Abstract
The present study aims to assess energy demand and supply in 100m sprint running. A mathematical model was used in which supply has two components, aerobic and anaerobic, and demand has three components, energy required to move forward (C), energy required to overcome air resistance (Caero), and energy required to change kinetic energy (Ckin). Supply and demand were equated by using assumed efficiency of converting metabolic to external work. The mathematical model uses instantaneous velocities registered by the 1997 International Association of Athletics Federations world champions at 100 m in men and women. Supply and demand components obtained in the male champion were (in J/kg) aerobic 30 (5%), anaerobic 607 (95%), C 400 (63%), Caero 83 (13%), Ckin 154 (24%). Comparatively, a model that uses the average velocity of the male and female 100m champions overestimates Ckin by 37 and 44%, respectively, and underestimates Caero by 14%. We argued that such a model is not appropriate because Ckin and Caero are nonlinear functions of velocity. Neither height nor body mass seems to have any advantage in the energetics of sprint running.
 energy cost of running
 acceleration
 air resistance
 anaerobic energy
 body size
it has been proposedthat the energetics of running might be appropriately described by using a powerbalanced model based on a supplydemand approach (11, 21, 22, 2527). In such a model, the supply side has two components (aerobic and anaerobic) and the demand side has three components [energy required to move forward (C), energy required to overcome air resistance (Caero), and energy required to change kinetic energy (Ckin)]
Concerning demand, the relative contribution of each component is very different depending on the running distance. The longer the running distance, the lower the contribution of Ckin and Caero (11). Thus, for shortdistance running, such as a 100m race, both Ckin and Caero make significant contributions to demand.
The most developed application of Eq. 1 concerns middle and longdistance running in which the average running velocity can be approximated by d/t, where d is the known distance covered (in meters) and t the measured running time (in s). In the model that uses the average velocity (d/t), Caero = k · η^{−1} · d ^{2} · t ^{−2}, where k is the air friction constant (in kg^{−1} · m^{−1}) and η = 0.5 (11, 21, 22), and Ckin = 0.5 · η^{−1} · d · t ^{−2}, where η = 0.25 (11, 21, 22).
In principle, this analytical model employed for the mathematical analysis of a wide range of running performances (21) can be applied to investigate the results of individual athletes. This has been done successfully by di Prampero and colleagues (10), who demonstrated good relationships between actual and predicted performances in 36 individuals at marathon and halfmarathon distances (42 and 21 km) by using average velocity in the model. For such longdistance races, demand is almost entirely described by C (92%) because Caero is low in still air, ∼0.3 J · kg^{−1} · m^{−1} or 8% of the total demand, and Ckin is negligible.
Such reliability of results was also obtained (11) over shorter distances (0.8–5 km) in 16 runners of intermediate level and 27 elite athletes tested by Lacour and colleagues. The model that uses average velocity indicates that running 0.8 km in world record time (101.11 s) would mean that Caero = 0.63 J · kg^{−1} · m^{−1}, or 14% of the total demand, and Ckin = 0.16 J · kg^{−1} · m^{−1}, or 3% of the total demand.
Also by using the average velocity, Peronnet and Thibault (21) obtained a powerbalanced Eq. 1 for a 100m sprint covered in 9.95 s. Aerobic and anaerobic supply [42 J/kg (6%) and 650 J/kg (94%) above resting, respectively] were balanced with C = 386 J/kg (56%), Caero = 104 J/kg (15%), and Ckin = 202 J/kg (29%). However, it is worth noting that, in a 100m race, the initial acceleration phase is so important that average velocity is not appropriate when calculating Ckin. Considering Ckin as 0.5 · η^{−1} · V ^{2} · d ^{−1}would be strictly correct only if the final velocity attained during the race is equal to the average velocity. This is far from being the case in sprinting. There is also a problem with Caero for the same reason: Ckin and Caero are nonlinear functions of running velocity. As pointed out by Capelli et al. (5), when analyzing the energetics of cycling over short distances, an improved energetic model could be achieved when instantaneous velocities, rather than average velocity, throughout the race is considered.
