Comparison of spatiotemporal gait characteristics between vertical climbing and horizontal walking in primates

Quadrupedal walking in primates is differentiated from that of most mammals by a suite of features (Cartmill et al., 2002; Demes et al., 1994; Granatosky et al., 2016b; Kimura et al., 1977; Reynolds, 1985; Schmitt, 1994; Schmitt and Hanna, 2004). First, primates tend to utilize diagonal-sequence footfall patterns (i.e. each hindlimb footfall is followed by a contralateral forelimb footfall) versus lateral-sequence footfall patterns (i.e. each hindlimb footfall is followed by an ipsilateral forelimb footfall) (Cartmill et al., 2002, 2007; Hildebrand, 1967; Rollinson and Martin, 1981; Shapiro and Raichlen, 2007; Vilensky and Larson, 1989). Second, quadrupedal primates exhibit relatively larger limb excursions, resulting in relatively long stride distances (Larson et al., 2000, 2001; Schmitt, 1998). Finally, most primates exhibit functional differentiation between the forelimbs and hindlimbs, where the hindlimb serves as the primary support and propulsive limb (Demes et al., 1994; Granatosky et al., 2018a; Hanna et al., 2017; Kimura et al., 1977; Schmitt and Hanna, 2004; Schmitt and Lemelin, 2002). It remains unresolved what selective factors initially drove primates to adopt these unusual locomotor characteristics.

One scenario that has received relatively little attention is that the mechanical requirements of vertical climbing may be the selective pressure for the evolution of certain aspects of the unusual pattern of primate quadrupedal walking (Hanna et al., 2017; Prost and Sussman, 1969; Vilensky et al., 1994). Climbing, often on vertical supports, is a critical and fundamental form of locomotion for arboreal animals during foraging, travel, escape or finding a safe resting place. As originally reported by Prost and Sussman (1969) and Vilensky and colleagues (1994), and later supported by Nyakatura et al. (2008), increasing support inclination increases the presence of diagonal-sequence gaits over lateral-sequence gaits in primates. This finding led Prost and Sussman (1969) and Vilensky et al. (1994) to propose that, as climbing became more important to the locomotor repertoire of early primates, the frequency of diagonal-sequence-gait utilization also increased. Those previous studies focused on increased inclination, but not pure vertical supports, which raises the question of how limb phase will vary with a 90 deg incline. Hirasaki et al. (1993) examined footfall patterns in Macaca fuscata and Ateles geoffroyi, and found that macaques showed mostly diagonal-sequence diagonal-couplet (DSDC) gaits (and trots), whereas spider monkeys tended more toward a pace when climbing a vertical support. Data collected by Hanna and Schmitt (2011) places Macaca fasicularis solidly in with M. fuscata in the use of DSDC gaits and trots. However, neither study compared explicitly to horizontal supports, although it is known that all three species tend to use DSDC gaits on horizontal supports. Macaques, therefore, retain the pattern seen on the horizontal supports while spider monkeys deviate in a direction toward laterally coordinated limb behavior. More recently, Hanna et al. (2017) reported that, during climbing, the forelimb experiences primarily tensile loads, while hindlimb loading is primarily compressive and propulsive, suggesting that functional differentiation of the limbs could have evolved in association with climbing and been maintained on level surfaces.

While evidence supporting Prost and Sussman (1969) and Vilensky and colleagues’ (1994) original hypothesis has been reported in a number of studies (Hanna et al., 2017; Nyakatura and Heymann, 2010; Nyakatura et al., 2008), direct comparisons between vertical climbing and horizontal walking remain scarce. Furthermore, comparisons that have been conducted between the two locomotor modes are usually limited in terms of species and body size range. Therefore, this study explores spatiotemporal gait variables of vertical climbing and horizontal walking in nine species of primarily arboreal primates across a range of body masses (0.18–9.77 kg) in order to test the hypothesis that the mechanical requirements associated with vertical climbing may represent a selective factor in the formation of certain aspects of the unusual pattern of primate quadrupedal locomotion.

