Energetics Of Bipedal Running II. Limb Design and Running Mechanics

Animals use more metabolic energy the faster they run and, on a mass-specific basis, small animals use more energy than do larger ones (Taylor et al. 1970). It has been hypothesized that the differences in the energy cost of running that occur with speed and body size result from the increased energy cost of generating muscular force with faster muscle fibers (Taylor et al. 1980; Heglund et al. 1982; Taylor, 1985; Kram and Taylor, 1990; Rome, 1992). Among quadrupeds and small bipedal hoppers, the energy cost of locomotion decreases in proportion to the time available to generate force, measured as the time for which the foot is in contact with the ground (Kram and Taylor, 1990). Bipedal runners have longer legs, longer times available to generate force with each step and should presumably be able to use slower, more economical muscle fibers than quadrupeds (Roberts et al. 1998), yet their energy consumption is the same (Fedak and Seeherman, 1979; Roberts et al. 1998). In the present study, we seek to determine whether differences in the volume of muscle required to generate force might explain why bipedal runners use more energy for a given rate of force generation.

The rate at which a running animal’s muscles consume metabolic energy is determined by the total volume of muscle that must be active to support the body and maintain movement and by the rate at which a unit volume of muscle uses energy. Experiments in which animals were trained to carry weights while running suggest that much of the active muscle acts to provide support for the body. When rats, dogs, humans and horses run with weighted backpacks, their energy consumption increases in direct proportion to the supported load (Taylor et al. 1980). Up to a load equivalent to 27 % of body mass, stride frequency and time of foot contact do not change, but the mean force that must be produced against the ground increases in direct proportion to the load. Glycogen depletion studies in rats show that the cross-sectional area of muscle showing glycogen loss also increases in direct proportion to the supported load (Armstrong and Taylor, 1982), indicating that the increase in energy consumption with loading results from an increase in active muscle volume to provide more force.

The muscle force that a running animal must produce to provide support is determined not only by its body weight but also by the design of its limbs. Differences in the mechanical advantage with which limb muscles operate through the skeletal lever system result in large differences in the muscular force required to support the body (Gray, 1968; Smith and Savage, 1956). Biewener (1989) has demonstrated that limb mechanical advantage changes in a regular way with animal size; the crouched posture of small animals increases the moment of the ground reaction force about the joints and reduces their muscle mechanical advantage relative to larger animals with more upright postures. Among quadrupedal mammals, the mass-specific muscle force required to support the body scales according to M_b^−0.26, where M_b is the body mass (Biewener, 1989). On average, a 30 g mouse must produce 10 times as much muscle force as a 300 kg pony to produce 1 N of force against the ground. If the active muscle fibers produce the same force per unit cross-sectional area, the mouse must recruit 10 times the cross-sectional area of muscle to support each newton of body weight.

The total volume of muscle necessary to support the body is equal to the product of the cross-sectional area and the fascicle length of the active muscle. Longer muscle fascicles produce the same force per unit cross-sectional area, but a larger volume of muscle is activated for each newton of force produced. Small quadrupeds have relatively shorter muscle fibers than large quadrupeds, proportional to M_b^0.26 (Alexander et al. 1981; mean of exponents for all muscle groups for non-hoppers). From large to small quadrupeds, the increase in cross-sectional area of muscle required to support the body weight during running, proportional to M_b^−0.26, is offset by a decrease in the relative lengths of the muscle fibers, proportional to M_b^0.26, such that the volume of muscle required per unit body weight is constant across size in quadrupedal mammals (Biewener, 1989; Kram and Taylor, 1990).

It has been suggested that differences in muscle mechanical advantage and active muscle volume might explain the variation in the energy cost of generating force during locomotion in insects that have markedly different limb postures (Full et al. 1990). Humans use more energy when they run in a crouched posture (‘Groucho running’) because of the poor mechanical advantage and increased muscle recruitment at the knee (McMahon et al. 1987). To determine the influence of muscle mechanical advantage and fiber length on running energetics in bipeds and quadrupeds, we compared dogs and turkeys of the same size. On average, bipedal runners use 1.7 times as much metabolic energy as quadrupeds for a given rate of force generation (Roberts et al. 1998). We hypothesized that this difference resulted from a greater volume of active muscle necessary to generate force in bipedal runners. We tested two possible causes of an increased active muscle volume: (1) a poorer muscle mechanical advantage for force production, and (2) a longer fascicle length of the active muscles.

