Kinematic and kinetic adjustments of Japanese macaques during quadrupedal walking on horizontal poles

TL;DR

Japanese macaques can walk stably on narrow poles without major changes, adjusting forces to stay balanced, showing how surface shape affects primate movement.

Contribution

This study reveals how Japanese macaques adjust gait mechanics on different pole widths, highlighting substrate geometry's role in primate locomotion.

Findings

Walking speed and limb flexion remained stable across pole diameters, indicating consistent locomotion.

Macaques reduced mediolateral forces and hindlimb braking on poles compared to flat surfaces.

Mechanical energy recovery via pendular exchange was similar on poles and flat surfaces.

Abstract

Primates employ a diagonal-sequence gait, exhibit compliant forelimbs and place greater weight on the hindlimbs, features regarded as adaptations to arboreal locomotion. Although substrate diameter is an ecologically critical factor influencing gait mechanics, its effects remain underexplored in larger catarrhine monkeys. This exploratory study examined how horizontal pole diameter affects the kinematics and kinetics of quadrupedal walking in Japanese macaques (Macaca fuscata). Four trained macaques walked on metal poles of three diameters (48.6, 76.3 and 114.3 mm) mounted on force plates while ground reaction forces, joint kinematics, center of mass (COM) dynamics and mechanical energy exchange were analyzed. Contrary to predictions, walking speed, stance duration, mediolateral COM fluctuations and limb flexion did not differ significantly between narrow and wide poles, indicating stable locomotion across tested diameters. The only detectable differences were a modest lowering of COM and a modest decrease of stride length on narrower poles, which was probably of negligible functional significance. In contrast, comparisons with flat-surface walking revealed substrate-specific adjustments: macaques minimized mediolateral forces, showed reduced hindlimb braking impulses and lifted the hindlimb earlier on poles. Despite these differences, mechanical energy recovery via pendular exchange remained similar (35–50%) between pole and flat-surface walking. These results demonstrate that Japanese macaques maintain stable quadrupedal locomotion on moderately narrow poles without major postural changes, while modulating mediolateral force production to reduce destabilizing moments in the frontal plane. Substrate geometry therefore plays a key role in shaping primate gait mechanics and offers insights into the evolutionary adaptations underlying arboreal locomotion.

Summary: Japanese macaques maintain stable gait during quadrupedal locomotion on moderately narrow poles, with reduced mediolateral forces, indicating the role of substrate geometry in primate gait mechanics for arboreal locomotion.

Figures

Experimental setup for measuring the kinematics and kinetics of quadrupedal walking on a horizontal pole in Japanese macaques. The macaques walked quadrupedally at a self-selected speed on horizontal poles of three different diameters (48.6, 76.3 and 114.3 mm; length ∼5 m) mounted on two force plates and were filmed using four digital cameras. The rectangular aluminium frames were covered with cloth, which the monkeys disliked stepping on, obliging them to walk on the poles instead.

Digitized marker positions and definitions of joint angles. Top: positions shown correspond to: (1) head of the fifth metatarsal (MT), (2) lateral malleolus of the fibula, (3) lateral epicondyle of the femur, (4) greater trochanter, (5) acromion, (6) lateral epicondyle of the humerus, (7) styloid process of the ulna and (8) head of the fifth metacarpal (5MC). COM, center of mass. Bottom: an example of digitization using DeepLabCut.

Three components of mean ground reaction force (GRF) in the ipsilateral forelimbs and hindlimbs during quadrupedal walking on a narrow and wide horizontal pole. Lines and shading represent mean±s.d. GRF in the anteroposterior (A), mediolateral (B) and vertical (C) direction. BW, body weight. The bars at the bottom indicate the average foot–ground contact duration of the forelimbs (F) and hindlimbs (H).

Mean joint angle profile during quadrupedal walking on a narrow and wide horizontal pole. Lines and shading represent mean±s.d. ankle (A), knee (B), hip (C), wrist (D), elbow (E) and shoulder (F) angle. The bars at the bottom indicate the average foot–ground contact duration of the forelimb (F) and hindlimb (H).

Comparison of normalized mediolateral fluctuation and COM height between the narrow and wide horizontal poles. Box plots (median, upper and lower quartiles and 1.5× interquartile range) show mediolateral fluctuation (A) and COM height (B) for all four macaques (symbols indicate individual data points).

Comparison of the three components of the GRF during walking on a narrow horizontal pole and a flat surface. Lines and shading represent mean±s.d. GRF in the anteroposterior (A), mediolateral (B) and vertical (C) direction.

Comparison of mean footfall patterns during quadrupedal walking on a horizontal pole and a flat surface. The bars indicate the contact times of the four limbs for horizontal pole (A) and flat-surface (B) walking. F, forelimb; iHL, ipsilateral hindlimb; cFL, contralateral forelimb; cHL, contralateral hindlimb. Note that two-point contact of the ipsilateral forelimb and hindlimb (asterisks) occurred on the horizontal pole, whereas this period was very short on the flat surface.

Fluctuation in potential energy (PE), kinetic energy (KE) and total external mechanical energy (TME) of the COM during quadrupedal walking on a narrow horizontal pole.