In an arboreal context, gait patterns may serve as a behavioral mechanism to enhance stability and safety (Cartmill et al., 2002, 2007; Hildebrand, 1967; Preuschoft, 2002; Rollinson and Martin, 1981; Shapiro and Raichlen, 2007). Habitual use of DSDC gaits is common among primates, some marsupials and the carnivoran Potos flavus (Cartmill et al., 2002, 2007; Granatosky et al., 2016b; Hildebrand, 1967; Karantanis et al., 2015; Schmitt and Lemelin, 2002). There are two primary hypotheses for the benefits of a DSDC gait in an arboreal context. First, DSDC gaits essentially minimize the amount of time that the limbs are arranged as a unilateral bipod (Cartmill et al., 2002; Shapiro and Raichlen, 2007). This may be especially important for arboreal quadrupedal primates while on arboreal supports because it reduces the chances of toppling off the support (Cartmill et al., 2007; Shapiro and Raichlen, 2007). The second explanation is based on the distinct limb-loading pattern observed in primates that essentially allows them to support the majority of their body weight on the hindlimb rather than the forelimb (Reynolds, 1985; Schmitt and Lemelin, 2002). By adopting a DSDC walking gait, the forelimb is able to test supports before the animal commits the majority of its body weight (Cartmill et al., 2002; Schmitt and Lemelin, 2002). If the support should fail due to local weakness, then the animal is able to utilize the horizontal lever effect to quickly pitch backwards onto the hindlimb and use its grasping hindfoot to secure a grasp, thereby preventing a fall (Larson and Demes, 2011; Larson and Stern, 2009; Reynolds, 1985; Young, 2012). It should be noted that this mechanism has only been addressed thoroughly during quadrupedal walking, and it is unclear what, if any, benefits shifting weight backwards to the hindlimbs would have during behaviors such as climbing, running or leaping.

While these explanations have been discussed in great detail in reference to arboreal quadrupedal walking, it is likely that DSDC gaits would not confer the same biomechanical benefits while animals are vertically climbing. The concern of falling forward on a flexible or failing branch articulated by Cartmill et al. (2002) is not significant during vertical movement, nor would hindlimb dominance be a particular advantage over forelimb dominance were a vertical support to bend or break. With this in mind, some have argued that lateral-sequence diagonal-couplet (LSDC) gaits promote stability, in relation to the location of the center of mass (COM) relative to the support polygon of the limbs (Cartmill et al., 2002, 2007; Lammers and Zurcher, 2011; Rollinson and Martin, 1981; Shapiro and Raichlen, 2005, 2007). Of course, the polygon of support model only applies when animals are walking on a horizontal, or at least mostly horizontal, support (Cartmill et al., 2002; Preuschoft, 2002). LSDC gaits are more often exhibited by terrestrial mammals (Cartmill et al., 2002, 2007; Hildebrand, 1967; Muybridge, 1887; Preuschoft, 2002), but are also used by some arboreal specialists, such as sugar gliders (Shapiro and Young, 2010), callitrichids (Nyakatura and Heymann, 2010; Nyakatura et al., 2008) and scansorial rodents (Schmidt and Fischer, 2010). Thus, there are no unambiguous biomechanical grounds for using diagonal- versus lateral-sequence gaits when climbing vertically.

Beyond specific gait patterns, animals can increase security and stability on arboreal substrates by altering the timing at which the limbs come into contact with and release the support (Cartmill, 1985; Dunbar and Badam, 2000; Isler, 2004; Isler and Thorpe, 2003; Karantanis et al., 2016; Preuschoft, 2002). For animals without claws, grasping hands and feet provide muscular support that prevent toppling off or rotating around arboreal supports (Cartmill, 1985; Congdon and Ravosa, 2016; Lammers and Gauntner, 2008). By maintaining longer contact times, animals are able to increase security and stability by assuring that they always have a powerful grasping hand or foot in contact with the support to prevent falling (Cartmill et al., 2002; Karantanis et al., 2015; Patel et al., 2015) and increase the total number of limbs in contact with the support at any one time, resulting in a more stable base throughout the stride (Cartmill et al., 2002; Isler and Thorpe, 2003; Shapiro and Raichlen, 2007).