Three adult wild turkeys (Meleagris gallopavo L.) (two female and one male; mass 5.3±2.3 kg; mean ± S.D.) and three adult dogs (Canis familiaris L.; breed, Teacup terrier) (two males and one female; mass 4.5±1.0 kg) were purchased from USDA-approved breeders. These particular species were chosen because they have approximately the same body weight and they perform well on the treadmill. The animals were fed commercial bird and dog diets ad libitum and were housed in large indoor/outdoor pens.

The energetic and kinematic data for turkeys are presented in Roberts et al. (1998). All energetic and kinematic measurements were made while the animals ran on a variable-speed treadmill. We selected speeds within the range of comfortable trotting speeds for the dog and running speeds for the turkey. All animals were trained for 20 min a day, 3 days a week for at least 6 weeks before measurements were made.

The rate of oxygen consumption was determined using an open-circuit indirect calorimetry system (Fedak et al. 1981). Animals wore loose-fitting masks through which air was metered at constant rates. A small sample of the collected gas was removed, dried, scrubbed for CO₂ and measured for O₂ content with a Beckman F3 oxygen analyzer. Calibration with a known flow rate of nitrogen allowed calculation of (Fedak et al. 1981). The system was found to be accurate to within ±2 %.

The rate of O₂ consumption was measured continuously during each trial. We used measurements taken after a steady-state level of had been maintained for 5 min or more. Several measurements were made on different days and averaged. We used an energy equivalent of 20.1 J ml⁻¹ O₂ to convert measurements of oxygen consumption to energy consumption (Blaxter, 1989). Ė_metab for a given running speed was defined as the metabolic rate minus the zero-speed (y intercept) rate. The cost coefficient was determined according to equation 1. The cost of transport was defined as the slope of the least-squares linear regression of Ė_metabW_b⁻¹ against speed.

Time of foot contact (t_c) was measured from high-speed video recordings of treadmill running. Animals were recorded using a NAC 200 high-speed video camera operating at 200 fields s⁻¹. Time of contact was determined by counting the number of fields during which the foot was in contact with the treadmill belt. Measurements were taken for each foot and averaged. Ten strides were analyzed for each animal at each running speed.

Fig. 1 summarizes schematically the method for calculating the effect of muscle force and fascicle length on the volume of active muscle required to support body weight during running. This calculation is based on two variables: (1) the extensor muscle force required to balance the ground reaction force moments at each joint and (2) the mean muscle fascicle length. Measurements of the fascicle lengths (l) and muscle moment arms (r) of individual muscles are averaged to calculate a composite extensor muscle for each joint, with a mean fascicle length and moment arm (Fig. 1A). Muscle moment arm values are combined with joint moments measured by video and force plate analysis (Fig. 1B) to calculate the muscle force F_m required to support body weight at each joint. The volume of active muscle is represented as a cylinder having a cross-sectional area A_cs that is proportional to the muscle force F_m at the joint and a length that is equal to the mean fascicle length of the extensors at the joint (Fig. 1C).

Anatomical measurements were taken from three limbs of two turkeys and three front and three hind limbs from two dogs. To take advantage of cadaver material available from other experiments, two of the animals dissected were not the same individuals used for the running measurements, but they were similar in weight and condition.