Tables

Table 1.. Japanese macaques used in the present study

Subject	Age (years)	Sex	Body mass (kg)	Effective hindlimb length (m)
Kurimatsu	9	Male	8.25	0.356
Shiotaro	8	Male	6.87	0.328
Matakichi	9	Male	6.49	0.310
Maro	4	Male	4.55	0.294

Table 2.. Inertial parameters of limb segments in the Japanese macaque

	Segment mass fraction (%)	COM location (%)
Trunk	70.0	52
Upper arm	3.2	46
Forearm	2.0	41
Hand	0.6	71
Thigh	5.6	41
Shank	2.7	40
Foot	1.0	62

Table 3.. Gait parameters of quadrupedal walking on a horizontal pole in Japanese macaques

Subject	Pole	n	Stride length (m)	Stride duration (s)	Velocity (m s−1)	Fr	Phase (%)
							A	B	C	D
Kurimatsu	Narrow	10	0.88±0.05	0.88±0.05	1.01±0.11	0.30±0.07	33.6±2.0	50.6±2.0	61.6±2.8	92.3±3.4
	Wide	8	0.96±0.04	0.86±0.09	1.13±0.15	0.37±0.10	31.5±1.9	49.5±1.5	57.9±2.4	89.8±2.3
Shiotaro	Narrow	8	0.82±0.03	0.93±0.07	0.89±0.09	0.25±0.05	32.6±2.9	49.2±2.8	58.6±2.1	92.7±2.3
	Wide	7	0.88±0.04	0.88±0.07	1.00±0.11	0.31±0.07	32.7±2.1	51.8±2.2	60.1±2.6	91.0±2.0
Matakichi	Narrow	4	0.77±0.05	0.78±0.07	1.01±0.16	0.34±0.10	33.7±3.0	51.3±3.3	59.0±2.0	91.8±3.9
	Wide	8	0.80±0.06	0.85±0.07	0.95±0.14	0.30±0.09	32.7±2.2	48.0±2.2	57.0±3.5	92.2±4.1
Maro	Narrow	5	0.72±0.01	0.87±0.07	0.83±0.06	0.24±0.04	38.9±1.3	52.0±2.1	60.7±2.5	93.6±0.8
	Wide	7	0.73±0.01	0.86±0.06	0.85±0.05	0.25±0.03	39.4±1.4	51.0±1.0	62.3±1.6	93.3±1.3

Full text

INTRODUCTION

Primates, though primarily quadrupedal, generally exhibit distinct differences in their gait compared with typical quadrupedal mammals such as dogs and horses. They most often use a diagonal sequence gait (DS) rather than a lateral sequence gait (LS) (Hildebrand, 1967, 1985; Cartmill et al., 2002; Schmitt and Lemelin, 2002), tend to have more compliant forelimbs due to greater elbow flexion–extension during stance (Larson et al., 2000, 2001; Larney and Larson, 2004; Schmitt, 1999), and usually support more body weight on the hindlimbs than on the forelimbs (Kimura et al., 1979; Reynolds, 1985; Demes et al., 1994), although there are notable exceptions among primate species.

These differences are considered to reflect evolutionary adaptations related to arboreal locomotion and grasping functions of the forelimbs and hindlimbs in primates. Primates are thought to have evolved from small, nocturnal mammals into a group specialized in navigating the terminal branches of tropical rainforests (Cartmill, 1974; Rasmussen, 1990; Sussman, 1991). This ecological shift is assumed to have favored the development of enhanced visuomotor control for reaching and grasping branches and exploiting foods such as flowers and insects. In association with these changes, the limbs became increasingly functionally differentiated: although both the forelimbs and hindlimbs play important roles in grasping and securing the substrate, the hindlimbs in many lineages contribute proportionally more to body weight support and propulsion. In addition, the need to maintain balance and minimize the risk of falling in a complex three-dimensional arboreal environment has been proposed as one factor favoring the use of DS walking in primates. In this gait, the hindlimb is placed in a location already contacted by the contralateral forelimb, allowing the animal to confirm the substrate's stability and safety. This strategy contributes to more stable and cautious locomotion, which is particularly important when traversing narrow or unpredictable substrates such as tree branches (Cartmill et al., 2002). Thus, the characteristic quadrupedal gait of primates is widely regarded as a consequence of fundamental adaptations to arboreal life. It should be noted, however, that comparative, developmental and theoretical studies suggest that its functional significance is more complex and probably reflects the interaction of arboreal habitat use with limb and trunk mechanics, body size and phylogenetic history, rather than a single, fine-branch-specific adaptation (Shapiro and Raichlen, 2005; Stevens, 2006; Wallace and Demes, 2008; Dunham et al., 2019; Usherwood and Smith, 2018; Cartmill et al., 2020).

