Tail loss and narrow surfaces decrease locomotor stability in the arboreal green anole lizard (Anolis carolinensis)

The post-anal tail is a synapomorphy of the phylum Chordata, yet is highly divergent in both form and function. For example, fish use an expanded tail for thrust generation during swimming, whereas other vertebrates such as opossums and chameleons use a prehensile tail for grasping narrow surfaces whilst climbing. Tails may be used to attract mates (e.g. swordtail fish and peacocks), as they are honest indicators of health in some taxa, and thereby are subjected to sexual selection (Andersson, 1994; Loyau et al., 2005). The tail may also be used for fat storage, facilitating survival during periods of unreliable food availability. For example, in a viviparous skink, Niveoscincus metallicus, caudal fat bodies comprise 55–78% of the total fat reserves in the body (Chapple and Swain, 2002).

Recently, tails have been shown to play an important role in dynamic stabilization during climbing (Jusufi et al., 2008), falling (Jusufi et al., 2008, 2010, 2011) and jumping (Gillis et al., 2009; Kuo et al., 2012; Libby et al., 2012) in geckos and anole lizards, and arboreal turning and running in primates (Larson and Stern, 2006). During vertical climbing, a gecko can counteract a slip-induced overturning moment by pushing its tail against the wall, much as one uses a bicycle stand to prevent a bicycle from falling over (Jusufi et al., 2008). When falling, a gecko can use the inertial properties of the tail to reorient the body and change how and where they land (Jusufi et al., 2008, 2010, 2011). Likewise, when jumping, lizards swing their tail to control body pitch angle (Higham et al., 2001; Gillis et al., 2009; Kuo et al., 2012; Libby et al., 2012). Following tail autotomy jump performance is compromised as lizards are unable to control angular rotation of their body, commonly landing on their back (Gillis et al., 2009). Using a robot to test this idea further, Libby and colleagues (2012) discovered that a body with an actively controlled tail experienced less rotation than one with a passive tail, a compliant tail, or no tail – which exhibited the greatest angular rotation.

Studies documenting tail function during running on narrow surfaces is much more limited. However, it is accepted that the large rotational inertia exerted by a heavy tail extended behind the body can limit turning performance (Carrier et al., 2001); therefore, active control of tail posture would be important to increase maneuverability and stability (Carrier et al., 2001; Briggs et al., 2012). Arboreal, tailed primates use a ‘tail-whip mechanism’, involving tail rotation in the direction of imbalance, to restore stability when they lose balance whilst moving on narrow branches (Larson and Stern, 2006). Vervets and long-tailed macaques have also been observed to use their tails as counterweights whilst walking, swaying the tail from side to side (Larson and Stern, 2006).

Interestingly, despite all these important physiological and locomotor functions, most lizard species use voluntary tail loss – or tail autotomy – as a mechanism for predator evasion (e.g. Arnold, 1984; Zani, 1996). During tail autotomy, the tail breaks along a region of weakness, either along a fracture plane located in the middle of a vertebral segment (Bellairs and Bryant, 1985) or between the vertebrae (Arnold, 1984). Longitudinal tail muscles then contract, limiting the amount of blood loss (Bellairs and Bryant, 1985).

The ease with which a lizard will lose its tail should be influenced by its costs and benefits (Vitt et al., 1977), phylogenetic history (Arnold, 1984), and caudal anatomy (Fisher et al., 2012; Ritzman et al., 2012). The length of the tail that can be lost has been correlated with caudifemoralis longus (CFL) muscle length, tail length and location of autotomy planes (Arnold, 1984; Fisher et al., 2012; Ritzman et al., 2012). Furthermore, since this muscle aids in femoral retraction (Snyder, 1952, 1954; Russell and Bauer, 1992; Nelson and Jayne, 2001), it has been hypothesized that fast lizards should have a longer CFL, and thus reduced ability for autotomy (Russell and Bauer, 1992).