Although cost of transport is not measured in this study, some might argue, based on studies of horizontal locomotion, that costs of force production would vary with stride length and contact time during climbing (Kram and Taylor, 1990; Pontzer, 2007a,b; Roberts et al., 1998). However, climbing costs do not appear to be associated with changes in step length relative to horizontal locomotion (Hanna and Schmitt, 2011). This is likely because the metabolic cost of climbing by primates can be primarily explained by the cost of performing muscular work against gravity, rather than the rate at which the force is produced (Hanna et al., 2008). Consequentially, vertical climbing results in relatively greater energetic costs compared with horizontal walking in primates at body sizes greater than 1 kg (Hanna et al., 2008). One strategy that large-bodied primates could theoretically use to reduce energetic costs of vertical climbing would be to increase step distance compared with horizontal walking. Thus, we were interested in how stride length varied with climbing in a wide sample of species, in respect to both climbing mechanics and energetic costs. It is worth noting in this context that primates typically take longer strides compared with non-primate species during horizontal locomotion (Alexander and Maloiy, 1984; Larson et al., 2000), and some data (Nyakatura et al., 2008) suggest that, during vertical climbing, primates may increase stride length even more so than in horizontal locomotion. Therefore, this study specifically examines stride length in climbing.

In this study, we have two primary goals: (1) to assess spatiotemporal patterns of vertical climbing in primates across a range of body sizes to determine general trends across species; and (2) to compare spatiotemporal patterns of vertical climbing to horizontal walking to assess the hypothesis that selective pressures associated with vertical climbing may be responsible for certain aspects of the unusual pattern of primate quadrupedal walking mechanics. To focus investigation, we will test the following hypotheses:

Hypothesis 1: diagonal-sequences gaits are the most commonly observed gait type for primates during horizontal walking and vertical climbing, and are not affected by variation in body mass between species.

Hypothesis 2: hindlimb and forelimb relative support duration and speed of movement are similar during vertical climbing and horizontal walking, and are not affected by variation in body mass between species.

Hypothesis 3: stride length, and the tendency to modulate it, are similar during vertical climbing and horizontal walking, and are not affected by variation in body mass between species.

Adult Loris tardigradus (Linnaeus 1758), Nycticebus pygmaeus Bonhote 1907, Cheirogaleus medius Geoffroy 1812, Eulemur mongoz Linnaeus 1766, Daubentonia madagascariensis Gmelin 1788, Saimiri sciureus (Linnaeus 1758), Macaca fascicularis Raffles 1821, Aotus nancymaae Hershkovitz 1983 and Aotus azarae (Humboldt 1811) were used in this study (Table S1). All species are primarily arboreal and commonly use both arboreal horizontal walking and vertical climbing as part of their normal locomotor repertoires (Cant, 1988; Curtis and Feistner, 1994; Fleagle, 2013; Fleagle and Mittermeier, 1980; Gebo, 1987; Glassman and Wells, 1984; Goodenberger et al., 2015; Nekaris, 2001). All data were obtained from animals housed at the Duke Lemur Center and Duke University Vivarium (Durham, NC, USA), Monkey Jungle (Miami, FL, USA), Stony Brook University (Stony Brook, NY, USA), and Michale E. Keeling Center (Bastrop, TX, USA).

All procedures were approved by the appropriate Institutional Animal Care and Use Committee (IACUC; West Virginia School of Osteopathic Medicine: 2007-1, 2008-1, 2009-4; Duke University: A104-09-03; A130-07-05, A270-11-10; State University of New York: 91-94-0131). The data collection procedures have been described extensively elsewhere (Granatosky, 2018a; Granatosky et al., 2016a, 2018; Hanna et al., 2017) and will be simply summarized here. Subjects were encouraged by food reward to climb a pole attached to a wall (climbing trials) or the ground (walking trials). The pole varied in diameter between 1.27 and 3.81 cm (Table S1). Pole diameters were chosen on the basis of previously published studies (Granatosky et al., 2016a; Hanna et al., 2017), which generally attempt to use the smallest pole the animals will utilize (Schmitt and Hanna, 2004). As the animals moved up/across the pole, they were video recorded (A601f; Basler AG, Ahrensburg, Germany, Sony Handycam, or GoPro Hero3+) at 60–120 frames per second [see Granatosky et al. (2016a) for information on data collection with GoPro cameras]. Only trials in which the animal was traveling in a straight path and not accelerating or decelerating (i.e. steady-state locomotion) throughout the climbing or walking trial, in which a full forelimb and/or hindlimb contacted the pole, and which exhibited a symmetric footfall sequence were retained for analysis. For all data, steady-state locomotion was determined by a combination of video, force and symmetry data following previously established methods (Byron et al., 2017; Granatosky et al., 2016a, 2018a; Hanna et al., 2017). For all trials, symmetry was determined using the methods of Cartmill et al. (2002), with a ±10 criterion such that the timing of opposite limb touchdown could vary between 40 and 60% of the stride cycle (50% indicates the timing of opposing limbs is exactly 1/2 of the cycle).