Muscle fascicle length (l), pinnation angle (θ), moment arm (r) and mass (m) of each of the extensors were measured at the ankle, knee and hip for the turkey and dog, as well as the wrist, elbow and shoulder of the dog. The intertarsal joint of the turkey is referred to as the ankle for ease of comparison. Measurements were made on muscles that act primarily as extensors. Proposed function determined from anatomy (Raikow, 1985; Gatesy, 1989; Miller et al. 1979; George and Berger, 1966) as well as electromyography measurements of muscle activity during stance (Gatesy, 1989; Tokuriki, 1973; Goslow et al. 1981) were used to determine which muscles to include (see Table 1). Two-joint muscles were included only if the extensor moment arm was greater than the flexor moment arm. Anatomical measurements were similar in method to those described by Alexander (1974). Muscle moment arms were determined by measuring the perpendicular distance from the line of action of each muscle to the center of rotation of the joint. The joints were held at approximately the joint angle observed at mid-stride for muscle moment arm and length measurements. For muscles with a variable fascicle length, a mean value was determined from several measurements. For pinnate muscles, the fascicle length was determined after the muscle had been dissected away and bisected with a sharp scalpel. Muscle fiber pinnation angle was determined using a protractor or small goniometer.

Because the relative contribution of individual muscles to the sum moment at a joint cannot be determined from external measurements, it is necessary to determine a mean muscle fascicle length and moment arm for each joint. We calculated a composite fascicle length and moment arm for each joint from the mean of the extensors at each joint, weighted by cross-sectional area (Biewener, 1989). This provided a measure weighted to the capacity of each muscle to generate force.

We assumed that the contributions of lateral ground reaction forces and inertial forces required to accelerate the limbs were small relative to the horizontal and vertical ground reaction forces (Clark and Alexander, 1975; Pandy et al. 1988). Measurement of ground reaction force moment at each joint requires measurements of the position of the center of rotation of the joint, the point of force application and the direction and magnitude of the resultant ground reaction force. The position was recorded using high-speed video, and force measurements were carried out using a Kistler model 9261A force plate mounted midway along a 15 m track.

The running speed along the track was recorded as the animal broke photocell beams from three photocells located 1 m apart. Three criteria were necessary for a run to be analyzed further. First, the animal had to run at the appropriate speed. We chose speeds that were close to the preferred speeds of the animals (a 2.0 m s⁻¹ trot for dogs, 3.5 m s⁻¹ for turkeys) and only accepted runs within 0.2 m s⁻¹ of this speed. Second, the animal had to run at a steady speed with little acceleration or deceleration. The horizontal force was integrated during the foot contact time, and only those runs in which the integrated accelerative and decelerative forces (i.e. impulse) differed by less than 25 % were accepted as steady-speed runs. This results in a net change in speed per stride of less than 0.1 m s⁻¹. The third criterion was that the turkey should land with one foot only on the plate; the dog had to land with the front foot and then the hind foot on the same side of the body in the same run. This condition was necessary in order to resolve the ground reaction forces for individual limbs.

Two channels of vertical (summed front and back of the force plate) and one channel of summed horizontal components of force were recorded using a Kistler 9261A force plate. The signal was collected on an Apple Macintosh computer after analog-to-digital conversion on a National Instruments board (no. NB MIO 16H). The force data, along with one channel of photocell output and a synchronization channel, were collected using a National Instruments Labview program. Data collection was triggered when the animal broke the first photocell beam. Force data were sampled at 1000 Hz and filtered in software with a Chebyshev low-pass filter with a cut-off frequency of 60 Hz. The point of force application was determined by the relative outputs of the vertical channels in the front and back of the force plate. The force plate was calibrated for force and position daily using a known mass.

Joint centers of rotation were determined by marking each joint and video-taping the animal running across the force plate. Images were digitized (Quickcapture frame-grabbing board) and analyzed using NIH Image software. Force and video measurements were synchronized using a light-emitting diode triggered by the photocells and recorded on video.

A spot of white latex paint was applied to the skin overlying the approximate joint centers, using bony landmarks to aid in locating the center of rotation. For turkeys, only the ankle could be marked with paint. The wing obstructed the view of the hip, and knee movement occurred independently of skin movement. These joint positions were determined with the aid of a small marker on the back over the hip, the ankle marker and measured leg segment lengths. In an immobile bird, the distance from the ankle to the center of rotation of the knee was determined. A line of this length was drawn on the digitized image from the ankle marker following the middle of the tibiotarsal muscle mass to locate the knee center of rotation. The same strategy was used to determine the position of the hip from the position of the knee, using the marker glued to the back of the animal just above the hip as a guide. Because there is virtually no movement of the skin on the back, this marker was fixed relative to the acetabulum, thus providing a good indicator of the position of the center of rotation of the hip.