Tail autotomy exerts some substantial costs (Vitt et al., 1977; Fox and McCoy, 2000), although how it actually affects fitness is debatable. For example, tail autotomy may lead to decreased short- or long-term survival (Fox and McCoy, 2000) as a result of decreased running performance (Martin and Avery, 1998), changes to escape behavior (Downes and Shine, 2001), increased energetic costs to regenerate the tail (Naya et al., 2007), and decreased reproductive fitness (Dial and Fitzpatrick, 1981). Yet even some of these findings may be contradictory (e.g. McElroy and Bergmann, 2013). Whereas some lizard species ran significantly faster on flat surfaces after losing their tail than with their original tail (Daniels, 1983; Brown et al., 1995), some exhibited decreased arboreal running performance (Brown et al., 1995). In other taxa, tail autotomy either did not affect overall or maximum running speed (Lu et al., 2010; Jagnandan et al., 2014), or it decreased running performance (Martin and Avery, 1998; Goodman, 2006).

These contradictory results make it difficult to determine how tail loss affects running performance and stability (McElroy and Bergmann, 2013; Jagnandan et al., 2014). Arboreal lizards, specialized for running on inherently challenging, narrow surfaces, and that are known to actively use their tail for stabilization, would serve as an excellent model system with which to address this issue. As a result, the goal of this study was to determine how tail loss affects running kinematics and performance in the green anole lizard, Anolis carolinensis. The green anole lizard is a diurnally active, arboreal lizard that autotomizes its tail in response to predation events. In their arboreal environment, these lizards move along narrow, springy surfaces (i.e. perches) that challenge locomotor stability. In order to maintain straight-line trajectories or to minimize turning angle, anoles frequently jump across a turn and use their tails for rotational control (Higham et al., 2001). In order to elicit the greatest potential kinematic compensations due to destabilizing perturbations, I ran the lizards on a flat track as well as on narrow perches that would limit the width of their base of support (Fig. 1B) whilst intact and following tail autotomy.

Based on previous findings for tail use during locomotion, I hypothesized that smaller diameter perches and tail loss would decrease the running stability of lizards, and that their locomotor kinematics would reflect compensations for the greater instability. Focusing on stable, constant-velocity trials, I expected that lizards would compensate for the decreased stability by running more slowly, decreasing mediolateral movements of the body (Ting et al., 1994; Hof et al., 2005), increasing foot contact time with the substrate (Schmitt, 1999), and assuming a more crouched posture (Schmitt, 1999). I expected these effects to be most pronounced following tail loss, on the narrowest surface.

Five adult male green anole lizards [Anolis carolinensis (Voigt 1832); mass: 5.0±0.5 g; snout–vent length (SVL): 6.2±0.7 cm; means±s.d.] were purchased from a lizard wholesale distributor in Louisiana and maintained on a 12 h:12 h, high-UVB (ReptiSun 10.0, Zoo Med Laboratories, San Luis Obispo, CA, USA) light:dark cycle at 28–31°C and automatically misted three times daily. Lizards were fed crickets or cockroaches three times per week, dusted in vitamin (ReptiVite, Zoo Med Laboratories) and calcium powder (Rep-Cal, Los Gatos, CA, USA). All experiments were performed in accordance with protocols approved by Temple University's Institutional Animal Care and Use Committee (ACUP no. 3487 and 4197).

All lizards were run along four different surfaces with intact and autotomized tails. Surfaces consisted of 1.2 m long dowels of three diameters (9.5, 15.9 and 19.0 mm) and a 2.4 m long flat surface. On average, lizards ran continuously along these surfaces for 0.3 m, with 0.5 m as the longest recorded sprint, suggesting that differences in track length did not limit running distance. I angled the dowels (‘perches’) at 6 deg to the horizontal to encourage the lizards to run along their length because lizards tended to jump, rather than run, when positioned horizontally. Flat surface trials were conducted on a 0 deg incline. Although these differences in incline could potentially confound observed kinematic differences, previous studies in anole lizards and other animals have shown that surface diameter tends to have a greater impact on running kinematics than incline (Spezzano and Jayne, 2004; Foster and Higham, 2012, 2014). All dowels were covered in nylon window screen material and the flat surface was covered in 200 grit sandpaper to provide traction.