Understanding the evolution of primate locomotion, therefore, requires detailed analysis of their quadrupedal gait. Given its central role in arboreal movement, quantitative biomechanical research on primate quadrupedalism has expanded in recent decades, with an increasing number of studies examining locomotion on arboreal substrates such as horizontal poles, ladders and branches under ecologically relevant conditions (Granatosky et al., 2016; Druelle et al., 2021; Higurashi et al., 2009, 2010; Higurashi and Kumakura, 2021; Nyakatura and Heymann, 2010). One important aspect of arboreal locomotion is substrate diameter. Locomotion on narrow substrates, such as small tree branches, poses a greater risk of falls and demands precise control of body posture and limb placement. Previous studies have begun to address this issue by examining the impact of substrate diameter on gait mechanics using a horizontal pole across various primate species (Schmitt, 1999; Schmitt et al., 2006; Schmidt, 2005; Stevens, 2008). Research has shown that clear diameter-dependent modifications in gait occur in common marmosets (Young, 2009; Young et al., 2016), but in squirrel monkeys (Young, 2009; Schapker et al., 2022) and mouse lemurs (Shapiro et al., 2016), the effect appears to be less pronounced. Quadrupedal walking of capuchins has also been investigated (Wallace and Demes, 2008; Carlson and Demes, 2010). Although Schmitt and colleagues (Schmitt, 1999; Schmitt et al., 2006; Schmidt, 2005) have included large-bodied catarrhines in their comparative analyses, most detailed studies on diameter-dependent gait modifications have focused on small- to medium-sized platyrrhine monkeys. Consequently, the influence of substrate diameter on locomotion in larger catarrhines remains relatively less well documented, and, to our knowledge, no study has yet examined footfall pattern, whole-body kinematics and kinetics in an integrated manner for quadrupedal walking on narrow cylindrical supports.

The aim of this exploratory study was to investigate how the diameter of a horizontal pole influences the kinematics and kinetics of quadrupedal walking in Japanese macaques (Macaca fuscata). Japanese macaques are typical semi-terrestrial quadrupeds: they use the ground for roughly 60–70% of their locomotor activities and are therefore often classified as terrestrial, although they still make frequent use of arboreal substrates (Chatani, 2003). When walking on narrow substrates, animals must adjust their locomotion to maintain balance under more constrained conditions. At low speeds, they mainly rely on quasi-static stability, keeping the center of mass (COM) within the base of support. As speed increases, however, dynamic balance becomes more important: stability is maintained by controlling whole-body momentum and step-to-step adjustments even when the COM moves relative to, and can transiently fall outside, the base of support (Gàlvez-Lòpez et al., 2011). We therefore tested the following four hypotheses. (1) As pole diameter decreases, walking speed would decrease and limb contact duration would increase, because macaques adopt a more cautious gait strategy to ensure stability. (2) As pole diameter decreases, fluctuations of the COM in the mediolateral direction would be reduced, keeping the COM more above the substrate center and resulting in a smaller mediolateral ground reaction force. (3) Macaques would lower their COM height to improve stability on narrower poles. (4) The elbow and knee joints would remain more flexed throughout the stance phase, leading to a more crouched posture that facilitates balance control on narrow substrates. Furthermore, to elucidate the effects of substrate conditions on quadrupedal locomotion, we compared the present results for ground reaction forces, footfall patterns and mechanical energy recovery during quadrupedal walking on horizontal poles with those obtained on a flat surface in our previous study (Ogihara et al., 2012).

MATERIALS AND METHODS

Four regularly trained Japanese macaques (Macaca fuscata Blyth 1875) from the Suo Monkey Performance Association (Kumamoto, Japan) participated in the experiment (Table 1). We have previously investigated the kinematics, kinetics and energetics of bipedal locomotion in these trained macaques (Hirasaki et al., 2004; Nakatsukasa et al., 2004, 2006; Ogihara et al., 2005, 2007, 2009, 2010, 2012, 2018; Blickhan et al., 2018, 2021, 2024a,b), but here we analyzed the kinematics and kinetics of quadrupedal walking on a horizontal pole. The experiment was approved by the Animal Care Committee at the University of Tokyo (A2024S001). Although the macaques used in this study are trained to walk bipedally, they still spend the great majority of their daily life moving and behaving quadrupedally in their home cages. Based on our observations, their quadrupedal gait appears typical for Japanese macaques, and we do not expect their capacity for quadrupedal walking to be altered by the bipedal training. A previous study has reported skeletal and postural adjustments related to bipedal behavior, particularly an increased lumbar lordosis in the upright posture (Hayama et al., 1992), but these changes are mainly expressed during bipedal stance and are unlikely to have a major influence on quadrupedal locomotion. All experiments were conducted in May 2024. For each individual, data were collected within a single experimental session on one day. Each session lasted approximately 1–1.5 h per animal, including practice and rest periods between trials.

