Locomotor Performance and Energetic Cost of Sidewinding by the Snake Crotalus Cerastes

The loss of limbs has evolved independently within several major groups of vertebrates (Gans, 1974; Edwards, 1985). The best known and most diverse of these lineages is the snakes, a monophyletic group (Cadle, 1987) containing more than 2000 species (Duellman, 1979). Snakes may use different locomotor behaviours depending upon speed and the topography of the locomotor surface (Gray, 1946; Gans, 1974; Jayne, 1986). One mode of limbless locomotion, apparently unique to snakes, is sidewinding (Gans, 1974). Sidewinding is the primary locomotor behaviour of some specialized species in their natural environment, yet many other species of snake will sidewind when they encounter smooth surfaces such as sand or mud (Gans, 1974; Jayne, 1986).

In contrast to the lateral undulatory and concertina modes of limbless locomotion, during sidewinding there is no sliding contact with the substratum (Mosauer, 1932; Gray, 1946; Gans, 1974; Jayne, 1986). A sidewinding snake simultaneously bends and lifts its body while maintaining two regions of static contact between its belly and the ground. The resulting pattern of movement leaves a series of disconnected tracks that are oblique relative to the overall direction of travel (Mosauer, 1935; see Fig. 1). Although these general aspects of kinematics are well documented, no study of sidewinding has systematically determined speed, endurance or the scaling of these measures of locomotor performance.

For diverse groups of limbed animals, the energetic cost of locomotion appears to be determined primarily by the mass of the animal (Bennett, 1985; Full, 1989). For snakes, the choice of locomotor mode significantly affects locomotor energetics. For example, the energetic cost of concertina locomotion is seven times greater than that of lateral undulation (Walton et al. 1990). Furthermore, this variation in energetic cost between modes of limbless locomotion is large compared to that of the net cost of transport of limbed lizards of similar mass (John-Alder et al. 1986). The objectives of this study examining sidewinding locomotion are (1) to determine the scaling of burst speed and endurance with size, (2) to determine the energetic cost of this mode and (3) to compare these values with those for different modes of limbless locomotion.

The sidewinder (Crotalus cerastes Hallowell) is a small species of rattlesnake which chiefly inhabits the sandier regions of the Sonoran and Mojave Deserts of North America (Klauber, 1972). This snake employs sidewinding as its primary locomotory mode (Mosauer, 1932; Klauber, 1972). Snakes used in this study were captured in San Bernardino Co., California, USA (CA Dept. Fish and Game permit no. 0690 to B.C.J.). After experiments, snakes were measured [snout-vent length (SVL) and total length (TL) to the nearest millimetre], weighed (to the nearest gram) and sexed (Fitch, 1987). Locomotor performance was measured for 41 snakes (20<SVL<55cm; 7<mass<147g), ranging from newborn (<1 month) to adult (>4 years). A subset of eight adult snakes (mean 5VL=51.0cm, range 45.3−54.8cm; mean TL=55.1cm, range 49.5−58.4cm; mean mass=110g, range 79−147 g) was used to determine the energetic cost of sidewinding.

Adult and subadult snakes were fed newborn rats every third week and all snakes had access to water ad libitum. Captive sidewinders were housed individually in either small aquaria (191) or plastic shoe boxes with a 12 h light : 12 h dark photoperiod, daytime photothermal gradient of 20−32°C and night-time temperatures of approximately 20 Key words: burst speed, endurance, oxygen consumption, cost of transport, Crotalus cerastes. C. All trials were conducted in a temperature-controlled room maintained at 30±l°C. This temperature is within the range of body temperatures of surface-active C. cerastes (Moore, 1978). Prior to the performance and cost trials, paint marks were placed at three positions along the mid-dorsal line of the snake to facilitate kinematic measurements.