All lizards were run a minimum of ten times on each surface to obtain baseline kinematics with their intact tail before 75% of their tail was autotomized. Tail autotomy was performed by firmly grasping the tail at 25% tail length from the tail base (immediately posterior to point T2; Fig. 1A) until the lizard voluntarily initiated autotomy. Following tail removal, lizards were run a minimum of ten times within 2 days of tail autotomy.

For all trials, lizards were marked with 29 reflective markers that were either 1.6 mm diameter reflective spheres or 1 mm square pieces of reflective tape (Fig. 1A). Trials were recorded with a synchronized six-camera infrared, high-speed video system (Motion Analysis Corporation, Santa Rosa, CA, USA) filming at 500 frames s^–1. The associated software (Cortex version 2.5.0.1160) auto-tracked the points and reconstructed their three-dimensional (3D) locations with an accuracy of approximately 0.2 mm.

Three trials per individual on each surface for each treatment (pre- and post-autotomy) were selected for detailed analyses based on the following criteria: (1) steady, continuous run (i.e. the average stride-to-stride velocity varied by less than 25%); and (2) a minimum of three complete strides. If more than three trials met both of these criteria, then I selected the three fastest trials with the most strides. As a result, all trials included in this study represent the fastest, stable, steady recorded runs, and excludes trials in which the lizard was obviously correcting for instability. A total of 120 trials consisting of 508 strides are included in the analyses.

Raw point positions were imported into MATLAB R2013a (MathWorks, Natick, MA, USA) and smoothed with a quintic spline (spaps function; tolerance=0.001) before analysis (Fig. S1). All 3D coordinates were rotated such that the lizard ran parallel to the positive x-axis. The +z-axis pointed vertically upwards, and the +y-axis pointed to the left of the running lizard, following right-hand-rule conventions.

To assess effects of autotomy and surface breadth on running stability, I quantified variables representing running speed, and timing characteristics and positions of limb and body points. For the timing characteristics, I measured stride frequency, duration and duty factor. Duty factor was defined as the proportion of a full stride that the foot was in stance. Since lizards ran quadrupedally, I calculated these stride parameters for each foot and averaged the values for the left and right sides to obtain separate trial means for the front and hindfeet.

Position variables included stride length, stance width, mediolateral axial body and tail excursions, and effective limb lengths (eLL) for the forelimb and hindlimb at midstance. Average stride length per trial was calculated as the 3D distance travelled from the first to final foot down by the same foot, divided by the number of strides. Because stride lengths did not differ among the feet, I arbitrarily selected the hindlimb with the most strides for obtaining the average stride length for each trial. Stance width was calculated as the y-distance between consecutive footfalls of contralateral feet, with separate values for forefeet and hindfeet. Mediolateral excursions were calculated as the difference between the maximum +y and −y excursions for each stride, for each of the axial body points (Fig. 1A). These values were then averaged to obtain a single quantity for each point along the body, for each trial. Effective forelimb length (eLL_fore) was calculated as the 3D distance from the pectoral girdle midpoint to the wrist for both forelimbs. As a result, greater eLL_fore values indicated more extended shoulder and elbow joints. Effective hindlimb length (eLL_hind) was calculated as the 3D distance from the pelvic girdle midpoint to the ankle point for both hindlimbs. Greater eLL_hind values represented more abducted and extended hip and knee joints. All eLL values were averaged for each limb pair across all strides within each trial to obtain a single value for eLL_fore and eLL_hind that was representative for the trial.

Traditionally, a body is considered to be stable and in-balance if the vertical projection of the center of mass (CM) falls within a base of support (Ting et al., 1994). This condition for maintaining balance requires the body to be in stasis. However, the velocity of the center of mass plays an important role in determining an individual's ability to control balance (Pai and Patton, 1997; Hof et al., 2005); therefore, this definition provides an inaccurate measure of stability under dynamic situations (Pai and Patton, 1997). Although Hof and colleagues (2005) developed a model that incorporates velocity into margin of stability calculations, this has only been validated for walking gaits, and not for running gaits as in this study. As a result, I present here calculations for the static margin of stability, only, as a metric for instantaneous stability at any point in time, with these caveats taken into account.