For single-camera recordings, calibration for distance was performed by using a known length in the view of the camera in the same plane as the animal was moving. For data collected with two cameras (i.e. all vertical climbing data and horizontal walking data on L. tardigradus, C. medius, N. pygmaeus, A. azarae and E. mongoz), calibration was accomplished with a three-dimensional (3D) cube with known corners. Speed was determined from this calibration as the average velocity of the animal over the view of the camera, by the 3D position of the head marker from the initial view in the cameras to the last view in the camera [see Hanna and Schmitt (2011), Hanna et al. (2008) and Hanna et al. (2017) for details]. For data collected from a single laterally positioned camera (i.e. horizontal walking data on D. madagascariensis, S. sciureus, M. fascicularis and A. nancymaae), the subject's speed was calculated by digitizing a point on the subject's head over the entire stride based on a known distance on the runway [see Granatosky (2018a), Granatosky et al. (2016a) and Granatosky et al. (2018) for details]. From video recordings, we considered diagonality, gait type, percentage of limb support, stride duration, forelimb and hindlimb duty factor, stride length, speed, and stride frequency (see Table S2 for information about variables). As some of these parameters are size dependent, we used the hindlimb length to calculate the dimensionless measures of stride length, speed, and stride frequency. Hindlimb length was collected from D. madagascariensis, S. sciureus, M. fascicularis and A. nancymaae while animals were anesthetized. Hindlimb length from the remaining species were measured from space-calibrated video recordings.

Due to low sample sizes, A. nancymaae and A. azarae were grouped together and analyzed as Aotus sp. We tested for statistically significant differences in the utilization of gait types and percentage of limb support between horizontal and vertical arboreal locomotion using a χ² test. We used a series of Mann–Whitney U-tests to determine whether diagonality, dimensionless speed, dimensionless stride length, and forelimb and hindlimb duty factor varied significantly between horizontal and vertical arboreal locomotion. Regression models were constructed to examine the impact of both stride frequency and stride length on speed using their dimensionless counterparts. The impact of each parameter was compared using a Fisher r-to-z transformation to test significance of the difference between two correlation coefficients. No statistical comparisons were conducted on untransformed speed, stride frequency, stride duration or stride length between the two conditions, but means and standard deviations are still reported as untransformed data.

This study used a series of regression analyses to assess the effects that variation in body masses between species may have on spatiotemporal gait variables (i.e. diagonality, forelimb and hindlimb duty factor, dimensionless stride length, speed, and stride frequency, and the correlation coefficients for dimensionless speed as a function of dimensionless stride length and stride frequency) during horizontal walking and vertical climbing. As some evidence suggests that branch diameter may influence gait patterns in primates (Stevens, 2008), an additional series of regression analyses were used to assess the effects that variation in relative branch diameter [i.e. branch diameter/body mass^1/3; see Stevens (2008)] may have on spatiotemporal gait variables during horizontal walking and vertical climbing. As phylogenetic history may be an additional and relevant confounding factor on the association between spatiotemporal gait variables and body mass and relative branch diameter, phylogenetically independent contrasts were used for all intraspecific regression analyses. The logarithmic transformation of all variables was used for all analyses. All phylogeny-based analyses were performed in R (Ver. 3.4.2) using phytools (Revell, 2012), ape (Paradis et al., 2004), nlme (https://CRAN.R-project.org/package=nlme) and geiger (Harmon et al., 2007). For all phylogenetic analyses, the phylogeny was constructed by pruning a recent super timetree (Hedges et al., 2015) to include only the species in our study (Fig. 1).

