Although our ancestors ran on natural surfaces without shoes, most modern recreational and competitive runners do so on artificial surfaces with cushioned shoes. In running, the leg muscles generate force to cushion the impact with the ground. External cushioning (surfaces or shoes) may reduce the muscular effort needed for cushioning and thus reduce the metabolic cost of running (13). To test this idea, we specifically investigated how the cushioning properties of surfaces and shoes affect the metabolic cost of running.
The elastic and viscoelastic properties of surfaces (17,18,20) and treadmills (15,16) can negatively, neutrally, or positively affect the metabolic cost of running. For example, Lejeune et al. (17) found that running on sand was 1.6 times more expensive than on a firm floor surface. In contrast, Pugh (20) found no metabolic difference between running on an artificial rubberized track and a traditional cinder track. But, when McMahon and Greene (18) designed and built a “tuned” indoor running track with substantial elastic recoil, they found that competitive times for the distance running events were faster on the new track, suggesting reduced metabolic cost. Classic and modern research-grade treadmills have rigid decks, comparable in vertical stiffness to asphalt or concrete surfaces. However, many modern treadmills, especially those used for fitness, have decks with fixed or adjustable stiffness and damping qualities that appear to increase the metabolic cost of running (15). In contrast, Kerdok et al. (16) built a unique research treadmill with adjustable vertical stiffness and minimal damping which reduced the metabolic cost of running by as much as 12% with a surface deflection of ∼2 cm. Kerdok et al. also found that the subjects ran with less flexed knees on the low-stiffness treadmill, which presumably reduced the knee extensor torque required and thus reduced the metabolic cost of generating force with the quadriceps muscles.
Like surfaces and treadmills, running shoes have been shown to have negative (8,9,19), neutral (8,13,20,22), and positive (13) effects on the metabolic cost of running. For example, Perl et al. (19) found that running in heavy shoes (with substantial damping properties) required more metabolic energy than lightweight, “minimal” shoes. Frederick et al. (14) established that shoe mass incurs a predictable metabolic penalty (1% per 100 g per shoe). However, in a different study, Frederick et al. (13) observed that, despite the mass of the shoes, the submaximal rate of oxygen consumption (V˙O2) did not differ between running in well-cushioned shoes and barefoot. Note that the treadmill used in that study had a rigid deck. To try and explain their results, Frederick et al. (13) hypothesized that submaximal V˙O2 during barefoot running includes a “cost of cushioning” the body and thus lightweight, well-cushioned shoes might reduce the metabolic cost of running. Indeed, Franz et al. (11) showed that when shoe mass is factored out, cushioned shoes can require 3%–4% less metabolic power than running without shoes. However, Franz et al. could not definitively attribute the energy savings specifically to the shoe cushioning.
Differences in footstrike (e.g., rearfoot vs midfoot) and/or other shoe-related factors (e.g., heel height, flexibility, motion control elements) could conceivably enhance or blunt any metabolic cost savings due to cushioning. Thus, we designed an experiment to isolate and measure the metabolic effects of shoe cushioning in a novel way—by attaching the same foam used in running shoe midsoles to the belt of a rigid-decked treadmill. This approach isolated the independent variable (cushioning) and eliminated possible confounding shoe construction factors. This allowed us to measure the effects of shoe cushioning without having our subjects wear shoes. We first hypothesized that the metabolic cost of unshod running would decrease on a cushioned surface. We also hypothesized that, on the normal rigid treadmill surface, unshod running would have approximately the same metabolic cost as running with lightweight, cushioned running shoes due to counteracting effects of shoe mass and shoe cushioning.
We present data for 12 healthy runners (10 M/2 F; mean ± SD, age = 30.2 ± 9.1 yr, mass = 68.5 ± 6.5 kg, and height = 174.5 ± 5.9 cm). These subjects reported running an average of 79.4 ± 60.5 km·wk−1, of which 59.5 ± 50.0 km·wk−1 (range = 11–177 km·wk−1) were barefoot or in minimal shoes. Subjects reported that their typical training speed averaged 3.5 ± 0.6 m·s−1 (range = 2.9–4.8 m·s−1). Our sample size was based on the recommendations of Frederick (12), who reported that, with an expected coefficient of variation of 1.5%–2% for repeated within-day measurements of oxygen uptake, a 1%–2% mean difference could be resolved with a sample size of 10–15 subjects. Indeed, with careful attention to detail, Roy and Stefanyshyn (21) were able to discern economy differences of 1% between shoe conditions using a sample size of 13. Thus, we collected data for 14 subjects. However, we had to exclude the data from two subjects because they exhibited RER > 1.0, indicating that they were not in metabolic steady state (3). Subject inclusion criteria were as follows: >18 yr of age; midfoot strike preference both shod and unshod; run at least 25 km·wk−1, of which at least 8 km·wk−1 were barefoot or in minimal running footwear (e.g., Vibram Five Fingers®) for at least 3 months; injury free; self-reported ability to sustain 3.3 m·s−1 running pace for at least 60 min; and meeting the medical criteria of the American College of Sports Medicine for minimal risk for exercise (1). On the basis of subject reports and our inclusion criteria, completing the experimental protocol was of low to moderate intensity and duration for all subjects. We included only runners with a midfoot strike pattern because asking runners to rearfoot strike without shoes on a hard treadmill surface might have increased the risk of injury. The University of Colorado Institutional Review Board approved the study protocol, and all subjects gave their written consent after being informed of the nature of the study.
To verify that the subjects preferred to run with a midfoot strike pattern (4), we asked them to run at their typical training pace across a 30-m runway equipped with a force platform (Advanced Mechanical Technology Inc., Watertown, MA) to which a sheet of paper was affixed. We attached small pieces of felt marker to each subjects’ right foot at 90%, 70%, and 33% of their foot length (measured between the heel and the distal end of the second toe). We collected the force plate data at 1000 Hz and tracked the center of pressure relative to the data points provided by the ink dots left on the paper as per Cavanagh and Lafortune (4). We classified subjects as midfoot strikers if the center of pressure at initial contact was between 33% and 70% of foot length and rearfoot strikers if the center of pressure originated posterior to the 33% mark (4).
During a single experimental session, subjects completed a 5-min standing trial, a 10-min unshod running acclimation trial (with no surface cushioning), and then four 5-min running trials. A 3-min rest period separated each of the running trials. In all running trials, subjects ran at a speed of 3.35 m·s−1 on a Quinton 18-60 motorized treadmill (Quinton Instrument Company, Bothell, WA) that we modified to have a calibrated digital readout for speed. Note that this treadmill has a rigid steel deck and a thin belt with no significant cushioning or damping properties. For the duration of the experiment, subjects wore very thin, slip-resistant yoga socks for traction and hygienic purposes.
In random order, subjects completed one shod (Nike Free 3.0 V2; ∼211 g per shoe) running condition on the normal treadmill belt surface and three unshod running trials: on the normal treadmill belt surface (Unshod 0 mm), with 10-mm-thick slats of foam attached to the belt (Unshod 10 mm), and with 20-mm-thick slats of foam attached to the belt (Unshod 20 mm) (Fig. 1). The foam slats (length × width × thickness; 18.8 cm × 33.7 cm × 10 mm and 21.4 cm × 37.3 cm × 20 mm) consisted of the same material used in the midsole of the Nike Free running shoes (Phylite®; 60% Phylon® and 40% rubber with an Asker Type C durometer reading of 52–58). We drilled two 2.5-cm-diameter holes along the left and right lateral edges of each foam slat through which we sewed short loops of 2.5-cm-wide nylon webbing. We sewed continuous strips of hook Velcro® to two 2.5-cm-wide straps of nylon webbing (one left and one right) and routed the strap through the small loops. We glued strips of the hooked part of the Velcro® to the left and right edges of the treadmill belt. Thus, the slats of foam could be easily put in place and removed. Overall, we created a “tank-tread” of foam slats that covered the entire length of the treadmill belt.
During the running trials, we offered verbal instructions to each subject to maintain a midfoot strike pattern whether shod or unshod. Further, we confirmed foot strike throughout each trial visually as well as with high-speed video recordings (Casio EX-FH20; 210 frames per second). We did not control the stride frequency or stride length so as to compare normal unshod and shod running. We determined each subject’s contact time and stride frequency from the video recordings using Windows Movie Maker (Microsoft Corporation, Redmond, WA) averaged over five consecutive strides.
During the standing and running trials, we used an open-circuit respirometry system (TrueOne 2400; Parvo Medics, Sandy, UT) to analyze the subject’s expired gases and calculate the STPD rates of oxygen consumption (V˙O2) and carbon dioxide production (V˙CO2). Before each experiment, we calibrated the system using reference gases and a 3-L syringe. We averaged V˙O2, V˙CO2, and RERs for the last 2 min of each 5-min trial. As noted, two subjects had to be excluded because their RER values exceeded 1.0. The RER values for each of the 12 remaining subjects were below 1.0. We report not only gross V˙O2 values in milliliters per kilogram per minute (mL·kg−1·min−1) but also the average standing value (mean ± SD: 4.84 ± 0.39 mL·kg−1·min−1) to allow calculation of net V˙O2. We normalized V˙O2 and V˙CO2 using the subject’s body mass while unshod. From V˙O2 and V˙CO2, we calculated gross metabolic power in watts per kilogram (W·kg−1) using Brockway’s equation (2). We agree with Fletcher et al. (10) who suggested that metabolic power is more representative of running economy than V˙O2 alone, but we report both metabolic power and V˙O2 for the convenience of the reader.
A Shapiro–Wilk test and Mauchly test of sphericity respectively confirmed that metabolic cost, contact time, and stride frequency were normally distributed (P > 0.24, P > 0.08, and P > 0.07, respectively) and each had equal variance across conditions (P = 0.63, P = 0.15, and P = 0.62, respectively). A repeated-measures ANOVA then tested for significant main effects of cushioning (0, 10, and 20 mm) on V˙O2, gross metabolic power, contact time, and stride frequency. When a significant main effect was detected, we performed post hoc pairwise comparisons. We also compared shod and unshod conditions using paired-samples t-tests. We used a criterion of P < 0.05 for statistical significance.
Treadmill surface cushioning significantly decreased V˙O2 and metabolic power. On average, V˙O2 and metabolic power for unshod running were 1.47% (P = 0.015) and 1.63% (P = 0.034) less on 10 mm of foam cushioning compared to the rigid surface, respectively (Table 1 and Fig. 2). However, those measures for running on 20 mm of foam cushioning were not significantly different from those for running on the rigid surface (P = 0.602 and P = 0.605, respectively). We did find considerable individual variation with respect to the effect of surface cushioning on metabolic demand (Table 2 and Fig. 3). However, 10 of the 12 subjects had lower V˙O2 values and 8 of the 12 subjects required less metabolic power for the 10-mm-thick foam surface compared to the rigid surface. On the rigid treadmill surface, V˙O2 and metabolic power for running unshod and shod were not significantly different (P = 0.533 and P = 0.182, respectively).
Average stride frequencies for unshod running on the cushioned surfaces were not significantly different from unshod running on the normal rigid surface, but stride frequency was a modest 2.5% slower during shod running (P < 0.01) (Table 1). Hence, stride length was 2.5% longer (average of ∼5 cm). Likewise, ground contact times were not different between the unshod conditions, but contact time was 5.9% longer during the shod condition (P < 0.01) (Table 1).
In this study, we tested the metabolic cost of cushioning hypothesis, which states that running involves a “cost of cushioning” the body against impact (13). Specifically, we quantified the isolated effects of shoe cushioning on the metabolic cost of running, while controlling for footstrike pattern, barefoot/minimalist running experience, and footwear. Supporting our first hypothesis, we found that, on average, the metabolic cost of unshod running was significantly reduced when subjects ran on a 10-mm-thick foam-cushioned surface compared to a normal rigid treadmill surface. Supporting our second hypothesis, on the normal, rigid treadmill surface, the metabolic cost of unshod running was not significantly different from running with lightweight, cushioned running shoes.
To further clarify, our first hypothesis stated that the metabolic cost of unshod running would decrease with a cushioned surface. While 10 mm of surface cushioning did elicit a lower metabolic cost than the rigid treadmill surface alone, 20 mm of surface cushioning did not, on average, further reduce metabolic cost. We suspect that there may be an optimal cushioning thickness for each individual, which minimizes his/her metabolic power demand. This optimum likely depends on many factors including cushioning hardness (durometer), body mass, and footstrike preference.
The elastic and viscoelastic (damping) properties of running shoes and surfaces can combine to influence the metabolic cost of running. Kerdok et al. (16) found that the metabolic cost of running on an elastic, adjustable-stiffness treadmill steadily decreased with decreased stiffness. In contrast, we did not find that the metabolic cost of running steadily decreased with thicker foam cushioning. It is likely that the treadmill surface of Kerdok et al. had much less damping than our foam surfaces and thus, that of running shoes. Indeed, excessive damping may have negative metabolic effects. In another treadmill running study, Hardin et al. (15) reported that metabolic cost increased with a lower-stiffness treadmill surface that also had greater damping. Treadmill surface properties should be considered when interpreting studies of footwear energetics and biomechanics and when designing future studies.
Our data also support our second hypothesis, that is, unshod running would have approximately the same metabolic cost as running with lightweight shoes due to counteracting effects of cushioning and mass. In a previous study from our laboratory, Franz et al. (11) investigated the effects of adding mass to the feet on the metabolic cost of shod and unshod running. In accordance with the classic findings from Frederick et al. (14), Franz et al. (11) found that every 100 g of mass added to each foot increased the rate of oxygen consumption by ∼1%, both with and without shoes. Based on this “1% rule” alone, we would expect that running in the 210-g shoes used in the present study would be 2.10% more expensive than running unshod on the normal rigid treadmill surface. However, we found that those conditions elicited similar metabolic costs. Running unshod on the 10 mm of foam cushioning (approximately the thickness of the shoe midsole in the forefoot region) afforded an energetic savings of 1.63%. Thus, it appears that the positive effects of shoe cushioning counteracted the negative effects of added mass, resulting in a metabolic cost for shod running approximately equal to that of unshod running.
While our stride kinematics results for unshod versus shod conditions were similar to those of previous studies, our data for unshod running on cushioned surface conditions provide new insight. We found that stride frequency was 2.5% slower for shod running, which is similar but somewhat less than the 3.3%, 3.4%, 3.9%, 5.1%, and 5.7% values reported by Franz et al. (11), Divert et al. (8), De Wit et al. (6), Divert et al. (7), and Squadrone and Gallozzi (22), respectively. Those studies varied in what factors were controlled for (e.g., footstrike type). A slower stride frequency at a fixed speed equates to longer strides while running shod compared to unshod. What aspects of shoes are responsible for the longer strides? Franz et al. (11) found that the longer strides were not due to shoe mass. Other authors have suggested that shorter strides are selected during barefoot running to reduce loading or, in other words, shoe cushioning allows for longer strides with similar loading. In the present study, we found that stride frequency did not differ between unshod running on the hard surface and the cushioned surfaces. Thus, the stride frequency/length differences between unshod versus shod running cannot be attributed to the cushioning properties of the shoe. Similarly, our contact time data for unshod versus shod running were akin to those of previous studies. We found that contact time was 5.9% longer for shod running, similar to the 2.4%, 4.1%, 5.0%, 5.7%, and ∼9% differences reported by Divert et al. (7), Kerdok et al. (16), De Wit et al. (6) Divert et al. (8), and Clarke et al. (5), respectively. But again, we found that running unshod on hard and cushioned surfaces had similar contact times. Thus, the contact time differences between unshod and shod cannot be attributed to the cushioning properties of the shoe. These topics may deserve further investigation.
Our study had several limitations. First, to maintain consistency, all subjects ran in the same model of running shoes; therefore, our findings may not translate to other running shoe models. Simply due to the demographics of our volunteers, our subjects were predominately male. However, we have no reason to expect different results for female runners. Finally, we only studied two thicknesses of one specific foam cushioning material.
Our results have implications for the design of running shoes and/or track surfaces. Many competition track surfaces are extremely hard, presumably to enhance sprint performance. Despite the prevalent track hardness, spiked shoes designed for middle- and long-distance track running events have almost no cushioning under the midfoot and forefoot. Our results suggest that distance running spikes with midfoot/forefoot cushioning (or the use of racing flats) could enhance performance. Alternatively, track surfaces could be made with soft, yet elastic properties, allowing athletes to run without shoes. However, given the interindividual variability that we noted, the benefits of a soft track would not be uniform and thus possibly considered to be unfair. Our data suggest that the design of competition shoes for road racing on paved surfaces should not overemphasize weight minimization at the expense of cushioning.
Future studies on the energetics, biomechanics, and neuromuscular control of running on different surfaces could be fruitful. For example, to identify the metabolically optimal thickness of foam cushioning, future studies could compare more experimental conditions (e.g., 5, 10, 15, 20, 25 mm). A complementary approach would be to compare foam surfaces of the same thickness (e.g., 10 mm) but with different hardness properties (i.e., durometer values). Because small reductions in the metabolic cost of running are most meaningful for competitive runners, it would be useful to repeat our study at faster running speeds on more aerobically fit subjects. In addition, it may be interesting to study how runners adapt over time to different cushioned surfaces. Further, we have not yet elucidated the biomechanical or neuromuscular basis for why metabolic cost was reduced on the 10-mm-thick cushioned surface. EMG measurements may be able to detect a reduction in muscle activity, but the trial-to-trial variability may preclude the detection of small differences.
In summary, we found that a moderate thickness of foam cushioning generally reduced the metabolic cost of running. In addition, the metabolic cost of running did not differ between unshod and shod conditions, presumably because the positive effect of cushioning was counteracted by the negative effect of shoe mass. Overall, our data provide clear support for the cost of cushioning hypothesis of Frederick et al. (13).
Nike, Inc., donated the foam cushioning and shoes used in this study but was not involved in the conception, planning, design, or interpretation of the study. Rodger Kram is a paid consultant for Nike, Inc.
The results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. American College of Sports Medicine. ACSM’s Guidelines for Exercise Testing and Prescription
. 7th ed. Baltimore (MD): Lippincott Williams & Wilkins; 2006. p. 368.
2. Brockway JM. Derivation of formulae used to calculate energy expenditure in man. Hum Nutr Clin Nutr
. 1987; 41C: 463–71.
3. Brooks GA, Fahey TD, Baldwin KM. Exercise Physiology: Human Bioenergetics and Its Applications
. 4th ed. New York (NY): McGraw Hill; 2004. p. 928.
4. Cavanagh PR, Lafortune MA. Ground reaction forces during distance running. J Biomech
. 1980; 13: 397–406.
5. Clarke TE, Frederick EC, Cooper LB. Effects of shoe cushioning upon ground reaction forces in running. Int J Sports Med
. 1983; 4 (4): 247–51.
6. De Wit B, De Clercq D, Aerts P. Biomechanical analysis of the stance phase during barefoot
and shod running. J Biomech
. 2000; 33: 269–78.
7. Divert C, Mornieux G, Baur H, Mayer F, Belli A. Mechanical comparison of barefoot
and shod running. Int J Sports Med
. 2005; 26 (7): 593–8.
8. Divert C, Mornieux G, Freychat P, Baly L, Mayer F, Belli A. Barefoot
-shod running differences: shoe or mass effect. Int J Sports Med
. 2008; 29 (6): 512–8.
9. Flaherty RF. Running economy and kinematic differences among running with the foot shod, with the foot bare, and with the bare foot equated for weight [dissertation]
. Springfield (MA): Springfield College; 1994. p. 106.
10. Fletcher JR, Esau SP, MacIntosh BR. Economy
of running: beyond the measurement of oxygen uptake. J Appl Physiol
. 2009; 107: 1918–22.
11. Franz JR, Wierzbinski CM, Kram R. Metabolic cost of running barefoot
versus shod: is lighter better? Med Sci Sports Exerc
. 2012; 44 (8): 1519–25.
12. Frederick EC. Measuring the effects of shoes
and surfaces on the economy
of locomotion. In: Nigg BM, Kerr BA, editors. Biomechanical Aspects of Sport Shoes and Playing Surfaces
. Calgary (Canada): University of Calgary; 1983. pp. 93–106.
13. Frederick EC, Clarke TE, Larsen JL, Cooper LB. The effect of shoe cushioning on the oxygen demands on running. In: Nigg BM, Kerr BA, editors. Biomechanical Aspects of Sports Shoes and Playing Surfaces
. Calgary (Canada): University of Calgary; 1983. pp. 107–14.
14. Frederick EC, Daniels JT, Hayes JW. The effect of shoe weight on the aerobic demands of running. In: Bachl N, Prokop L, Suckert R, editors. Current Topics in Sports Medicine
. Vienna (Austria): Urban & Schwarzenberg; 1984. pp. 616–25.
15. Hardin EC, van den Bogert AJ, Hamill J. Kinematic adaptations during running: effects of footwear, surface, and duration. Med Sci Sports Exerc
. 2004; 36 (5): 838–44.
16. Kerdok AE, Biewener AA, McMahon TA, Weyand PG, Herr HM. Energetics
and mechanics of human running on surfaces of different stiffnesses. J Appl Physiol
. 2002; 92: 469–78.
17. Lejeune TM, Willems PA, Heglund NC. Mechanics and energetics
of human locomotion on sand. J Exp Biol
. 1998; 201 (Pt 13): 2071–80.
18. McMahon TA, Greene PR. Fast running tracks. Sci Am
. 1978; 239 (6): 148–63.
19. Perl DP, Daoud AI, Lieberman DE. Effects of footwear and strike type on running economy
. Med Sci Sports Exerc
. 2012; 44 (7): 1335–43.
20. Pugh LG. Oxygen intake in track and treadmill running with observations on the effect of air resistance. J Physiol
. 1970; 207: 823–35.
21. Roy JP, Stefanyshyn DJ. Shoe midsole longitudinal bending stiffness and running economy
, joint energy, and EMG. Med Sci Sports Exerc
. 2006; 38 (3): 562–9.
22. Squadrone R, Gallozzi C. Biomechanical and physiological comparison of barefoot
and two shod conditions in experienced barefoot
runners. J Sports Med Phys Fitness
. 2009; 49 (1): 6–13.
Keywords:© 2014 American College of Sports Medicine
ECONOMY; ENERGETICS; ENERGY COST; SHOES; BAREFOOT