For this study, the static stability margin was calculated according to the definition by Ting and colleagues (1994). For each frame, the feet in stance were identified as forming the base of support. The static stability margin is the minimum two-dimensional distance of the vertical projection of the CM to each edge of the base of support. If the CM fell within the base of support, the static stability margin was assigned a value greater than zero. If the CM falls outside the base of support, the system was determined to be statically unstable, and the static stability margin was assigned a negative value. What this means is that if the lizard were to suddenly stop when statically unstable, it would fall over. However, when running at high speeds, assuming the lizard continued along a predetermined trajectory, a negative static stability margin indicated the lizard was relying on dynamic stability to continue undisturbed forwards locomotion (Koditschek et al., 2004).

Seeing that the tail comprises 11.1% of total body weight (Legreneur et al., 2012), removal would result in an anterior shift of the center of mass by 5.6% SVL (0.32±0.0029 mm for this study). As a result, all stability margins calculated for post-autotomy treatments included a slight anterior shift of the center of mass to account for repositioning due to tail loss. All calculated values were averaged to obtain a single value for each trial representing the overall stability of an individual run.

Multiple regressions were run to test for effects of speed and size on each of the analyzed kinematic variables. Where a significant relationship was obtained, the residuals of those variables were used for further analysis, to control for speed- and size-related differences. All variables were then tested for effect of surface and tail autotomy using a two-way, fixed-effect (model I) ANOVA, blocked by individual as a random variable (Zar, 1999). Post-hoc pairwise comparisons, when appropriate, were completed with a Tukey's honest significant difference (HSD) test. Comparisons of frontlimb and hindlimb variables were completed with one-tailed paired t-tests. All statistics were performed in JMP 10.0.0 software (SAS, Cary, NC, USA). Unless otherwise stated, all means are presented as means±s.e.m.

In this study, I explored how surface breadth and tail loss affected lizard running performance. I determined whether they compensated for tail loss by examining whole-body kinematics, limb kinematics, footfall parameters and tail movements on four different surface widths (9.5, 15.9 and 19.0 mm, and flat) before and after autotomy and loss of 75% of the tail. I expected that running performance would decrease on narrower surfaces and following tail loss, and that locomotor kinematics would exhibit signs of destabilization and compensation, such as a more crouched posture, body stiffening, increased tail excursions and shorter but quicker steps with increased duty factor.

To correct for effects of size and running speed, a multiple regression was run with SVL and running speed as predictor variables (Table 1; Table S1). All multiple regressions were significant at least for speed, so residuals were extracted as the size- and speed-corrected values. These residuals were used for the remaining analyses exploring surface diameter and tail autotomy effects. Running speed was the sole exception in that ‘relative running speed’ was calculated by dividing the running speed by SVL, resulting in units of body lengths per second (BL s⁻¹).

Lizard body size had no significant impact on forefoot stance width, although hindfoot stance width decreased slightly (β=−0.228) with increased size (Table 1). This demonstrates that for the animal sizes studied here, forefoot stance width was limited more by perch diameter and running speed than SVL. As expected, decreased perch diameter tended to force foot positions into narrower stance widths (Fig. 1B, Fig. 2A,B; Table 2), although no significant difference was detected between the 15.9 and 19.0 mm diameter perches (Fig. 2A,B; Table 2). On all surfaces, hindfoot stance width was greater than forefoot stance width (paired t-tests, P<0.0001). Forefoot and hindfoot stance width increased significantly with running speed (Table 1).

Stability margin (SM) increased slightly with speed, but not SVL (Table 1). The stability impacts of the narrower surfaces were reflected in the static stability margin calculations (Fig. 2C; Table 2). SM was significantly greater on the flat surface, than on the perches (Fig. 2C; Table 2), but did not differ significantly among the perches despite narrower stance widths on smaller diameter perches.