The sample consisted of 780 strides. Summary statistics for the variables of interest are presented in Table 1. Frequency data for the gait type and percentage of limb support are presented in Tables 2 and 3, respectively.

For almost all species, diagonality was significantly lower during vertical climbing compared with horizontal walking (P≤0.033; Table 1 and Figs 2 and 3). Only S. sciureus (hereafter referred to as Saimiri) demonstrated similar diagonality between the two locomotor modes (P=0.254). Accordingly, this drop in diagonality was associated with a significant increase in walking trots and LSDC gaits for all species except Saimiri (Table 2).

Patterns of forelimb and hindlimb duty factor between the two locomotor modes also revealed significant differences for most comparisons, although the direction of the effects were less consistent (Fig. 4). As with diagonality, there was no significant difference (P=0.452) in hindlimb duty factor observed in Saimiri between vertical climbing and horizontal walking. Forelimb duty factor was significantly (P≤0.041) lower during horizontal walking compared with vertical climbing for almost all species except L. tardigradus (hereafter referred to as Loris), N. pygmaeus (hereafter referred to as Nycticebus) and M. fascicularis (hereafter referred to as Macaca). For these taxa, no significant differences (P≥0.289) were observed. Hindlimb duty factor was significantly (P<0.001) lower for E. mongoz (hereafter referred to as Eulemur) during vertical climbing, but all other species demonstrated significantly lower hindlimb duty factors during horizontal walking (P≤0.044).

When comparing limb support patterns between horizontal walking and vertical climbing, it is clear that supporting the body with two contralaterally positioned limbs is the most common arrangement for both locomotor modes. This is followed by having three limbs in contact with the support, followed by supporting the body as pair of ipsilateral couplets. For half of the species [Loris; Nycticebus; A. nancymaae and A. azarae (hereafter referred to collectively as Aotus); and Eulemur], no significant differences were observed between the frequency of limb support patterns between horizontal walking and vertical climbing. For the other half of the sample, the proportion of stride that had all four limbs in contact with the support increased during vertical climbing. Additionally, C. medius (hereafter referred to as Cheirogaleus), Saimiri and D. madagascariensis (hereafter referred to as Daubentonia) exhibited an increased proportion of the stride that had all three limbs in contact with the support during vertical climbing. In contrast, Macaca demonstrated a decrease in three-limb contact and a rise in diagonal-couplet support when climbing vertical supports (Table 3).

For most species, dimensionless speed tended to be significantly (P≤0.034) lower during vertical climbing compared with horizontal walking (Fig. 5A). There was no significant difference (P=0.163) in dimensionless speed observed in Eulemur between vertical climbing and horizontal walking. Dimensionless stride length followed a similar pattern. For most species, dimensionless stride length tended to be significantly (P≤0.001) lower during vertical climbing compared with horizontal walking. The opposite pattern was observed in Saimiri (P<0.001). There was no significant difference (P≥0.133) in dimensionless stride length observed in Loris, Cheirogaleus or Daubentonia between vertical climbing and horizontal walking (Fig. 5B).

The construction of a stepwise regression model on the effects of dimensionless stride frequency and length on dimensionless speed (Figs S1 and S2) showed varying relationships across species. For all species except Aotus, both dimensionless stride frequency and length had a significant relationship on dimensionless speed. During horizontal walking in Aotus, dimensionless stride frequency had a significant relationship with dimensionless speed, but dimensionless stride length did not. However, during vertical climbing for this species, the opposite pattern was observed. Among the other species, four different solutions to modulation of speed during vertical movement were detected (see Tables S3 and S4 for correlation coefficients and P-values). Cheirogaleus, Nycticebus, Saimiri and Daubentonia exhibited a strong correlation between speed and stride frequency during climbing, whereas each of the other species exhibited differing relationships between speed, stride frequency and stride length.