The macaques walked quadrupedally on horizontal metal poles of three different diameters (48.6, 76.3 and 114.3 mm), which were mounted on two force plates (FR4060-07, Bertec, Columbus, OH, USA) at a height of approximately 300 mm from the force plate surfaces (approximately 350 mm from the concrete floor) (Fig. 1). The macaques were guided by a personal coach. Each 40 cm-long pole was rigidly fixed to a rectangular aluminium frame, which was securely screwed to the corresponding force plate. The total length of the horizontal pole was approximately 5 m. For reference, based on computed tomography (CT) data from a ∼10 kg Japanese macaque (Ogihara et al., 2009), hand, shoulder and hip breadths are approximately 37, 161 and 149 mm, respectively. When pole diameter is normalized to 1, these correspond to about 0.8, 3.3 and 3.1 diameters for the narrow pole (48.6 mm) and about 0.3, 1.4 and 1.3 diameters for the wide pole (114.3 mm); because the macaques in this study were smaller (5–8 kg), the relative diameters would be slightly larger than these conservative estimates. During DS walking in Japanese macaques, hand contact with the substrate is followed by that of the ipsilateral foot. Consequently, the ipsilateral hand and foot make contact with the substrate with their points of contact positioned very close to each other. This combination of spatial and temporal proximity makes it challenging to independently measure the ground reaction force (GRF) vectors of the ipsilateral hand and foot. Therefore, in the present study, the sum of the ipsilateral hand and foot GRF vectors was measured using the first force plate, while the sum of the contralateral hand and foot GRF vectors was measured using the second force plate. Ground reaction torques generated by the hand grip on the pole could, in principle, affect the balance of whole-body angular momentum. However, in the present experiments, the horizontal pole was relatively thick, and the monkeys rested their hands on the top of the pole rather than firmly gripping it. Therefore, we expect grip-related torques around the longitudinal axis of the pole to be small compared with the moments produced by the limbs, and we did not include them in our analysis.

Experimental setup for measuring the kinematics and kinetics of quadrupedal walking on a horizontal pole in Japanese macaques. The macaques walked quadrupedally at a self-selected speed on horizontal poles of three different diameters (48.6, 76.3 and 114.3 mm; length ∼5 m) mounted on two force plates and were filmed using four digital cameras. The rectangular aluminium frames were covered with cloth, which the monkeys disliked stepping on, obliging them to walk on the poles instead.

The macaques walked along the horizontal pole at a self-selected speed, and three components of the GRF vectors were recorded at 100 Hz. Walking sequences in which the macaques maintained a steady gait were selected. To exclude trials where the macaques walked too slowly or too fast, only sequences where the time interval from hand contact on the first force plate to foot-off of the contralateral foot from the second force plate was within one standard deviation of the mean for that macaque and pole diameter were included in the analysis. Kinetic data were collected under three pole diameter conditions (narrow, medium and wide). However, preliminary analyses showed that the medium condition did not differ substantially from the other two (Fig. S1), and these data were therefore excluded from further analysis, yielding a total of 57 valid recordings across individuals (at least four trials per individual per condition).

The macaques were simultaneously recorded at 60 Hz using four digital cameras (60Hz, 1920×1080 pixels; Everio GZ-G5, JVC, Yokohama, Japan) (Fig. 1). The cameras were time synchronized at the moment when the forelimb first contacted the instrumented pole. Two cameras were placed on the right side and two on the left side of the animal, at a distance of approximately 2 m from the center of the instrumented section of the pole, providing overlapping views for sagittal-plane reconstruction (Fig. 1). From the motion images, the stride duration and footfall patterns (limb phase defined as the percentage of the stride duration in which each limb contacted or lifted from the substrate) of each walking sequence were measured. In this study, stride duration was defined as the time interval from the initial contact of the hand on the first force plate to the next contact of the same hand. In addition, four temporal parameters were determined with the stride cycle normalized to 100% (Hildebrand, 1965, 1966; Abourachid, 2003): (1) the timing of touchdown of the ipsilateral hindlimb relative to the ipsilateral forelimb (phase A); (2) the timing of touchdown of the contralateral forelimb relative to the ipsilateral forelimb (phase B, symmetry); (3) the timing of lift-off of the forelimb relative to its own touchdown (phase C, duty factor of the forelimb); and (4) the timing of lift-off of the hindlimb relative to the touchdown of the ipsilateral forelimb (phase D). The hindlimb duty factor was calculated as phase D−phase A, and diagonality as 100−phase A.

For the kinematic analysis, we likewise focused on the narrow and wide conditions. Because 3D reconstruction required extensive manual processing, we selected three trials for each individual in each condition (24 trials in total). From the valid trials for each condition, we chose those for which walking speed and kinematic patterns fell within the typical range observed for that individual and treated them as representing that individual's average performance. Such selected sequences were digitized semi-automatically from the video recordings using DeepLabCut, a deep learning-based markerless pose estimation toolbox that enables tracking of user-defined body parts across frames (Mathis et al., 2018). For this study, a set of anatomical landmarks was manually labeled in a subset of frames to train a neural network, which was then used to predict the 2D positions of the landmarks in all frames. A total of more than 400 frames were manually labeled to train the DeepLabCut network. The network was trained for 100,000 iterations with the default DeepLabCut parameters until the training loss converged. As the estimated 2D coordinates were not always accurate, we carefully checked them visually and made manual corrections as needed using Argus, an open-source motion analysis software (Jackson et al., 2016). The 2D coordinates obtained from multiple camera views were subsequently reconstructed into 3D coordinates using triangulation, which was also performed using Argus. Argus allows the integration of camera calibration data, frame synchronization and 3D reconstruction of landmark trajectories via the direct linear transformation (DLT) method. Spatial calibration was performed using a 585 mm cubic frame placed on the force plates. Based on this procedure, the mean (±s.d.) absolute error in measuring the distance between two points 1 m apart was 2.3±3.2 mm. A total of 16 joint positions (eight on each side) were digitized (Fig. 2): (1) head of the fifth metatarsal, (2) lateral malleolus of the fibula, (3) lateral epicondyle of the femur, (4) greater trochanter, (5) acromion, (6) lateral epicondyle of the humerus, (7) styloid process of the ulna and (8) head of the fifth metacarpal. In addition, the positions of the eyes and tail base were digitized, resulting in a total of 19 landmarks (Fig. 2). The positional data were low-pass filtered at 14 Hz using a second-order, zero phase shift digital filter (Ogihara et al., 2010). Based on these marker coordinates, the sagittally projected joint angles of the ankle, knee, hip, wrist, elbow and shoulder were calculated (Shimada et al., 2017). We did not quantify frontal-plane joint angles because, with the cameras positioned laterally to optimize sagittal-plane tracking, depth resolution was reduced and frontal-plane angles could not be estimated with sufficient accuracy.