On average, lizards ran faster on broader surfaces (Fig. 3A; 9.5 mm: 6.1±0.5 BL s⁻¹; 15.9 mm: 6.5±0.4 BL s⁻¹; 19.0 mm: 8.4±0.8 BL s⁻¹; flat: 7.7±0.7 BL s⁻¹) in part due to longer stride lengths (Fig. 3B), which compensated for the lower stride frequencies (Fig. 3C). These sprint speeds were comparable to average speeds measured in other studies on green anole lizards (Foster and Higham, 2014), but were lower than the maximum sprint speeds reported elsewhere (Losos and Irschick, 1996; Vanhooydonck et al., 2006; Husak et al., 2015; Sathe and Husak, 2015). Lizards ran with a diagonal gait on all surfaces (Fig. 4A,B). Forelimb duty factor was greatest on the perches, whereas hindlimb duty factor was greatest on the 9.5 mm perch and flat surfaces (Fig. 4C,D, Table 2). Animals also ran with a more upright posture on flat surfaces than they did on the perches, as shown by significantly greater effective limb lengths for both the forelimbs and hindlimbs on flat surfaces (Table 2, Fig. 3D).

Figure 5 shows that mediolateral excursions of the axial body points increased caudally when running on all surfaces. Side-to-side movements of the head and body points anterior to the pelvis did not differ with perch diameter, but were significantly greater on the flat surface (Table S2). In contrast, mediolateral excursions of the pelvis and tail points increased distally on the narrowest surface such that lateral excursions on the narrowest surface were not statistically different from those on the flat surface, with lateral excursions significantly lower on the intermediate perch diameters. In contrast, the distal-most tail points (T6–8, Fig. 1A) varied so much in their excursions that they were not significantly different among all surfaces analyzed.

Contrary to expectations, tail loss had no measurable impact on calculated static stability margins (Fig. 2C, Table 2). However, it did have a significant impact on all other metrics for foot placement and timing characteristics (Table 2).

Stance width decreased where there was a significant effect due to tail loss (Table 2). Forelimb stance width decreased significantly on only the 19.0 mm surface (Fig. 2A), although there appeared to be a non-significant trend towards narrower forelimb stance width on all perches. Hindlimb stance width narrowed following tail loss on the 15.9 mm and flat surfaces (Fig. 2B). In all cases, the hindlimbs remained more sprawled than the forelimbs (paired t-test, P<0.0001).

Lizards ran faster following tail loss on the narrowest two perches only (Fig. 3A). Tail loss had no statistical impact on running speeds on the 19.0 mm perch or the flat surface, although the mean values tended to be higher. Higher running speeds were associated with higher stride frequency (Fig. 3C) and decreased stride length (Fig. 3B). Whereas hindlimbs tended to become more crouched following autotomy, forelimbs tended to assume a more erect posture (Fig. 3D). Tail loss was also associated with lower forelimb duty factor on the 9.5 mm perch, but a higher duty factor on the flat surface (Fig. 4C), and greater hindlimb duty factor on all but the narrowest perch (Fig. 4D).

Tail autotomy had no impact on mediolateral movements of the body (Table S2). However, on all perches, the remaining tail points (T1and T2) increased their lateral excursions following tail loss. On the flat surface, tail loss had no detectable impact on lateral excursions (Fig. 5).

Previous studies point towards the importance of the tail for stabilization during falling (Jusufi et al., 2008, 2010, 2011) and jumping (Gillis et al., 2009; Kuo et al., 2012; Libby et al., 2012) in lizards, and also when running in arboreal primates (Larson and Stern, 2006) and robots (Briggs et al., 2012). As a result, it is expected that tail loss could negatively impact an animal's locomotor performance; yet studies are conflicted in their conclusions.

In this study, I examined the consequences of tail loss on the locomotor performance and kinematics of an arboreal lizard species, the green anole lizard (Anolis carolinensis). Anole lizards are an interesting group with which to explore the interplay of tail loss and locomotor performance because they are a diverse genus, and are known to exhibit clear morphological patterns that correlate with their habitat choice, grouping them into ecomorphs (Williams, 1969; Moermond, 1979; Losos, 1990, 1992; Losos et al., 1998; Beuttell and Losos, 1999). For example, it is widely believed that anole lizards that tend to move dynamically on unstable surfaces (e.g. grass bush species) have slim bodies and disproportionately long tails. In contrast, slow locomotors, such as twig anoles, tend to have dorsoventrally compressed bodies and short, prehensile tails (Moermond, 1979; Irschick and Losos, 1998). In other words, tail length has often been associated with surface breadth and locomotor style in lizards (Ballinger, 1973; Kohlsdorf et al., 2001) and other arboreal animals (Larson and Stern, 2006).