Overall, the phylogenetic signals in the variables analyzed (i.e. diagonality, forelimb and hindlimb duty factor, dimensionless stride length, speed, and stride frequency, and correlation coefficients for dimensionless speed as a function of dimensionless stride length and stride frequency) were low in the sample species during both horizontal walking and vertical climbing, and were not significantly different (Blomberg's K≤1.04, P≥0.094; Pagel's λ≤0.838, P≥0.079). Phylogenetically independent contrasts revealed no significant relationships between body mass or relative support size and any of the variables analyzed, with the exception of diagonality, which had a significant (P=0.01) positive relationship (y=0.0752x+0.0008) with body mass during horizontal walking (see Table S5 for correlation coefficients and P-values).

The data presented here show that vertical climbing in primates does not demonstrate the same patterns observed during horizontal arboreal walking. Vertical climbing in primates is characterized by decreased diagonality and an increase in trots and lateral-sequence gaits, compared with horizontal walking. Furthermore, on vertical supports, the animals usually moved more slowly, with longer intervals of grasping the support. In the majority of the species, speed was regulated primarily by stride frequency, while stride length had a secondary, but often significant, effect. Beyond these variables, patterns were inconsistent across species, and phylogenetic and size-related patterns were not observed.

In almost all of the species analyzed, diagonality decreased during vertical climbing and, as a consequence, the proportion of walking trots and LSDC gaits increased. So what drives footfall sequence during vertical climbing? We believe that there are notable biomechanical advantages to switching from primarily DSDC gaits during horizontal walking to trots and LSDC gaits during vertical climbing. As originally modeled by Preuschoft (2002), on vertical substrates the animal must grasp the support either superior or inferior to the COM in at least one of its limbs, otherwise the animal is faced with a situation where it must use large muscular contractions to counteract the rotational torques acting on the COM due to gravity (Cartmill, 1985; Preuschoft, 2002). During DSDC gaits, positioning of the limbs means that there is a portion of the stride where the animal is supporting body mass on two feet placed close together near the COM (Cartmill et al., 2002, 2007; Shapiro and Raichlen, 2007). There are a number of strategies that animals can use to prevent this situation, including increasing duty factor so that more limbs are in contact with the support at any one time (i.e. three- and four-limb support), or by adopting LSDC gaits, which effectively eliminates moments where body support is solely dependent on two feet placed close together near the COM (Cartmill et al., 2007; Shapiro and Raichlen, 2007). Our data suggest that primates employ both strategies during vertical climbing.

The continued presence of DSDC gaits (albeit with lower diagonality and thus closer to walking trots) may be explained by a simple retention of neuromuscular gait parameters on both horizontal and vertical substrates (Schoonaert et al., 2016; Vilensky et al., 1994). However, the ability of primates to switch easily between gait types depending on different conditions (Granatosky et al., 2016a; Isler, 2004; Nyakatura and Heymann, 2010; Stevens, 2008) suggests that this is not a complete answer. Understanding the continued use of DSDC gaits during vertical climbing remains a wide-open question that deserves further exploration.

As mentioned above, almost all species in this study demonstrated lower speeds during vertical climbing compared with horizontal walking. This follows predictions by Schmidt and Fischer (2010) and Karantanis and colleagues (2015, 2016, 2017a,b,c) that, in arboreal conditions, animals may adopt slower speeds as a way to increase static stability. More interesting, however, are the strategies used by vertically climbing primates to regulate speed. For primates, increasing speed by increasing stride length is possibly safer in an arboreal setting, allowing for a longer reach of the forelimbs and reducing involuntary branch sway (Cartmill, 1985; Demes et al., 1994; Larson et al., 2000, 2001). Additionally, longer stride lengths also have energetic benefits that may be especially important during vertical climbing for large-bodied primates (Hanna et al., 2008; Karantanis et al., 2016). In our study, only Saimiri had significantly longer stride lengths during vertical climbing compared with horizontal walking. For the rest of the species, stride length during vertical climbing was either lower than, or the same as, during horizontal walking. Accordingly, almost all of the species in our study primarily regulate speed by increasing stride frequency rather than stride length (although both factors are important contributors in almost all species). Similar findings have been reported in geckos (Zaaf et al., 2001), cotton-top tamarins (Nyakatura et al., 2008), bonobos (Schoonaert et al., 2016), acacia rats (Karantanis et al., 2017c) and feathertail gliders (Karantanis et al., 2015). The regulation of speed by stride frequency versus stride length is thought to reduce body oscillations. During horizontal locomotion, dorsoventral body oscillations are directed perpendicular to the branch and in the vertical plane (Cartmill, 1985; Preuschoft, 2002). There are of course horizontal (anterior–posterior) and mediolateral oscillations of the COM as well, but they are both smaller. During vertical climbing, much of the oscillations are parallel to the substrate (Cartmill, 1985; Preuschoft, 2002). On thin, arboreal substrates, body oscillations may cause substrate oscillations. Therefore, reducing body oscillations during arboreal locomotion (horizontally or vertically oriented) may increase arboreal stability (Delciellos and Vieira, 2007; Karantanis et al., 2015, 2017c; Strang and Steudel, 1990). With this in mind, it is possible that the need to maximize stride length during vertical climbing may be less important than previously thought.