Digitized marker positions and definitions of joint angles. Top: positions shown correspond to: (1) head of the fifth metatarsal (MT), (2) lateral malleolus of the fibula, (3) lateral epicondyle of the femur, (4) greater trochanter, (5) acromion, (6) lateral epicondyle of the humerus, (7) styloid process of the ulna and (8) head of the fifth metacarpal (5MC). COM, center of mass. Bottom: an example of digitization using DeepLabCut.

Using the estimated 3D kinematics of whole-body motion during quadrupedal locomotion, we calculated the displacement profiles of the whole-body COM for each sequence following the method of Winter (2005), using segment mass and segment COM position data shown in Table 2 (Ogihara et al., 2011; Oku et al., 2021), determined using the anatomically based whole-body musculoskeletal model of the Japanese macaque (Ogihara et al., 2009). The mediolateral fluctuation of the COM was quantified as the standard deviation of its mediolateral displacement within a gait cycle; a larger standard deviation indicates a greater mediolateral fluctuation. The height of the COM was defined as the mean vertical position of the COM relative to the superior ridge of the pole. In order to account of the size difference of the subjects, the linear dimensions such as stride length, mediolateral fluctuation of the COM and COM height were normalized by dividing them by the effective leg length, defined as the mean distance between the hip joint and the head of the metatarsus during the stance phase (Raichlen et al., 2008; Ogihara et al., 2010). For size normalization of walking velocity, we also computed the Froude number (Fr) as the size-normalized velocity measure, defined as Fr=v²/g**L, where v is the walking velocity, g is the gravitational acceleration and L is the effective leg length. The velocity here was calculated as stride length divided by stride duration.

To characterize the gait mechanics of quadrupedal walking on a horizontal pole, we analyzed the temporal fluctuations of the whole-body COM potential energy (PE) and kinetic energy (KE) throughout a gait cycle, based on GRF data (Cavagna, 1975). The total external mechanical energy (TME) was defined as the sum of PE and KE, and the rate of energy recovery was calculated to estimate the proportion of mechanical energy conserved via the inverted pendulum mechanism (Cavagna et al., 1977; Griffin et al., 2004):

$$[eqn]$$

where ΔPE, ΔKE and ΔTME represent the total increments of potential energy, kinetic energy and total mechanical energy over one gait cycle, respectively (Cavagna et al., 1977). For this, normalized GRF profiles were averaged, and contralateral hand and foot forces were reconstructed under the assumption of bilateral symmetry. The net GRF acting on the COM was then obtained by summing the measured ipsilateral GRFs with the reconstructed contralateral GRFs, using the measured timing of contralateral limb contacts. Translational acceleration of the COM was calculated by dividing GRFs by body mass and subtracting gravitational acceleration. This acceleration was numerically integrated to estimate vertical displacement as well as vertical, anteroposterior and mediolateral velocities of the COM for each stride, from which PE and KE were derived (Ogihara et al., 2007). To satisfy the physical requirement that the integrated vertical GRF over one stride equals body mass×stride duration, a small offset correction was applied. The initial vertical velocity was set so that COM height at the beginning and end of the cycle matched, assuming steady-state cyclic locomotion. Percentage congruity, defined as the percentage of the stride cycle during which PE and KE changed in the same direction, was also calculated to quantify the extent to which PE and KE fluctuated in-phase (Ahn et al., 2004). The mechanical energy was divided by body mass for normalization.

To account for the non-independence of repeated trials within individuals, we analyzed the spatiotemporal variables using linear mixed-effects models. For each dependent variable, we fitted a separate model with pole condition (narrow versus wide) as a fixed effect and individual identity as a random intercept. Models were estimated by restricted maximum likelihood, and the significance of the pole effect was evaluated using Wald z-tests, with P<0.05 considered statistically significant. For mediolateral COM fluctuation, the response was log transformed to improve numerical stability, and the mixed-effects model was fitted using the Powell optimizer. These analyses were performed in Python using the statsmodels package.