Green anole lizards are categorized as crown or trunk-crown ecomorphs, and are frequently found perched on the distal-most tips of branches. Their escape behavior upon approach by a researcher is predictable, and involves running towards the base of the tree or bush, and then dropping off the branch and into the leaf litter, if they continue to be pursued (author's personal observation). Upon capture, they are slow to autotomize their tail. In this study, several lizards did not autotomize their tail when grasped for more than 15 min, suggesting greater value of the tail for general survival (Vitt et al., 1977).

Because tails have been credited with playing an important role in balance (Larson and Stern, 2006; Jusufi et al., 2008; Gillis et al., 2009; Kuo et al., 2012; Libby et al., 2012), I expected that tail loss would be associated with clear decrements in locomotor performance, and would have negative consequences on locomotor stability. By running lizards along perches of different diameters, I expected to observe the greatest performance decrement and kinematic differences when running on the narrowest perch without a tail, in comparison with running with an intact tail on a flat surface. Although these trends were consistent for some variables (e.g. running speed and most stride parameters), other kinematic parameters were most similar between the narrowest and widest surfaces, probably due to stability and performance trade-offs.

An assumption of this study was that narrow perches would force a narrower stance width, and thereby decrease the stability margin of a running lizard. Surprisingly, although lizards did present with a narrower stance width on the smallest (9.5 mm) versus largest (19.0 mm) diameter perches (Fig. 2A,B), their static stability margins did not differ, increasing significantly only on the flat surface (Fig. 2C). Seeing that stability margins are calculated as a combination of CM position relative to foot placement, the similarity in stability margins among the perches is likely to be a reflection of sufficient stability compensation via other mechanisms on narrow perches.

The static stability margin is defined as the minimum horizontal distance from the vertical projection of the CM to edges of the base of support (Ting et al., 1994). For this study, I designated the mid-body point as an approximation of the lizard's CM location when the lizard is dangled with its legs positioned at right angles to the body and tail held straight (Fig. 1A). Although it is reasonable to expect that the CM position would shift during running as a result of lateral body and tail undulations (Fig. 5) and cycling of the limbs, based on results elsewhere, I expected that CM movements in the horizontal plane would be small. Force data collected on the similarly sized western skink (Eumeces skiltonianus; SVL: 5.2 cm) running at a similar speed as that observed here (7.7 BL s⁻¹) showed that the CM shifted 0.3 mm in the fore–aft directions, and 0.2 mm mediolaterally (Farley and Ko, 1997). These movements are small enough to be contained within the error margins of 3D kinematic reconstructions in this study (see Materials and methods). Impacts of tail loss on an anterior shift in the CM position were taken into account for post-autotomy stability margin calculations (see Materials and methods). In this study, whole-body kinematics show that excursions of the mid-body point and those cranial to it do not differ among the perches and between tail conditions (Fig. 5), with all axial changes due to perch diameter limited to tail movement. This suggests that increased mediolateral tail excursions were necessary to compensate for instabilities imposed by decreased perch diameter.

It remains notable, however, that the results did indicate narrower forefoot and hindfoot stance widths on the 9.5 mm surface than on all other perches (Fig. 2A,B) but no differences among calculated stability margins (Fig. 2C). This suggests that although lizards assumed a wider stance width on the 15.9 and 19.0 mm surfaces, these differences were slight and the step-to-step variability in foot placement combined with increased tail movements under less stable conditions were able to sufficiently compensate for differences in locomotor stability.