An alternative mathematical procedure to that of modeld/t can be identified in the literature (25) in which supply and demand are equated by using instantaneous velocities V_{t}
. When using the model V_{t}
, Eq. 1
becomes
The model V_{t} was used with assumed supply and calculated demand so that the speed curve and running performance were predicted (25). It was also used inversely with measured distancetime data of the 100m sprint and predicted kinetics of anaerobic metabolism (27).
For the first time in Athens, on the occasion of the 1997 International Association of Athletics Federations (IAAF) World Championships in Track and Field, laser apparatus were used to obtain the speed curves of worldclass sprint runners (4). For example, the speed curves of both the male 100m world champion (MWC) and female 100m world champion (FWC) were registered.
The aim of the present study was to assess energy supply and demand in 100m world champions, considering Caero and Ckin. Because Caero and Ckin are nonlinear functions of the running velocity, our hypothesis is that a model, which uses instantaneous velocities rather than average velocity, can provide more valid estimated results. For a complete list of terms and their default values, refer to Table1.
METHODS
Van Ingen Schenau et al. (25) first proposed the theoretical model we used.
Supply in the model.
It was assumed that the kinetics of the aerobic pathway above resting follow the characteristics of a first order system from the initiation of a vigorous exercise (25) according to Paer_{t} = MAP(1 − e^{−t/τ1}), where Paer (in W/kg) represents the aerobic power above resting, MAP is the subject's maximal aerobic power above resting (in W/kg), and τ_{1} (in s) is the time constant for reaching MAP at the onset of supramaximal efforts. A MAP of 18.4 W/kg was obtained by van Ingen Schenau et al. (25) in sprinters of intermediate level who had a τ_{1} of 26 s, which was in close agreement with WardSmith and colleagues (26, 27) and Péronnet and colleagues (21,22).
In existing models of the energetics of sprinting, the anaerobic power under conditions of allout exercise is represented by Pana_{t} = Pmax · e^{−t/τ2}, where Pmax represents the maximal anaerobic power (in W/kg) and τ_{2} (in s) is the parameter governing the rate of anaerobic energy release.
Therefore, in the present analysis, supply, which includes aerobic and anaerobic components, is represented by
Demand in the model.
Demand includes parameters with numerical values fixed on the basis of prior knowledge: C, k, and η.
C, in Eqs. 1 and 2 , has been measured in a wide range of studies under submaximal conditions by using the steadystate rate of oxygen uptake at a given running velocity. C was found to be essentially constant at any speed between 10 and 20 km/h (e.g., Ref.11). Van Ingen Schenau et al. (25) used a value around 4.25 J · kg^{−1} · m^{−1}, but, above resting, it should be 4 J · kg^{−1} · m^{−1} (11,15). It is not yet known whether C remains the same at running velocities >20 km/h.
Caero (J · kg^{−1} · m^{−1}) is k · V ^{2}, and, consequently, the power lost to air friction during sprinting is k · V ^{3} (in W/kg), where k is calculated from the values of air density (ρ; in kg/m^{3}), frontal area of the runner (Af; in m^{2}), and Cd, according to k = 0.5 · ρ · Ap · Cd. The ρ is calculated by using barometric pressure (Pb; in Torr) and air temperature (T°, in °C), according to ρ = ρ_{0} · Pb · 760^{−1} · 273 · (273 + T°)^{−1}, where ρ_{0} = 1.293 kg/m is the ρ at 760 Torr (101.3 kPa) and 273°K.
Af was obtained by using the runner's mass (in kg) and height (in m) according to Af = (0.2025 · height^{0.725} · mass^{0.425}) · 0.266.
The value of Cd (0.9) is the same as that of van Ingen Schenau et al. (25).
For our calculation of k, both the mass and height of the male and female runners (75 kg and 1.75 m and 64 kg and 1.78 m, respectively) were used. Also introduced in the calculation were the Pb (760 Torr) and T° in Athens (25°C). The very limited effect of air humidity on ρ was not considered.
The men's 100m final was run with a +0.2 m/s assisting wind (w), whereas in the women's final, w = +0.4 m/s. Therefore, the power lost because of air friction was calculated as k · (V− w)^{3}, and Caero was calculated as k · (V − w)^{2}, where V − w is the runner's relative velocity to the air.
η Of converting metabolic to external work.
Van Ingen Schenau et al. (25) pointed out the necessity to consider η in powerbalanced models. Di Prampero and colleagues (10, 11) and Peronnet and colleagues (21, 22) made the same choice, although WardSmith and colleagues did not (26, 27). In the present model, according to others (10, 11, 21, 22, 25), it was considered that Caero and Ckin should include a numerical value of η. Caero in demand is Caero = η^{−1} · k · V ^{2}and Ckin = η^{−1} · [0.5V _{(t+1)} ^{2}− 0.5V_{t} ^{2}]. In the energetics models, which use the average velocity (11, 21, 22), it is usually considered that, during the acceleration phase of the sprint, the η for the transformation of metabolic energy into kinetic energy is roughly η = 0.25. Indeed, no recovery of elastic energy has been considered during this early phase in which the running velocity is low, and consequently the overall running η must approach the η of muscular contraction (11). Conversely, in association with Caero, it is usually considered that, when speed increases, η can reach 0.5 partly because of the storage and recoil of elastic energy in exercising muscles (11).
With the above considerations in mind, the iterative procedure used in the present study offers the opportunity to include in the calculation of Caero and Ckin a value of η that increases in a linear way depending on instantaneous running velocity. The value of η might increase from 0.25 within the first movements of “pushing power” to 0.5 at maximal running velocity (V _{max}; in m/s) because “reactive” power increases progressively. Thus η at any instant t is obtained by 0.25 + (0.25 · V _{max} ^{−1}) · V_{t} , where V _{max} was 11.8 m/s for the MWC and 10.7 m/s for the FWC.
Models.
The detailed calculation made with the model we used, modelV_{t}
, is
This mathematical model was compared with the model that uses modeld/t
Measurements.
The speedtime curves of the 100m MWC and 100m FWC were obtained from laser measurements (4). In brief, laser apparatus (LAVEG Sport, Jenoptik, Jena, Germany) were placed 15 m behind the start line at a height of ∼1.7 m. The laser beams were directed to the lower part of the runner's back in the upright position. Consequently, measurements were not possible during the first instants of the race when the runners were crouched in their starting blocks. Speed values were obtained from that instant when the athlete began to raise the trunk until he/she had crossed the finish line (Figs.1 and2). Any error due to the difference in height of the system in relation to the height of the lower back is considered insignificant.
The system operated at 50 Hz and measured the distance covered by the runner every 0.02 s. The system was scaled by using the exact metric distance between laser apparatus and start line (15 m) and finish line (115 m). Video cameras operating at 50 Hz were placed perpendicular to the running direction on the upper stands at the 30, 50, and 60m line, which allowed the distancetime results at regular 10m intervals over 100 m to be measured (Figs. 1 and 2). Laser and video measurements were compared, and there was a 0.10 ± 0.06m (n = 10) average difference between video and laser measurements for MWC and 0.09 ± 0.06 m (n = 10) for FWC.
Recorded speed curves of the MWC and FWC were fitted by using the differential Eq. 4 and the leastsquare method. In Eq.4 , five parameters had values fixed on the basis of prior knowledge as detailed in sections Supply in the model andDemand in the model; they are MAP, τ_{1}, C, k, and η. Two parameters were allowed to float free, Pmax and τ_{2}, so that the solver in Microsoft Excel could calculate the optimal values.
RESULTS
Running performances.
The 100m sprint finals at the Sixth IAAF World Championships in Athens 1997 were won by Maurice Greene (MWC), who recorded 9.86 s (w + 0.2) including a 0.134s reaction time, and Marion Jones (FWC), who recorded 10.83 s (w + 0.4) including a 0.160s reaction time.
Speed curve modeling.
Figures 1 and 2 show the predicted and calculated speed curves obtained by laser and video measurements of the 100m champions. The better fit with model V_{t} (Eq. 4 ) was obtained with Pmax = 90.7 W/kg and τ_{2} = 12.1 s for MWC and Pmax = 76.2 W/kg and τ_{2} = 13.2 s for FWC.
Energy cost of running.
Average velocity for MWC and FWC were, respectively, 100 m · (9.86 s − 0.134 s)^{−1} = 10.28 m/s and 100 m · (10.83 s − 0.160 s) = 9.37 m/s.
By using model d/t, the energy cost due to acceleration (Table 2) was obtained according to Ckin = η_{b} ^{−1} · 0.5 · V ^{2} · d ^{−1}, where η_{b} = 0.25, d = 100 m, andV = 10.28 m/s for MWC and V = 9.37 m/s for FWC.
Aerodynamic cost (Table 2) was obtained according to Caero = η_{a} ^{−1} · k · (V− w)^{2}, where η_{a} = 0.50 and k = 0.0036 m^{−1} · kg^{−1} for MWC and k = 0.0040 m^{−1} · kg^{−1} for FWC;V was 10.28 and 9.37 m/s for men and women, respectively, and w = 0.2 m/s for MWC and w = 0.4 m/s for FWC. The contributions to demand of C, Caero, and Ckin are presented in Table 2.
By using model V_{t} , the instantaneous changes in both Ckin and Caero were obtained (Figs.3 and 4). To compare Caero and Ckin with the values obtained with modeld/t, those instantaneous values were averaged (Table 2).
Ckin is the difference between the energy required to accelerate the body in the early phase of the race and the energy recovered during the final deceleration phase. The former amounted to 171 J/kg in MWC and 140 J/kg in FWC, and the latter to 17J /kg in MWC and 18 J/kg in FWC. This means that 10–12% of the energy of acceleration is recovered before the end of the race.
Ckin was overestimated with model d/t by 37% in MWC and by 44% in FWC (Table 2). As a consequence, modeld/t predicted 1) Pmax 20 and 14% higher than model V_{t} in MWC and FWC, respectively; 2) a higher decrease in the rate of anaerobic energy release; and 3) an 8% higher predicted contribution of anaerobic metabolism.
Caero was underestimated by 14% in both champions with modeld/t.
DISCUSSION
The present analysis of supply and demand during sprint running was based on the speedtime curves of MWC and FWC 100m world champions recorded by laser apparatus on the occasion of the 1997 IAAF World Championships (4). A mathematical model, which uses the recorded instantaneous velocities, model V_{t} , was employed to describe the components of supply and demand. Results were compared with another model, which uses the average velocity (21).
Model V_{t}: sensitive parameters.
The model we used includes seven parameters (Eq. 4 ). Five of them, MAP, τ_{1}, C, k, and η, have numerical values fixed on the basis of prior knowledge, and only two, Pmax and τ_{2}, were allowed to float free. To assess the importance of each parameter into the model, a sensitivity analysis was undertaken. By using supply characteristics obtained for MWC and FWC, the effects on predicted performance of a ±10% change to the initial value of each parameter was tested.
First, even the largest ±10% variation in MAP and τ_{1}resulted in only minor changes in predicted performance (0.02 s in MWC and 0.03 s in FWC). This is consistent with the small contribution (5%) of aerobic metabolism to energy supply obtained, according to other studies (21, 22, 2527).
Figures 5 and6 show how the final time could be affected by the other five parameters of Eq. 4 . This analysis clearly demonstrates that Pmax has the greatest influence on the predicted running time, followed by C, η, τ_{2}, and to a lesser extent k. As a consequence, the supplydemand approach is really informative only if it can be ensured that Pmax, C, η, and τ_{2} are reasonably assigned or predicted.
Parameters with assigned values.
The assigned values for C and η have been plausibly derived from previous literature (e.g., Ref. 11). However, the values are partly unsecured. This is probably the weakest area of such mathematical models.
First, the numerical value for C was fixed on the basis of experimental data obtained at submaximal running speeds, although much higher speeds are concerned in our 100m sprint model. Second, one should consider the great variability of C among different subjects in submaximal speeds. For example, in a group of middledistance runners, Lacour et al. (16) found a range of variation of 20% according to previous results (4, 24) over the range 3.5–4.2 J · kg^{−1} · m^{−1}. Assuming C amounts to ≤3.92 rather than 4 J · kg^{−1} · m^{−1} in our model, it would be enough for MWC to set a new world record (Figs. 5 and 6). The reason is that C is 63–67% of demand in 100m sprint (Table2).
Concerning η, there is no evidence in the literature that η might be around 0.25 during the most propulsive phase of acceleration. Nevertheless, a recent study (6) supported the 0.25 value, noting a lack of correlation between leg stiffness and acceleration during sprinting. The authors argued that forward power rather than reactive power is required during the acceleration phase. Conversely, as leg stiffness correlates to maximal running velocity, this could be a further argument for assuming an η of ∼0.5 at high speeds.
An η of 0.228 was used in model V_{t} used by van Ingen Schenau and colleagues (25). However, we observed that the predicted speed curves in this work were far from those that have been obtained with laser measurement. If the speed of MWC is considered, assuming η = 0.228, the predicted Pmax will be 135 W/kg and τ_{2} at 9.6 s.
According to WardSmith and colleagues (26, 27), η is not usually assigned to Caero and Ckin. They argued that the entire dissipation of thermal power is taken into account in the magnitude of C. Unfortunately, a standard value of C, e.g., 3.96 J · kg^{−1} · m^{−1}(26) was used in this model so that no additional thermal energy concerning Caero and Ckin can be taken into account. This leads to the predicted Pmax being only twofold MAP, which is not realistic in highly trained sprinters.
To conclude, there are many assumptions in the present model, particularly with regard to η.
Predicted parameters: anaerobic metabolism.
Pmax and τ_{2} were the two parameters allowed to float free in our model. This means that predicted values depend on other assumed parameters in the model. Pmax was predicted at around 90 and 75 W/kg in MWC and FWC, respectively (Table 2); at an η of 0.25, Pmax = 23 and 19 W/kg, respectively. This is very similar to the results of Pmax obtained in male sprinters by Van Ingen Schenau and colleagues (25) using cycling experimentation. Additionally, in sprinters of national level clocking around 10.6 s for men and 11.6 s for women, cycling Pmax measured in freeaccelerated conditions amounted to 20 and 17 W/kg, respectively (1). These results are a first argument against using modeld/t for sprint analysis because this model predicts a much higher Pmax (Table 2).
Again, WardSmith et al. (26) obtained values of Pmax that do not compare with those presently obtained because his mathematical framework did not include η. In the present study, Pmax (75–90 W/kg) was about four to five times higher than MAP (18 W/kg), which is realistic for highly trained sprint runners.
Although there is a consensus in the literature about using τ_{2} for ∼30 s in models (21, 22, 25, 26), lower time constants for anaerobic energy release were presently predicted in world champions (Table 2). First, this rapid decrease in anaerobic energy supply is a consequence of η increasing rapidly to 0.5 so that the balanced external work in demand is moderate. Assuming η is correct, there are many arguments for present values of τ_{2} around 12–13 s as realistic for elite sprint runners. For instance, it has been observed that the higher the initial maximal power, the higher the decline in power in the following seconds of exercise (18). This is mostly due to the great participation of the glycolysis in the first few seconds of maximal exercise (2, 12) and the inhibitory effect of proton on phosphofructokinase. As MWC and FWC both had high predicted Pmax, it is not surprising that the anaerobic rate decreased quickly.
By using a monoexponential model to describe anaerobic, energy release is approximate. Anaerobic energy supply relies actually on different components (stored phosphagens, stored oxygen, and glycolysis), each within their own, different time constants (27). The following may be reasonable estimates of time constants for each component: phosphagens = 9 s, stored oxygen = 3 s, and glycolysis = 35 s (e.g., Ref. 8). When most of the measured rates of release are for longer periods, typically 30–210 s (19, 25), the overall time constant will be dominated by the rate of release of glycolytic energy. Over shorter periods, the more rapid components will dominate. So, fitted over 10 s, time constants are likely to be shorter than when applied to longer periods of exercise. Additionally, by allowing for a faster rate of release in specifically trained athletes, and particularly a faster rate of release of energy from stored phosphagens, an overall time constant of 12–13 s seems very likely.
Taking into account the predicted Pmax and τ_{2} (Table 2), we obtained an anaerobic contribution to energy supply amounting to 607 and 557 J/kg in MWC and FWC, respectively. On the basis of the direct metabolic approach of anaerobic ATP turnover in 6 and 10 s of maximal exercise (2, 12), these values might be underestimated. By making space for ATP derived from oxygen stores (i.e., attached to hemoglobin and myoglobin and available in the lungs), which probably amounts to 6 ml/kg (120 J/kg) (19) and has a rapid half time in use, the strictly anaerobic ATP turnover might be 490 J/kg for MWC and 440 J/kg for FWC. However, direct examination of muscle metabolites (2) provided 130 mmol ATP turnover (±5,200 J) per kilogram active muscle mass (5,200 J/kg) in 10 s. By assuming that a quarter of the body mass is highly active during sprinting, this leads to an ATP turnover of 1,300 J/kg, a threefold higher value than that presently predicted in worldclass 100m runners.
Others have used postcompetition blood lactate concentrations as indicators of anaerobic energy expenditure during 400 and 800m races (15). We are aware of post100m peak blood lactate concentrations ([Lac]_{b}) noticeably different between Caucasian and African runners. In a group of Caucasian sprint runners with performance times of 10.54–10.69s, [Lac]_{b}reached 14.6–16 mM (17), whereas in Africans with an average time of 10.70s, [Lac]_{b} reached only 8.5 mM (13). On the basis of calculations made by others (15, 19), the amount of energy yielded by the glycolysis in Africans could be ±500 J/kg. Supply, obtained as the sum of oxygen stores (120 J/kg), glycolysis in Africans (±500 J/kg), and depletion in phosphagens (≥400 J/kg), is higher than the supply predicted in our model. Nevertheless, we can conclude that our understanding of whole body anaerobic capacity is too limited. Perhaps modeling the running performance is the most appropriate approach to study anaerobic energy supply in brief maximal exercises as suggested by others (27).
Energetics of sprint running.
The energetics of sprint and middledistance running might differ for several reasons. First, because of the comparative high speeds sustained in sprint running, aerodynamic resistance is a nonnegligible component of demand. Second, and more importantly, the cost of acceleration (Ckin) is most likely to be high in worldclass sprint runners. This is because 1) the speed increases from 0 to 9 m/s in <1.8 s in men and from 0 to 8 m/s in <1.9 s in women and2) the acceleration phase represents ∼60% of the race (4). Accordingly, the contribution of Ckin amounted to 20–25% of the total demand in our work (Table 2, Figs. 3 and 4). This means that the model used to estimate Ckin has to be well defined, as a small error in Ckin can lead to quite large discrepancies in both predicted supply and demand. This confirms that using modeld/t, where Ckin is not correctly defined, discourages the modeling of the energetics of short running distances. There is, as yet, no reason to reject model d/tfor distances ≥0.8 km on the basis of Ckin being ≤3% of demand. Unfortunately, it might not be possible to create a model for 200 and 400m races. As both are not maximal exercises, modeling the anaerobics by using a simple firstorder system is obviously not correct. Furthermore, Ckin might be incorrectly approximate with modeld/t, taking into account its relative contribution to the energy demand. Finally, Caero could be of importance at such high speeds, but the effect of wind could be difficult to assess as the runner changes direction during the race.
The model V _{t} we used makes it possible to obtain instantaneous changes during the race in both Ckin and Caero (Figs. 3and 4). Interestingly, after ∼6 s of running, both MWC and FWC begin to decelerate. This means that they are paying back kinetic energy so that part of Caero requires no more supply. Demand is almost entirely set by C, so that, in practical terms, the runner just has to replace his or her limbs in the correct position before contact with the ground and bounce on the track. Consequently; it is not surprising that muscle stiffness, evaluated by using vertical rebounds, usually correlates with running at maximal velocity (6, 17).
The estimated Caero amounted to 80 J/kg (Table 2). This is fairly plausible on the basis of wind tunnel or other experiments (7,14). Any changes in body size, Cd, or ρ would affect Caero through changes in k, as shown by k = 0.5 · ρ · Af · Cd (see methods). We show that a ±10% change in k leads to ∼0.07 s advantage in running time (Figs. 5 and 6). Especially in worldclass women, we noted some difference in height of comparable body mass. In the recent 2001 World Championship in Edmonton, Marion Jones (FWC in the present study; weight = 64 kg, height = 1.78 m) lost her title to Zhanna Pintussewich (weight = 64 kg, height = 1.64 m). On the basis of their 6% difference in Af, the present model predicts a 0.05s advantage to Pintussewich in still air or moderate assisting wind. However, with a head wind of 2 m/s, this advantage reaches 0.1 s. Height in sprint runners can be an advantage with regard to stride length, but on the basis of its role in Caero, it can be a disadvantage.
Differences in body mass affect the energetics of sprint running in a more complex manner. The supplydemand equation is balanced in J/kg body mass in our work. Therefore, any increase in body mass, which is less involved in power generation, would increase demand without changing supply. Increased upper body mass would increase both Ckin and Caero but in different proportions. We assessed that, in MWC, an additional 5 kg upper body mass would increase Ckin by ±5% and Caero by 0.4%, resulting in a performance time of 10.03 s (Fig.7), which is obviously the lower limit for a worldclass running time. Inversely, supposing that upper body muscle mass has few other advantages, a 5kg decrease resulted in a running time of 9.69 s (new world record) mainly due to the advantage of a 5% reduced Ckin.
More than body mass itself, the repartition of that mass seems to affect C. Previous investigations (20, 23) demonstrate that adding external masses to limb extremities widely increases the energy cost of moving forward. In sprinting events in which extremities are successively accelerated and decelerated at high rates, the lighter the extremities are, the lower C is.
To conclude the effect of body size on sprintrunning performance, height presents no advantage in terms of energetics, but rather the runner's mass should power muscles that are as proximal as possible. It is interesting to note that runners increase in height and mass as their running distance increases from 60 to 400 m. It may be that Ckin is very important over short distances, as our analysis suggests, but as distance increases, size also confers other advantages, such as better reuse of elastic energy (6).
In conclusion, speed curves of world champions can be successfully used to model the energetics of sprint running. Our model provides realistic insights on components of both demand and supply, ensuring that Pmax, the η of converting metabolic into external work, and C can be reasonably assumed or predicted. Unfortunately, η and C are unsecured, whereas Pmax receives supporting results in the literature.
A MWC who runs 100 m in 9.86 s might yield 640 J/kg, including 95% anaerobic energy to balance an energy demand composed of 24% Ckin and 13% Caero. The remaining cost of C in J · kg^{−1} · m^{−1} was supposed to be the same as that required in running middle or long distances. A high initial level of power output, amounting to at least 90 W/kg in men and 75 W/kg in women, might be a prerequisite for toplevel results. Little is known about the rate of decrease in maximal power during the race. A time constant for anaerobic energy release, amounting to 12–13 s, has not been confirmed. Neither height nor body mass seems to have any advantage in the energetics of sprint running.
Footnotes

Address for reprint requests and other correspondence: L. M. Arsac, Faculté des Sciences du Sport, Université Victor Ségalen Bordeaux 2, Domaine Universitaire, 12 Ave. Camille Jullian, 33607 Pessac cedex, France (Email:laurent.arsac{at}ubordeaux2.fr).

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

First published November 30, 2001;10.1152/japplphysiol.00754.2001
 Copyright © 2002 the American Physiological Society