RESULTS

Table 3 summarizes the means±s.d. of gait parameters for each subject. The smallest individual (Maro) exhibited slightly lower values but, overall, stride length was approximately 0.8–0.9 m, stride duration 0.8–0.9 s and velocity 0.9–1.1 m s^−1^ across pole diameters. The corresponding Froude number ranged from 0.25 to 0.37, indicating a normal gait at a moderate speed. The duty factors of both forelimbs and hindlimbs were slightly below 60%, and diagonality exceeded 60%, indicating a DS, diagonal-couplet gait typical of primate quadrupedal locomotion. The limb phase for the touchdown of the contralateral forelimb relative to the forelimb was confirmed to be approximately 50%.

Analysis of the gait parameters using linear mixed-effects models indicated that only stride length differed significantly between pole conditions. Stride length was found to be greater on the wide pole (estimate for wide versus narrow: 0.045 m, P<0.001). However, as a sensitivity analysis using participant-level mean stride length (one value per individual per condition; n=4 paired observations), neither a Mann–Whitney U-test (U=6, P=0.67) nor a paired Wilcoxon signed-rank test (W=0, P=0.07) detected a significant difference between pole conditions. Walking velocity and Froude number tended to be higher on the wide pole, but these effects did not reach statistical significance (velocity: 0.056 m s^−1^, P=0.074; Fr: 0.034, P=0.081). The stride duration and four limb phases (A–D) were also not significantly affected by pole condition (P=0.845, 0.213, 0.373, 0.242 and 0.116, respectively).

Fig. 3 compares the mean anteroposterior, mediolateral and vertical GRF profiles between the two pole conditions. Data represent the means of all trials from both sides of all four macaques. Time is expressed as a percentage of the interval from forelimb touchdown (0%) to ipsilateral hindlimb lift-off (100%), and forces were normalized to body weight. Longitudinal GRFs are plotted as negative for braking and positive for propulsion, whereas transverse GRFs are plotted as positive in the lateral direction. The vertical GRF on one side reached its maximum when the ipsilateral forelimbs and hindlimbs were simultaneously in contact with the substrate. In the longitudinal direction, the propulsive force generated by the forelimb was largely counterbalanced by the braking force produced by the ipsilateral hindlimb. Consequently, the forelimbs acted predominantly as net braking elements, whereas the hindlimbs served primarily as net propulsive elements. Transverse GRFs remained close to zero throughout the stride cycle. Overall, the waveforms were very similar between the two conditions, showing virtually no effect of pole diameter on GRF profiles in the present study.

Three components of mean ground reaction force (GRF) in the ipsilateral forelimbs and hindlimbs during quadrupedal walking on a narrow and wide horizontal pole. Lines and shading represent mean±s.d. GRF in the anteroposterior (A), mediolateral (B) and vertical (C) direction. BW, body weight. The bars at the bottom indicate the average foot–ground contact duration of the forelimbs (F) and hindlimbs (H).

Fig. 4 compares the mean sagittally projected joint angle profiles between the two pole conditions. Joint angles were defined as positive for protraction of the hip and shoulder joints, flexion of the knee and elbow joints, and dorsiflexion of the ankle and wrist joints. Angles were set to zero when the adjacent segments were aligned in a straight line (Fig. 2). In Japanese macaques, joint angle profiles revealed a generally crouched posture, with flexed knee and elbow joints and marked plantarflexion of the wrist in late stance phase. However, the overall joint and segment angle profiles were almost identical between the two pole diameter conditions, indicating that pole diameter had no effect on kinematic patterns of quadrupedal walking in Japanese macaques.

Mean joint angle profile during quadrupedal walking on a narrow and wide horizontal pole. Lines and shading represent mean±s.d. ankle (A), knee (B), hip (C), wrist (D), elbow (E) and shoulder (F) angle. The bars at the bottom indicate the average foot–ground contact duration of the forelimb (F) and hindlimb (H).

Fig. 5 compares the normalized mediolateral fluctuation and normalized height of the COM between the narrow and wide pole conditions. The mediolateral fluctuation, being very small, amounting to only about 1.5% of the effective leg length, did not differ significantly between conditions (P=0.314; mixed-effects model fitted to log-transformed values). In contrast, the normalized COM height was significantly greater in the wide condition (P=0.001), although the difference was modest, corresponding to approximately 2% of the effective leg length.

Comparison of normalized mediolateral fluctuation and COM height between the narrow and wide horizontal poles. Box plots (median, upper and lower quartiles and 1.5× interquartile range) show mediolateral fluctuation (A) and COM height (B) for all four macaques (symbols indicate individual data points).

Fig. 6 compares the GRF profiles during quadrupedal walking on the narrow horizontal pole with those obtained on a flat surface in our previous study (Ogihara et al., 2012), highlighting the effects of substrate conditions on locomotion. Data from the present study represent the mean of individual mean waveforms from the four macaques, whereas the flat-surface data represent the mean of individual mean waveforms from three different macaques reported in Ogihara et al. (2012). The vertical GRF profiles were generally similar between the two substrates. In the longitudinal direction, however, a pronounced braking force was observed at hindlimb contact on the flat surface, but this effect was not observed on the pole and generation of the propulsive force was more continuous. Conversely, braking at forelimb contact appeared greater on the pole than on the flat surface. In the transverse direction, the flat surface condition showed relatively larger medial forces generated by the hindlimbs, whereas these forces were suppressed on the pole. Both forelimb and hindlimb forces in the transverse direction were generally reduced when walking on the pole compared with walking on the flat surface.