Surface breadth had a significant effect on all variables tested. As expected, lizards ran faster on wider surfaces. Consistent with other studies on arboreal quadrupeds (Cartmill, 1985; Larson and Stern, 2006; Lammers, 2009), forelimb stance width was always narrower than hindlimb stance width, indicating that the hindfeet tended to be placed lower and more laterally on the perches. On all perches, lizards assumed a more crouched posture (Fig. 3D) and increased forelimb duty factor (Fig. 4C), relative to that on the flat surface, probably to compensate for greater locomotor instability (Fig. 2C). A similar postural modification has been reported among arboreal primates, which adopt a compliant walking gait when moving along narrow surfaces. This bent-limb posture lowers CM height, decreasing the potential rolling moment about the perch. Additionally, by increasing elbow flexion, primates increase stride length and contact time whilst decreasing peak substrate reaction forces and bone loading (Schmitt, 1999).

Hindlimb duty factor also differed among the surfaces, but followed a different trend from the forelimbs, increasing on the narrowest (9.5 mm) perch and flat surfaces, and decreasing on the intermediate diameter perches (Fig. 4D). This difference among limb pairs is not surprising because of differential function of the forelimbs and hindlimbs among multi-legged runners (Full and Tu, 1991; Lee et al., 1999; Biewener, 2003; Chen et al., 2006). Whereas the forelimbs usually produce decelerative forces during steady-state running, the hindlimbs serve a primarily propulsive role (Cavagna et al., 1977; Heglund et al., 1982; Autumn et al., 2006; Chen et al., 2006). This differential limb function aids mediolateral stabilization and maneuvering among sprawled runners (Jindrich and Full, 1999), and pitch stabilization during speed changes among parasagittal runners (Lee et al., 1999). The opposing impacts of surface breadth on hindlimb duty factor suggest an associated change in locomotor function with whole body destabilization. On the flat surface, where the lizards were most stable, the hindlimbs still served a largely propulsive role, with greater relative contact periods enabling more force to be exerted against the ground for a greater duration of a step. This was also associated with the higher running speeds observed on the flat surface (Fig. 3A). In contrast, when on the narrowest surface, running speed was low and the greater hindlimb contact time probably facilitated locomotor stabilization.

Although mediolateral excursions of the axial body points anterior to the pelvis were similar on all perches, excursions of the pelvis and tail points were greatest on the narrowest 9.5 mm perch and the flat surface. On the narrowest surface, these larger mediolateral movements probably acted to stabilize the CM. In particular, caudal to the second tail point (T2), mediolateral excursions on the 9.5 mm surface matched or exceeded that on the flat surface. Most caudally, variability in stride-to-stride tail motions was large enough to obscure statistical significance. However, the range of tail tip motion on the 9.5 mm surface exceeded that on all other surfaces (Fig. 5).

On the flat surface, large tail excursions were probably a side-effect due to anatomical constraint, rather than being necessary for locomotor stabilization, as was observed on the narrow perches. The caudofemoralis muscle originates from the tail axial skeleton and inserts on the femur. In addition to playing an important role in limb retraction during high speed running (Reilly, 1995; Nelson and Jayne, 2001), muscle activation also causes the tail to move towards the side of muscle contraction (Gatesy, 1990; Reilly, 1995). On flat surfaces, lizards ran faster (Fig. 3A) and took longer strides (Fig. 3B), requiring greater retraction and protraction of the limbs and rotation of the pectoral and pelvic girdles. As a result, the longer stride lengths on the flat surface were likely to cause increased lateral tail excursions, due to the muscular linkage between the tail and femur.