High-speed X-ray cine film was used to check the accuracy of joint markers in both the dogs and the turkeys (Siemens cineoradiographic apparatus with an Eclair GV16 camera). We placed radio-opaque markers on the skin and filmed the animals running on a treadmill. We could visualize the center of rotation of the joint directly with the X-ray film and observe how reliably our external markers indicated the joint axis. The largest error in joint marking occurred at the most proximal joints and was less than 0.5 cm.

The total muscle force also included a component of force contributed by two-joint muscles that extended one joint but also flexed another. For example, if the gastrocnemius muscle is active to extend the ankle, it will also produce a small flexion moment about the knee that must be balanced by knee extensors. The muscle flexor moments produced by two-joint muscles were calculated from the flexor moment arm and the force required to balance ground-reaction-based extensor forces, assuming that force was distributed equally (by cross-sectional area) across all joint extensors. In both dogs and turkeys, muscle flexion moments from two-joint hip and ankle extensors increased the extensor force required at the knee by approximately 60 % over the force required to balance ground reaction forces only. The extensor force required to balance these antagonist muscle forces was less than 2 % of the total extensor force required in the limb in dogs and less than 9 % in turkeys. It was assumed that any other forces due to co-contraction of antagonist flexor muscles were negligible. This assumption is supported by the close agreement between muscle forces measured directly by force buckles and the forces calculated from force-plate measurements in the ankle extensors of hopping kangaroo rats Dipodomys merriami (Biewener et al. 1988).

The energetic costs of running were similar in trotting dogs and running turkeys (Fig. 2). The speed range of these gaits overlapped between 1.5 and 2.0 m s⁻¹. At these two speeds, the measured metabolic rate was approximately 15 % higher for turkeys, but this difference was not significant (two-tailed t-test, P=0.09).

The slope of the relationship between mass-specific energy cost (Ė_metabW_b⁻¹) and running speed is the ‘cost of transport’ (E_transW_b⁻¹). It represents the amount of energy used by the muscles to move a unit body mass a unit distance. The cost of transport was not significantly different for dogs and turkeys (Table 2). Transport cost changes in a regular way with body mass, and the values for the dog and turkey given in Table 2 are close to the values predicted for their size (Taylor et al. 1982).

Turkeys took longer steps than did dogs running at the same speed (Table 2). Step length, equal to the product of running speed and time of contact, was twice as long for turkeys. This is explained by the long tibiotarsal and tarsometatarsal bones of the turkeys, giving them almost twice the leg length and a 1.6-fold greater hip height at mid-stance than dogs (Table 2). The longer steps of the turkeys meant that each foot was in contact with the ground 1.8 times longer than the feet of the dogs. A longer foot contact time provides more time to generate force and presumably allows the turkeys to use slower, more economical muscle fibers.

The muscle force required to support weight is determined by the magnitude of the ground reaction force and the mechanical advantage of the muscle. Changes in the ground reaction force moment arm, R, that occur during a step result in a pattern of muscle force that does not necessarily parallel changes in the ground reaction force with time (Fig. 3). For example, hip extensors in the turkey experience their peak forces early in the step, while the knee extensor force is maximal later in the step, after the peak ground reaction force. The ground reaction force can also act to extend a joint, indicating an opposing flexor muscle force. These moments were small relative to the moments that required extensor muscle force and were recorded as zero extensor force. The total integrated flexion moment summed for all joints was less than 8 % of the total extensor moment in dogs and less than 2 % in turkeys.