Comparison of the three components of the GRF during walking on a narrow horizontal pole and a flat surface. Lines and shading represent mean±s.d. GRF in the anteroposterior (A), mediolateral (B) and vertical (C) direction.

Fig. 7 compares the footfall patterns during quadrupedal walking on the narrow horizontal pole with those obtained on a flat surface (Ogihara et al., 2012). In both conditions, the contralateral hindlimb was already in contact with the ground at the time of forelimb touchdown. However, on the pole, this hindlimb lifted off earlier, resulting in a period during which only the ipsilateral forelimb and hindlimb supported the body (asterisks in Fig. 7), whereas this period was virtually absent in quadrupedal walking on the flat surface. Accordingly, the major difference between the two substrates was reflected in phase D. A Mann–Whitney U-test comparing the phase D values between the narrow horizontal pole (Table 3) and the flat surface (table 1 of Ogihara et al., 2012) indicated that phase D was significantly smaller in the pole condition than in the flat surface condition (one-sided, P=0.026).

Comparison of mean footfall patterns during quadrupedal walking on a horizontal pole and a flat surface. The bars indicate the contact times of the four limbs for horizontal pole (A) and flat-surface (B) walking. F, forelimb; iHL, ipsilateral hindlimb; cFL, contralateral forelimb; cHL, contralateral hindlimb. Note that two-point contact of the ipsilateral forelimb and hindlimb (asterisks) occurred on the horizontal pole, whereas this period was very short on the flat surface.

Fig. 8 compares the mean fluctuation profiles of normalized PE, KE and TME in the four subjects during walking on the pole. Only the results for the narrowest pole are presented, as the GRF profiles were virtually identical between the narrowest and widest poles, and the primary aim was to compare mechanical energy during horizontal pole walking in the present study with flat-surface walking reported in our previous study (Ogihara et al., 2012). PE and KE profiles oscillated approximately out of phase and each showed two peaks within a stride, suggesting that some degree of mutual exchange between PE and KE of the COM occurred during quadrupedal walking on a horizontal pole in the Japanese macaque. The PE profile was sinusoidal, with peaks occurring at approximately 30–40% and 80–90%, and troughs at approximately 10% and 60% of the stride cycle in all subjects. The KE profile showed peaks occurring nearly at the same instants, indicating that PE and KE fluctuated partly out of phase. Consequently, fluctuation in TME was small. However, the calculated percentage congruity values were 43%, 39%, 25% and 30% for Kurimatsu, Shiotaro, Matakichi and Maro, respectively, indicating that PE and KE were not perfectly out of phase. The calculated percentage recovery values were 34.9%, 40.1%, 47.2% and 51.1% for the same subjects, respectively.

Fluctuation in potential energy (PE), kinetic energy (KE) and total external mechanical energy (TME) of the COM during quadrupedal walking on a narrow horizontal pole.

DISCUSSION

In this study, we examined how horizontal pole diameter influences the kinematics and kinetics of quadrupedal walking in Japanese macaques. Analysis of the gait parameters using linear mixed-effects models indicated that, apart from a modest decrease in stride length on the narrow pole (∼0.05 m), spatiotemporal variables were largely unaffected by pole diameter. Specifically, walking speed and Froude number did not decrease, stance duration did not increase, mediolateral fluctuations of the COM were not reduced, and elbow and knee joints were not maintained in a more flexed posture on the narrow pole. Thus, none of the four proposed hypotheses were supported, and the null hypothesis could not be rejected. These findings indicate that quadrupedal walking of Japanese macaques remains stable on poles of the diameters tested in this study. Maintaining stability on narrow substrates is generally considered more demanding in the frontal plane, because mediolateral GRFs generate moments around the COM that can displace it beyond the base of support. While such conditions would probably cause considerable swaying in humans, they did not pose a substantial challenge for the macaques. It must be noted, however, that the normalized COM height was significantly lower in the narrow condition than in the wide condition. However, this difference corresponds to only about 2% of the effective leg length of an approximately 30 cm Japanese macaque, which is approximately 6 mm in absolute terms. This slight lowering of the body may simply reflect the fact that, on the narrower pole, the hands tend to be positioned below the superior edge because they must be placed on the lateral rather than the upper surface of the pole. Even if this were not the case, a 6 mm difference in COM height is negligible in terms of balance control. For instance, with a COM height of 300 mm, if the COM sways laterally by 10 mm, lowering the COM by 6 mm would reduce this displacement by only about 0.1 mm. This minimal change indicates that the reduction in COM height did not substantially contribute to enhanced stability during walking.

However, a comparison between horizontal poles and flat surfaces revealed that quadrupedal walking on poles is characterized by reduced mediolateral GRFs and a relative decrease in the braking component at hindlimb touchdown. The reduction in mediolateral forces can be explained by the fact that large mediolateral forces would generate moments around the COM and displace it outside the narrow support base; thus, the macaques modulated their gait to minimize such forces. The smaller braking impulse of the hindlimbs appears to result from the geometric constraints of walking along a straight line on the pole. On a flat surface, the hindfoot typically lands beside the ipsilateral forefoot, which is still in contact with the ground. On the pole, however, this placement is not possible, forcing the hindfoot to land more posteriorly, thereby reducing the braking component at touchdown.

Furthermore, the present study found that the hindlimb lifted off earlier on the pole, resulting in a period during which only the ipsilateral forelimbs and hindlimbs supported the body, whereas such a period was virtually absent during quadrupedal walking on the flat surface (Fig. 7). This difference was observed despite our initial expectation that support duration would generally be longer for both forelimbs and hindlimbs on the pole, given its reduced stability. One possible explanation is a geometric constraint: maintaining three points of contact in a straight line along the pole may be anatomically and mechanically difficult, and such a colinear arrangement does not enlarge the base of support. Consequently, the macaques may have reduced the duration of this posture. Another possible explanation is that having fewer than three points of contact may enhance maneuverability. Having fewer supporting limbs may allow quicker adjustment of the COM, make it easier to control mediolateral forces, and even provide an advantage in maneuverability despite the higher intrinsic instability of the pole. The present study therefore suggests that balance control is not the sole determinant of gait strategy on narrow supports.

Mechanical energy recovery due to pendular exchange was compared between walking on a horizontal poles and on a flat surface (Ogihara et al., 2012), and was found to be similar (35–50%) in the two conditions. This is consistent with Druelle et al. (2021), who reported no statistically significant difference in energy recovery between ground and pole walking in baboons (43% versus 32%, respectively). Percentage recovery during quadrupedal locomotion on poles has also been examined in squirrel monkeys (Miller et al., 2024), in which values were likewise relatively low (25%). It has been reported that percentage recovery during quadrupedal walking in primates is generally lower than that of typical quadrupeds (Cavagna et al., 1977; Griffin et al., 2004; Usherwood et al., 2007; Ogihara et al., 2012). The present results indicate that it remains essentially unchanged. Our previous study suggested that the relatively low percentage recovery in primates arises because the amplitudes of PE and KE fluctuations do not match well, reflecting the compliant nature of their locomotor mechanics (Ogihara et al., 2012). The present study confirmed a similar tendency during pole walking, thereby supporting this interpretation.

As noted above, during pole walking, the mediolateral forces were minimized. In contrast, in human bipedal walking, although the absolute magnitude of mediolateral forces is small, they play an important role in balancing the angular momentum about the vertical axis, because the moment arm around the COM is large in the horizontal plane particularly at heel contact (Negishi and Ogihara, 2023). On the pole, however, such a mechanism cannot be effectively utilized. Nevertheless, cancellation of whole-body angular momentum around the vertical axis remains important even during pole walking. Because the swinging limbs strongly affect whole-body angular momentum, it is advantageous for the hindlimb and the contralateral forelimb to move largely in phase, thereby canceling each other's effects. This pattern is realized in DS walking, whereas in LS walking the forelimbs and hindlimbs of the same side swing together, making such cancellation difficult. Thus, it may be that selection pressures on effective control of angular momentum around the vertical axis in arboreal environments contributed to the adoption of DS walking in primate species. Although this interpretation cannot be confirmed definitively, the present findings are consistent with such a possibility.

Limitations and perspectives

One limitation of the present study is that the poles used were probably not narrow enough to elicit substantial postural adjustments. Further experiments using thinner poles will be necessary to determine whether narrower substrates impose more pronounced demands on balance control during quadrupedal walking. In addition, in natural arboreal substrates, branch diameter and compliance are often coupled: larger branches tend to be less compliant, whereas smaller branches tend to be more compliant. Although our experimental design intentionally varied pole diameter under controlled conditions, this necessarily decouples properties that are typically correlated in ecological settings. A useful next step would therefore be to test supports that preserve this natural covariation (e.g. combining smaller, more compliant and larger, stiffer supports), and to quantify how macaques adjust their gait and balance when diameter and compliance vary together rather than independently.

Another limitation is that we examined only steady quadrupedal walking at a self-selected speed. It remains unclear how gait strategies might change under different locomotor conditions, such as faster speeds, varying slope angles or differences in pole compliance and surface properties. Future studies incorporating a wider range of locomotor tasks, as well as experimental manipulation of pole conditions, will provide further insight into the adaptive mechanisms underlying quadrupedal locomotion on narrow supports. From an ecological perspective, the metal poles used here do not replicate the frictional and compliant properties of natural tree branches. Thus, caution is needed when extrapolating them to arboreal locomotion on rough, compliant branches in the wild.

Finally, to avoid overinterpreting potentially noisy measures, the present study focused on sagittal-plane angles, which can be reconstructed more reliably with our current setup. In future work, we will expand the camera configuration by adding additional viewpoints and optimizing camera placement to improve depth resolution, enabling more robust quantification of frontal-plane joint angles, particularly at the distal joints.

Supplementary Material

10.1242/jexbio.252010_sup1Supplementary information