These data showed a statistically significant increase in running speed following tail loss on the two narrowest surfaces, and a non-significant increasing trend on the remaining surfaces tested. These findings contradict the only other study, of which I am aware, which report decreased sprint performance of green anole lizards following tail loss (McElroy and Bergmann, 2013). This difference may be attributed to several factors that are not mutually exclusive: (1) I ran the lizards along a horizontal, flat trackway and perches inclined at a shallow 6 deg angle, whereas McElroy and Bergmann ran their lizards up a flat trackway placed at a much steeper 30 deg incline. It is possible that steep inclines have a different impact on running performance following tail loss, as joint function is known to change during incline running in other vertebrates (Gabaldón et al., 2004; Roberts and Belliveau, 2005; Lee et al., 2008). (2) I recorded all post-autotomy trials within 1–2 days of initiating autotomy, whereas McElroy and Bergmann recorded their trials 14 days post-autotomy. Following autotomy, it is known that body condition declines as resources are diverted towards regeneration (Dial and Fitzpatrick, 1981; Maginnis, 2006; Naya et al., 2007; Wrinn and Uetz, 2007; Fleming et al., 2009; Jagnandan et al., 2014) and could decrease maximum sprint speeds (Bateman and Fleming, 2009; Fleming et al., 2009). In this study, the autotomized tail represented 6–10% of the intact body weight, which could have enabled greater running speeds immediately after tail loss because lizards had less weight to carry. Because I ran the lizards soon after tail autotomy, this avoided the potentially confounding factor of changes in body condition. Likewise, it is possible that lizards were more motivated to run rapidly shortly following tail autotomy because it is often associated with extreme life or death situations. (3) Finally, this study focused on average speeds achieved during the fastest steady runs, whereas McElroy and Bergmann reported maximum sprint speeds in each trial. As a result, their reported running speeds are substantially greater than those recorded in this study, and represent what lizards can achieve during burst sprints involving brief, high accelerations. This is unsurprising, as it is well known that intermittent locomotion can increase peak locomotor performance, and is a common strategy employed during regular, undisturbed locomotion in nature (Weinstein and Full, 1998; Gleeson and Hancock, 2001; Weinstein, 2001). The findings in this study therefore address locomotor parameters that represent the performance capacity of sustained locomotion in these lizards.

Surprisingly, while I expected that tail loss would induce instability and force lizards to seek a wider stance width, the results did not support this expectation. Where differences were detected, lizards narrowed rather than widened their stance width (Fig. 2A,B). Furthermore, I found that tail loss did not impact stability margin (Fig. 2C). It is possible that this is an artifact of having focused the analyses on steady, balanced running only, during which the stability margin can only fluctuate within a narrow margin and still maintain stable locomotion. An alternative explanation is that lizards are able to sufficiently adjust their kinematics following tail loss to compensate for the imposed destabilization. Observed changes in kinematic parameters such as a more erect forelimb and a more crouched hindlimb posture, increased hindlimb duty factor, greater stride frequency, decreased stride length and greater mediolateral excursion of the residual tail tip (T2), all support the latter conclusion.

The changes in limb posture suggest weight transfer towards the hindlimbs (Lee et al., 2004; Krause and Fischer, 2013), although why that might occur as a consequence of tail loss remains unclear. Electromyography could reveal whether such a postural shift permits the lizard to engage larger muscle groups for stabilization or facilitating greater running speeds.

Mediolateral excursions of the residual tail tip (T2) increased following tail autotomy. Because stride length did not increase following tail loss, it is unlikely that greater limb excursions were responsible for increased tail movements. Instead, increased tail excursions were likely to be correcting for minor stride-to-stride instabilities. Comparing these data with unstable trials in which lizards stumbled could yield additional insight into this question.

This study examined how lizard running kinematics changed following tail autotomy and when subjected to a range of surfaces that would challenge their running stability. Results suggest that whilst both manipulations affect stability, surface breadth manipulations do so by narrowing the base of support, and therefore the stability margins. How tail autotomy affects locomotor stability is more difficult to assess and is probably more dependent on dynamic CM movements rather than the static calculations used here. Kinematic adjustments largely manifested themselves in longer stance periods per stride, more crouched hindlimb postures and increased tail excursions. These combined results suggest that active adjustments are necessary to compensate for both of these locomotor perturbations, but the differences in kinematic response suggest that they each introduce unique challenges requiring different control strategies to maintain steady, constant speed locomotion.

Thank you to Rebecca Fisher, Rob Kulathinal and Jeanne Wilson-Rawls for discussions related to this study, to Kenro Kusumi for discussions and comments on the manuscript, to Dallas Malzi, Amber Dai and Laura Dallara for assistance with data collection and analysis, and to Scott Armstrong for help with editing.