Biewener (1989) defined the effective mechanical advantage (EMA) as the mean mechanical advantage, ⁠, of the limb extensors averaged over a step. We calculated the EMA over the period of the stride when the measured muscle force was greater than 25 % of the measured maximum muscle force (Biewener, 1989). The EMA of the turkey hind limb, 0.451±0.052 (mean ± S.E.M.), was not significantly different from that of the dog hind limb, 0.652±0.089 (P=0.12, two-tailed t-test), or the dog fore limb, 0.298±0.025 (P=0.06, two-tailed t-test). The scaling of EMA across body mass for a number of species of mammals (Biewener, 1989) predicts an EMA of 0.337 for the fore limb and 0.385 for the hind limb for a mammal of the same body mass as the dogs in the present study. There are no comparable values published for birds.

The EMA expresses the mean of the ground reaction force produced per unit muscle force for all of the extensors and is useful for comparing the mechanical design of the limbs of different animals. For the purposes of an energetic comparison, however, we required a measure of muscle force that could be summed for all joints and that included the entire force-generating period. We divided the integrated muscle force, ∫F_m (the area under the muscle force curve in Fig. 3), at each joint by the integrated ground reaction force, ∫F_g (the area under the ground reaction force curve in Fig. 3). This gives a measure of the amount of muscle force required at each joint to produce 1 N of ground reaction force (Fig. 4). In both turkeys and dogs, the highest muscle forces occurred at the distal joints. When all joints were included, the muscle force required to support each newton of body weight, ∫F_sum/∫F_g, was 9.8±0.8 for turkeys compared with 8.0±0.6 (mean ± S.E.M.) for dogs. This 1.2-fold difference was not significant (P=0.15, two-tailed t-test) (Fig. 4).

Running turkeys used a greater muscle volume to generate a unit of force than dogs (P=0.003, two-tailed t-test) because of their longer muscle fascicles (Fig. 5). If we assume that the dog and turkey muscles produce the same force per active cross-sectional area of muscle, the product of mean muscle force and fascicle length is proportional to the volume of active muscle (Figs 1, 5). Thus, the turkey uses 2.5 times the active limb muscle volume to generate each unit of ground reaction force as the dog (P=0.003, two-tailed t-test). Much of this difference can be explained by the high forces and relatively long fascicles in the ankle of the turkey. The volume of muscle required to support weight at the ankle of the turkey was at least as much as that used for all the joints of the front and hind limb combined in the dog. The was due to both high muscle forces (Fig. 4) and long fascicles at the ankle in turkeys (Table 1). The difference in distribution of active muscle volume in dogs and turkeys was reflected in the distribution of total extensor muscle mass (Table 1). For example, the total extensor mass available at the ankle of the turkey is more than four times the combined muscle mass at the wrist and ankle of the dog.

The limb extensor muscles of the turkey have much longer fascicles than those of a similarly sized dog (Table 1). To determine a fascicle length that is relevant to the energetic cost of generating force, we have calculated an ‘active fascicle length’, L_act, which is the mean fascicle length at each joint weighted by the muscle force produced at that joint (Fig. 6B). This should provide a measure of the mean length of the active muscles. For example, very long fascicles of the extensor muscles that act about a particular joint would not contribute to the active fascicle length if no force is generated (see equation 6). The active fascicle length for the turkey was 3.3±0.4 cm, 2.1 times greater than the value of 1.6±0.1 cm for the dog (mean ± S.E.M., P=0.01, two-tailed t-test). Turkeys use muscles that are, on average, 2.1 times longer while producing similar muscle forces, providing an explanation for their 1.9-fold greater cost coefficient (Fig. 6; Table 2).

We started with the hypothesis that running bipeds use more muscle to run than do quadrupeds, to explain why their longer legs and slower rates of force generation do not result in a lower energy cost. Our data indicate that the longer muscle fascicles of the turkey require them to recruit more than twice the volume of limb muscle as a running dog to support their body weight. This helps explain the higher cost coefficient of running bipeds compared with quadrupeds and provides a link between limb design and running energetics.