One of the major goals of this study was to test the hypothesis that the unusual spatiotemporal gait characteristics observed during horizontal walking in primates are present because of the effect of using these characteristics during vertical climbing. This argument is based on previous studies showing an increase in diagonal-sequence gaits on inclined supports (Prost and Sussman, 1969; Vilensky et al., 1994), and on recent suggestions by Hanna et al. (2017) that limb-loading patterns that characterize primates may have arisen in association with vertical climbing rather than horizontal locomotion. However, in regards to spatiotemporal patterns, the data presented here show that many of the gait characteristics observed during vertical climbing in primates are quite similar to the patterns observed during quadrupedal locomotion in non-primate mammals (Demes et al., 1994; Hildebrand, 1967; Muybridge, 1887; Rollinson and Martin, 1981; Schmitt and Lemelin, 2002). During vertical climbing, primates appear to select gaits that maximize security and stability. LSDC gaits, increased duty factor, decreased speed, and the regulation of speed through stride frequency are all ways that an animal can increase stability on a vertical support (Cartmill, 1985; Karantanis et al., 2015, 2017c; Shapiro and Raichlen, 2007; Strang and Steudel, 1990). Vilensky and Larson (1989) emphasized behavioral flexibility – the ability to alter gait characteristics depending on environmental challenges – as an important feature associated with primate origins. This flexibility may have been essential to invasion and occupation of an arboreal, fine-branch niche by the earliest primates as it allowed successful movement on both vertical and horizontal substrates (Cartmill, 1992; Jenkins, 1974). Recent evidence, however, suggests that movement on arboreal substrates requires all mammals, regardless of taxonomic affiliation, to demonstrate high behavioral flexibility (Granatosky, 2018b).

These data reject the idea that the suite of spatiotemporal gait features observed in primates during horizontal walking are in some way linked to selective pressures associated with mechanical requirements of vertical climbing. Instead, the use of different gait characteristics during climbing and horizontal movement emphasizes the innate flexibility of arboreal mammals to adjust spatiotemporal variables to meet different substrate needs. Although these data do not definitively address the selective pressures leading to the evolution of DSDC gaits, functional differentiation of the limbs, or large limb excursions, they certainly provide for a broader discussion of the implications of these gaits in different environments. In short, primates do not simply use the same motor program for all surfaces. They differ in spatiotemporal gait characteristics on the ground, on horizontal supports and on vertical supports. The data presented here, in conjunction with previous studies on primates (Granatosky, 2018b; Granatosky et al., 2016a; Isler, 2004; Nyakatura and Heymann, 2010; Stevens, 2008) and non-primate mammals (Granatosky, 2018b; Karantanis et al., 2015, 2017a,b,c; Shapiro and Young, 2010), highlight the importance of behavioral flexibility for mammals to effectively traverse a complex, 3D arboreal environment.

We thank Erin Ehmke, David Brewer and Meg Dye at the Duke Lemur Center for all their help with animal training and data collection. Without their help, we would not have been able to complete this study. We thank Andrew A. Biewener and the two anonymous reviewers for their comments and inspiration that improved the overall quality of this work.