Burst speed trials were conducted on a 3-m long track, with a floor of smooth, dense rubber matting. Preliminary experiments revealed that this surface minimized slippage by snakes which could impair rapid locomotion. The vertical sides of the track (30 cm high) were adjusted to a width approximating the SVL of the individual being tested. A camera, suspended directly over the track, video-taped (60 fields s⁻¹) each trial with a shutter speed of 0.001s and a time display to the nearest 0.01 s was recorded on each video field. Two trials (12 days apart) were conducted for each snake. By tapping the mat or the snake’s tail with a soft brush, each individual was encouraged to move down the track 3−5 times in rapid succession for each trial. Other studies have recognized the varying behavioural responses of individual snakes to performance testing (Arnold and Bennett, 1984). In our original sample, three sidewinders assumed a defensive posture, refused to crawl and could not be tested.

The motion of sidewinding involves periodic static contact between the snake and the ground. We determined the duration of each cycle of movement (±0.01 s) over the middle 1.5 m of the track for each replicate. For the single cycle that was completed in the shortest time, we measured the distance travelled over that time interval (Fig. 1) by using a customized computer program for digitizing data from individual video images. Burst speed was calculated by dividing this distance by the duration of that cycle.

Endurance trials were performed on a motorized treadmill with a tread surface of rubber-impregnated cloth, 100cm long and 45 cm wide. For each of the 41 snakes, a single test of endurance was made using a tread speed of 0.50 km h⁻¹. The endurance of six adult snakes (four of which were used to determine energetic cost) was also measured at speeds of 0.60, 0.70 and 0.80 km h⁻¹. Snakes were encouraged to keep pace with the treadmill by tapping either the treadmill’s surface or the snake’s tail. A trial was terminated when an individual failed to match tread speed three times in rapid succession. If an individual maintained locomotion for 180 min, the trial was terminated. The endurance for an individual was measured as the time (±0.5 min) from the beginning of its movement on the treadmill until the end of its trial.

To measure rates of oxygen consumption, a light-weight (1.9 g) clear plastic mask (0.8 mm thick Lexan) enclosed the entire head of the snake and was kept in place with a piece of foam dorsal to the snake’s braincase. The mask was designed not to interfere with respiration or sight, or to compress the enlarged supraocular scales. The back of the mask was open and the tapering front of the mask was attached to tubing connected to an oxygen analyzer. Air was drawn (225 ml min⁻¹) through the air line into columns of water absorbent (Drierite) and CO₂ absorbent (Ascarite) prior to entering the oxygen analyzer (Applied Electrochemistry, model S-3A). Oxygen content (% O₂) of the drawn air was recorded on a chart recorder.

We used tread speeds of 0.28, 0.35, 0.43, 0.50, 0.58, 0.65 and 0.73 km h^-1 to measure oxygen consumption and we video-taped each trial. From playback of the video tapes, we determined the frequency of movement and subjectively rated the quality of the trial. We selected time intervals for calculating average oxygen consumption based upon the consistency and quality of movement (matching tread speed, using the entire body to sidewind, and not climbing treadmill walls). Within each record of oxygen consumption, lasting for 6−15 min for each tread speed, average rates of steady-state oxygen consumption were calculated from points at every 2-s interval for the highest quality 2−3 min span. The average coefficient of variation for 30-s means of oxygen consumption within each of the 39 trials was 20 %, and only one of these trials had a significant correlation between the 30-s means and time. Pre-exercise rates of oxygen consumption were measured from masked snakes resting on the treadmill prior to their locomotor trials. All rates of oxygen consumption were corrected for standard pressure and temperature and calculated following the method of Withers (1977) for open-flow mask respirometry.

Most statistical analyses were performed using a microcomputer version of SPSS (PC+ SPSS). Statistical significance was designated at P<0.05. Burst speed values, endurance times and size variables of the snakes were log-transformed (base 10) prior to regression analysis of scaling. All regressions were calculated using the least-squares method.

Both TL and mass increased with increasing SVL for C. cerastes (Table 1). Male and female sidewinders differed in both TL (log TL=1.0071ogSVL—0.013sex+ 0.027, r²=0.999) and mass (logmass=2.7611ogSVL-0.021sex-2.700, r²=0.961) as indicated by the highly significant (P<0.001) coefficient of a coded variable for sex (0=male, 1=female). Hence, for a given SVL, these males have longer tails and are heavier than females. Despite its significance, sex of the snake had only small effects on TL and mass, as suggested by similar coefficients of determination (r²) for the regressions including sex and those using SVL alone (Table 1).

Burst speed of C. cerastes ranged from 0.5 to 3.7 kmh⁻¹ and increased as SVL and mass increased (Table 1). Fig. 2A illustrates the linear increase of log burst speed (kmh⁻¹) with increased logSVL, and a multiple regression indicated no significant quadratic effect of SVL. The increase in burst speed with SVL resulted primarily from a significant increase in the distance covered per cycle with increased SVL, as the maximum frequency of movement during burst speed did not change significantly with SVL (Table 1). A multiple regression revealed that sex had no significant effect on burst speed.

Endurance time at 0.50kmh⁻¹ ranged from 3.5 to more than 180min and showed a simple linear increase with both SVL and mass (Table 1; Fig. 2B). Males had significantly greater endurance (E) than females (logE=2.8531ogSVL-0.228sex-2.996, r²=0.594; P for coefficient of sex=0.024).

For a subset of six C. cerastes of similar size, Fig. 3A illustrates the decrease of endurance with speed. The mean±l S.D. endurance times (min) at 0.50, 0.60, 0.70 and 0.80kmh⁻¹ are 88.4±58.2, 72.9±61.8, 25.5±19.4 and 6.3±5.3, respectively. Compared to 0.50kmh⁻¹, C. cerastes moving at 0.80kmh^-1 experienced a 93% reduction in endurance. Despite the initial decline in endurance, there was no significant difference in endurance between 0.50 and 0.60kmh⁻¹ (paired t=0.806). At 0.70kmh^-−1, endurance had decreased significantly from that at 0.50kmh⁻¹ (paired t=3.171, P<0.05). Endurance time was highly variable at these speeds, evident from an adult (SVL=45.3cm, mass=79g) moving continuously for 481 min at 0.60kmh⁻¹ during a preliminary trial. Crotalus cerastes increases speed between 0.30 and 0.80 km h⁻¹ by linearly increasing the frequency of sidewinding (Fig. 3B), not the distance per cycle. The frequency of movement (cyclesmin⁻¹) at 0.80kmh⁻¹ is nearly 2.5-times that at 0.30kmh⁻¹.

For each of the eight C. cerastes, rates of oxygen consumption (m1O₂g^{− l} h⁻¹) were measured for each of its four slowest speeds (Fig. 4A). Rates of oxygen consumption increased linearly with speed between 0.28 and 0.50 km h⁻¹, and maximum was attained between 0.50 and 0.65 km h⁻¹ (Fig. 4A). Based on the greatest value of observed for each of the individuals during sidewinding, mean was 0.405 ml O₂g⁻¹ h⁻¹ (A=8, S.E. =0.026). As in previous studies, we related to speed with a regression using all 32 data points from 0.28 to 0.50kmh⁻¹. The slope of this regression (A=32), the net cost of transport, was 0.408 ml O₂g⁻¹ km⁻¹. Table 2 illustrates an alternative approach involving separate regressions for each individual in order to account for the lack of independence in the observations of oxygen consumption. The slopes of these separate regressions ranged from 0.194 to 0.576mlO₂g⁻¹ km⁻¹ and the mean of these separate slopes was identical to that of the regression utilizing all 32 values. The y-intercept of the regression shown in Fig. 4A is 0.182mlO₂g⁻¹h⁻¹, which is significantly greater than zero (t=4.237, P<0.001) and significantly greater than the mean pre-exercise of 0.108 ml O₂g⁻¹h⁻¹ (t=3.519, PcO.Ol).

The energetic cost per cycle of sidewinding for the four aerobically sustainable speeds is illustrated in Fig. 4B. These values were calculated by dividing massspecific for each individual at a particular speed by the number of cycles per hour performed during that measurement of ⁠. A one-way analysis of variance on the cost per cycle detected significant variation among the different speeds (d.f.=3, 28, F— 4.073, P<0.05). However, a Tukey multiple range test revealed that the only significant difference in cost per cycle occurred between 0.28 and 0.50kmh⁻¹. The mean costs per cycle were 0.375 μlO₂g^-1 cycle⁻¹ (A=8, S.E. =0.043) at 0.28kmh⁻¹ and 0.306 μlO₂g⁻¹ cycle⁻¹ (/V=24, S.E. =0.011) from 0.35 to 0.50 km h^-1.

As in other studies of speed and endurance, the size of C. cerastes had a marked effect on performance. For example, a small juvenile (SVL=25cm) would have a predicted burst speed and endurance time of 1.34kmh^-1 and 7.8min, respectively, compared to those for a large adult (SVL=50cm) of 2.35kmh⁻¹ and 56.2 min (Table 3). Jayne and Bennett (1990) used a similar protocol (T=30°C) to test the performance of Thamnophis sirtalis using lateral undulation. Hence, these data are the most useful for comparisons with varying sizes of C. cerastes. Table 3 summarizes values of size and locomotor performance predicted for a given SVL of these two species. Compared to C. cerastes, T. sirtalis with equal SVL are less than half the mass and possess longer tails (Table 3). The burst speed of T. sirtalis is greater than that of C. cerastes with the smaller individuals having the most divergent values. The endurance of T. sirtalis is less than that of C. cerastes with the larger individuals having the most divergent values.

Similar to this interspecific comparison of burst speed between lateral undulation and sidewinding, Jayne (1986) found that the burst speed of sidewinding by Nerodia fasciata was less than that of terrestrial and aquatic lateral undulation performed by the same species. These measurements of speed during snake locomotion suggest that sidewinding is not an extraordinarily rapid mode of locomotion. However, it is apparent from comparative kinematic studies that snakes can move using sidewinding on surfaces where lateral undulation is extremely difficult (Gray, 1946; Gans, 1974; Jayne, 1986).

Jayne and Bennett (1990) calculated the residual of mass predicted from SVL for T. sirtalis, and they found that this quantity, similar to other indices (Van Berkum et al. 1989) of overall condition, showed a significant quadratic relationship with locomotor performance. In other words, T. sirtalis that were too thin or heavy, based upon their SVL, showed a decrement in size-corrected (residual) burst speed and endurance. For our sample of C. cerastes there was no significant linear or quadratic effect of mass residual upon either burst speed residual or endurance residual. Furthermore, burst speed and endurance residuals were not significantly correlated for C. cerastes (r=0.290, P=0.066), whereas these two quantities were significantly correlated for T. sirtalis, though only weakly (r=0.263) (Jayne and Bennett, 1990).

To facilitate comparisons with the study of Walton et al. (1990) on the locomotion of the snake Coluber constrictor, we selected C. cerastes of similar mass, and we performed all experiments at an identical temperature (30°C). Comparing the size measurements of the sample of C. constrictor of Walton et al. (N=7, mean SVL=71cm, mean TL—95 cm, mean mass=103g) shows that the C. cerastes (N=8, mean SVL=51.0cm, mean TL=55.1 cm, mean mass=110g) used for energetic measurements were relatively stouter animals with shorter tails. The mean burst speed of these C. constrictor (N = 8 ⁠) is more than twice that of these adult C. cerastes (N = 8 = 2.35km h⁻¹) a highly significant difference (two-tailed t=7.80, P<0.001). Interestingly, the mean endurance times (E) at 0.60kmh⁻¹ of these C. constrictor (N=6, Ē=33min) and C. cerastes (Fig. 4, N=6, Ē=73 min) did not differ significantly (t=1.550), primarily because of the large variance associated with these measurements in both species.

The net cost of transport (NCT, in mlO₂g⁻¹km⁻¹) is widely used to compare the energetic cost of aerobically sustainable locomotion among different species as well as among different modes of locomotion (Schmidt-Nielsen, 1972). A twotailed i-test (d.f.=53, t=2.923, P<0.01) reveals that the NCT of 0.408 ml O₂ g⁻¹ km⁻¹ for C. cerastes sidewinding is significantly less than the NCT of 1.15 ml O₂g⁻¹ km⁻¹ for C. constrictor performing terrestrial lateral undulation (Walton et al. 1990). The NCT of 8.49mlO₂g⁻¹km⁻¹ for C. constrictor performing concertina locomotion in 10-cm wide tunnels is significantly greater than that for both lateral undulation (Walton et al. 1990) and sidewinding. Several studies of limbed locomotion have documented that NCT scales negatively with increasing mass (Taylor et al. 1982; Full, 1989). The scaling equation for the NCT of limbed lizards (John-Alder et al. 1986; Fig. 5) predicts a value of 1.14 ml O₂g⁻¹ km⁻¹ for a 103-g individual, and this value is indistinguishable from the NCT of lateral undulation by Coluber but intermediate to the NCTs for concertina motion by Coluber and sidewinding by Crotalus. As emphasized in Fig. 5, the more than 20-fold variation in NCT found among different locomotor modes within snakes of similar mass exceeds the amount of variation in NCT reported for lizards over three orders of magnitude of mass. Hence, different modes of limbless locomotion are of primary importance in determining the energetic cost of movement.

Previous studies (Walton et al. 1990; Jayne and Davis, 1991) have suggested that some of the factors contributing to the high NCT of snake concertina locomotion relative to lateral undulation are the greater frequency of movements, smaller distance travelled per cycle of movement, and sliding plus static frictional resistance. Despite the fact that C. constrictor travelled less distance per cycle during concertina locomotion compared to lateral undulation, the cost per cycle of movement was significantly higher during concertina locomotion (Walton et al. 1990).

The spacing of projections from the locomotor surface affects the frequency and amplitude of lateral undulatory movements of snakes (Jayne, 1986) but, in our studies, the two groups of snakes used to calculate NCT of sidewinding and lateral undulation had similar frequencies of movement. For example, at 0.50 km h⁻¹, C. cerastes (N=6, Fig. 3B) performed sidewinding and C. constrictor (N=5, Walton et al. 1990) performed lateral undulation with mean frequencies of 26.8 and 27.3 min⁻¹, respectively, and these values are not significantly different (t=0.319). These matching frequencies of movement at identical forward speeds indicate that the mean distance travelled per cycle must also be equal. Thus, the lower NCT of sidewinding compared to lateral undulation cannot be attributed to differences in the modulation of the frequency and amplitude of movements.

The cost per cycle of movement for aerobically sustainable speeds of sidewinding (N=32, Fig. 4B) is significantly lower (Mann-Whitney U=676, P<0.001) than that of terrestrial lateral undulation (N=23, Fig. 2B in Walton et al. 1990). A similar conclusion is reached when the comparison is restricted to one speed in common to both experiments. At 0.50 km h⁻¹, there was a significantly lower (t=4.479, P<0.01) cost per cycle for sidewinding (N=8, mean=0.28 μlO₂g⁻¹ cycle^-1), which was about half that of lateral undulation (N=5, mean=0.52 μl O₂ g⁻¹ cycle⁻¹). Thus, to explain the different NCTs of lateral undulation and sidewinding it is helpful to examine factors that could affect the cost per cycle of movement.

Snakes performing terrestrial lateral undulation propagate waves of bending posteriorly such that they push laterally and posteriorly against irregularities in the locomotor surface (Gray, 1946; Gans, 1974). The resulting reactive forces are directed anterio-medially and sum to a net forward reactive force (Gray and Lissman, 1950). All points on the body of the snake are propelled with a constant speed following a more or less sinusoidal path (Jayne, 1986). Hence, a laterally undulating snake must exert forces sufficient to (1) bend its body laterally, (2) push its sides against vertical surfaces, (3) overcome lateral sliding frictional resistance, and (4) overcome ventral sliding frictional resistance (=snake weight ×coefficient of friction).

During sidewinding, snakes must exert forces sufficient to bend the body laterally as it is pushed away from one region of static contact and pulled towards a more anterior region of static contact with the ground (Fig. 1). Sidewinders must also lift the body between regions of contact with the substratum. This movement is accomplished by dorsiflexion of the vertebral column. Hence, lateral flexion is common to both sidewinding and terrestrial lateral undulation. Both locomotor modes use the longissimus dorsi and iliocostalis muscles as the primary lateral flexors and these epaxial muscles are activated in an alternating unilateral fashion with a posterior propagation of muscle activity such that the contractile tissue is active as it shortens (Jayne, 1988a,b). These similarities in lateral bending suggest that similar amounts of work may be done for this portion of movement in the two locomotor modes.

The pattern of muscle activity that is unique to sidewinding is the bilateral activation of the most medial epaxial muscles in order to dorsiflex and lift the snake’s body away from the ground (Jayne, 1988a), and specialized sidewinding species have fewer vertebrae spanned by these muscles than generalized species (Jayne, 1982). Based on the low energetic cost of sidewinding, this lifting of the body appears to be energetically much more economical than the combined work that must be done during lateral undulation to overcome the ventral and lateral sliding frictional forces.

In lateral undulation, the proportion of total work done to overcome external frictional forces remains unclear. It is also unclear if sidewinding is energetically more costly in snake species lacking the specialized anatomical features common to proficient sidewinders. Compared to snake species that largely use other modes, sidewinding specialists have fewer vertebrae (Jayne, 1982), possess shorter segments of epaxial musculature (Jayne, 1982) and are small (TL less than 90 cm), moderately stout-bodied with relatively short tails (Broadley, 1962; Gans and Mendelssohn, 1972; Klauber, 1972; Moore, 1980). Jayne (1986) noted that the relatively longer tail of Nerodia fasciata was ineffective during sidewinding.

The mean pre-exercise rate of oxygen consumption (0.108 ml O₂g⁻¹ h⁻¹) is similar to the predicted resting metabolic rate of squamate reptiles of equivalent mass (110g) at 30°C (0.0968 ml O₂g⁻¹h⁻¹, Andrews and Pough, 1985). Maximal oxygen consumption during the locomotion of C. cerastes in our study was about half that observed for C. constrictor during lateral undulation (Walton et al. 1990). However, the lower energetic cost of sidewinding enabled C. cerastes to have a maximal sustainable aerobic speed (about 0.6 km h⁻¹) equal to that of C. constrictor performing lateral undulation (Walton et al. 1990). Maximal aerobic speeds are 28 % and 10 % of burst speed in Crotalus and Coluber, respectively. Our findings, and those of Walton et al. (1990), agree closely with those of Ruben (1976), who used different methods to obtain ⁠. Ruben (1976) found that Coluber had about twice and maximal whole-body levels of lactate about 1.5 times those of the rattlesnake Crotalus viridis. Interestingly, by using a different locomotor behaviour from Coluber, C. cerastes was able to obtain similar endurance and maximal aerobic speed as Coluber, which had much greater aerobic scope. Explaining the more than twofold difference in burst speed between Coluber and C. cerastes remains an intriguing problem. It is not obvious whether it is differing mechanics, morphology or physiological capacity that explains the different burst speed of C. cerastes sidewinding compared to that of C. constrictor laterally undulating.

 
• D

  Distance travelled per cycle of movement (cm)

 
• E

  Endurance time at a constant tread speed (min)

 
• f

  Frequency of movement (eyeless⁻¹)

 
• M

  Body mass (g)

 
• NCT

  Net cost of transport (mlO₂g⁻¹h⁻¹)

 
• SVL

  Snout-vent length of snake (cm)

 
• TL

  Total length of snake (cm)

 
• 

  Mean forward velocity (kmh⁻¹)

Financial support was provided by University of California Natural Reserve System and University of California, Los Angeles, to S.M.S., and by National Science Foundation grants BNS 8919497 to B.C.J. and DCB 88-12028 and BSR 89−18054 to A.F.B. We wish to thank C. Wallace and J. Davis for technical assistance. We are grateful to the University of California’s Granite Mountain Reserve for lodging while collecting snakes.