Our approach estimates the relative volume of active muscle only for extensors of the limbs. We could not estimate the muscle forces involved in stabilizing the trunk, head and neck. Microsphere measurements in miniature swine Sus scrofa indicate that non-limb muscles receive less than 20 % of the total blood flow to muscle during treadmill exercise (Armstrong et al. 1987), suggesting that the energy cost of fueling non-limb muscles is considerably less than that required for limb muscles. The muscular effort required in the non-limb muscles of running bipeds may be even lower than for quadrupeds. Thus, our estimate of active limb muscle volume may overestimate the difference in total active muscle volume between dogs and turkeys.

A critical assumption of our approach is that the force produced per unit cross-sectional area of active muscle, σ, is equivalent in running dogs and turkeys. This stress is determined by the force-generating capability of the muscle, P₀, and the relative shortening velocity, V/V_max, at which it operates. Few studies of isolated avian muscle have been performed, but in situ measurements suggest that the maximum force-generating capacity is within the range of published values for mammals (Biewener et al. 1992). Large differences in active muscle stress occur when muscles operate at different V/V_max in order to perform different tasks. Muscles contracting at an intermediate shortening velocity, where high power outputs are produced, generate only approximately one-third of the force generated during an isometric contraction (Hill, 1938). Elastic energy storage and recovery in tendons can reduce the energy cost of running by allowing muscles to operate nearly isometrically to produce a high active muscle stress and reduce the volume of muscle required to support the body (Roberts et al. 1997). For both bipedal and quadrupedal runners, the mechanical energy fluctuations of the body and limbs are unrelated to changes in metabolic energy consumption across animal size and speed (Heglund et al. 1982), and level running at a constant speed requires little net mechanical work. It seems likely that muscles operate on average close to isometrically in both dogs and turkeys, allowing similar force per unit cross-sectional area of active muscles.

Quadrupedal and bipedal runners show similarities in the mechanics of movement of the body and limbs (Cavagna et al. 1977; Gatesy and Biewener, 1991). Our results show that turkey limbs also have a mechanical advantage that is similar to that predicted for quadrupedal hind limbs (Biewener, 1989). However, there is a marked difference in the mass distribution in the limb between dogs and turkeys that may reflect different strategies for support and power production by limb muscles. Both dogs and turkeys generate their highest forces at the joints with the shortest muscle fibers. This is consistent with the suggestion that shorter muscle fibers can be used to provide more economical force. However, the turkey requires a much greater active muscle volume at the ankle, because it has longer muscle fascicles at that joint, nearly four times as long as the fascicles at the wrist of the dog. This difference may reflect the specialization of distal muscles of the quadruped for force production compared with muscles that produce the mechanical power necessary to accelerate, run uphill or jump. Turkey ankle muscles provide force nearly isometrically during level locomotion, but they also shorten considerably and generate power during uphill running (Roberts et al. 1997). Dogs generate the power for a jump in proximal muscles, using the muscles of the ankle and the wrist only to provide support (Alexander, 1974). While quadruped limbs have a characteristic taper, the muscle mass of the ankle extensors in turkeys is almost as large as the extensor mass at the hip and knee combined. This gross morphological difference may reflect fundamental differences in the way that mammalian and avian hind limbs operate to produce mechanical power for locomotion.

Our results suggest that bipedal turkeys require a greater volume of active muscle than quadrupedal terriers when they run, primarily because they have longer muscle fascicles. The difference in active fascicle length measured between dogs and turkeys was approximately the same as the difference in leg length, but this outcome was not inevitable; very long-legged birds do not have proportionately long muscle fascicles (Bennett, 1996). The longer legs of the turkey allow it to use slower, more economical fibers, but having longer fascicles means that more muscle must be active. This appears to explain the similar energy cost of running in the biped and quadruped.

We thank G. E. Goslow Jr, J. E. A. Bertram and R. Kram for experimental assistance. A. W. Crompton kindly performed the X-ray cine measurements. A. A. Biewener provided critical advice throughout the study. We thank P. Ramirez for assistance in care and training of the animals. This work was supported by NIH grant AR18140 to C.R.T. and an NSF Graduate Fellowship to T.J.R.