six of our monkeys were considerably smaller, being ≈5–7°. However, the head movements of a seventh cynomolgus (Cy101), were substantially larger, also ≈16°. The pitch movements of their animals were ∼60°/s, which was close to the head pitch velocities of ≈50–60°/s of the monkeys in our series. The frequency of pitch movements in Dunbar et al.'s animals was ≈1–2 Hz, which is less than that of the pitch movements in our series, but this would be dependent on the size of the animals and their step frequencies. Thus there is good agreement between the two studies for movements in the pitch plane, considering the fact that there can be variability between and among animals of the same and different species.

lVCR and head stabilization during locomotion

There was no evidence for activation of the lVCR, although there were clearly instances where the animals held the correct compensatory position, i.e., with the HFP in front. Presumably, instances of compensation were due to the strategy of the animal to maintain a given fixation point so that the head maintained its pointing direction when the head was perturbed by the stepping. Strategic compensatory head pointing was not found when the body was passively translated (Wei and Angelaki 2004), and it is likely that the lVCR is activated with a substantial gain only when there is a strategy to maintain gaze at a particular point of fixation coupled with an active movement state such as locomotion (Raphan et al. 2001; Wei and Angelaki 2004). Similar active fixation points are necessary to elicit the functioning of the lVOR (Paige and Tomko 1991; Schwarz and Miles 1991).

Action of the lVCR was originally postulated in humans during bipedal locomotion from the tendency for the naso-occipital (x) axis of the head to point toward a fixed location in space in front of subjects (HFP) while they were walking (Berthoz and Pozzo 1994) (Hirasaki et al. 1999; Imai et al. 2001; Moore et al. 1999; Pozzo et al. 1990). Hirasaki has recently found head pitch of similar amplitude in bonobos monkeys and gibbons is associated with a HFP in front of the animals during bipedal and quadrupedal locomotion (Hirasaki and Kumakura 2004). The frequency (2.5–3.8 Hz) and amplitude of the vertical linear accelerations of the trunk and head (≈0.3 g) at these frequencies would have been sufficient to activate the lVOR. However, a HFP was not present in the limited sample of data from Cercopithecus monkeys walking quadrupedally (Dunbar 2004) and occurred only intermittently in our own data. One explanation for this is that the animals were not required to watch a target in front of them. By analogy with the lVOR, which is not apparent unless activated by a near fixation point, the lVCR may also be dependent on the involvement of vision and on the presence of a near target.

Similarly there was generally no fixed relationship between head yaw and lateral (Y) translation of the head. To maintain a consistent head fixation point in the horizontal plane based on our coordinate definition (Fig. 1), head yaw velocity should lead head Y acceleration by 90°. However, the opposite was the case, i.e., the head Y acceleration (Fig. 10D, - - -) led the head yaw velocity by ≈90° (Fig. 10D, —), which would maintain the horizontal HFP in the back of the head. The cynomolgus monkeys, engaging in diagonal gait sequences, mostly had this characteristic. The rhesus monkeys, walking with lateral gait sequences, however, were more likely to generate a horizontal HFP in the front, although this was never consistent. Therefore it was unlikely that the lVCR was consistently activated in our experiments.

In conclusion, the major contribution of this study was to show that the gains and phases of head-on-body compensation, which we have attributed to the aVCR with a possible contribution of vision, were different from those determined in anesthetized or decerebrate animals or in alert humans or animals that were passively rotated with their trunks stabilized. During quadrupedal walking, limb movements generate trunk and head oscillations in yaw and roll and along the y axis at ≈1–2 Hz, and pitch and translations along the x and z axes of ≈2–4 Hz, a range that has previously been identified as the frequencies that would activate the aVCR (Keshner and Peterson 1995; Keshner et al. 1995; Peng et al. 1996; Peterson 1988; Peterson et al. 1988). In their studies, however, there was a reduction in gain in light and a dispersion of phase for rotations >1.5–2 Hz. In our study, the gains of the head-on-body movements were close to unity across the full range of natural walking velocities in pitch and roll. The phases were also close to 180° across this range. From this, we conclude that natural movement of the trunk is essential to maintain full activation of the aVCR in three dimensions. Thus the aVCR is no different from the aVOR in that the enhancement provided by top-down and bottom-up activation, i.e., by behavioral state, has a profound effect in enabling proper function of head stabilization reflexes.


This work was supported by National Institutes of Health Grants EY-11812, EY-04148, DC-05204, and EY-01867.


We thank J. Martinez, P. John, D. Ogorodnikov, and S. Tarasenko for technical assistance.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract