Grounded Running Reduces Musculoskeletal Loading : Medicine & Science in Sports & Exercise

Journal Logo


Grounded Running Reduces Musculoskeletal Loading


Author Information
Medicine & Science in Sports & Exercise 51(4):p 708-715, April 2019. | DOI: 10.1249/MSS.0000000000001846



The authors of “Grounded Running Reduces Musculoskeletal Loading” (1) found an error in their method to determine running speeds resulting in an overestimation of the reported values. The following values should have been used in the manuscript:

As such, the second paragraph of the Introduction should read:

The running population is diverse spanning a range from novice runners to elite athletes. Besides running experience, there is also a large range in fitness levels, running styles, distances run, and running speeds. To illustrate the latter, one of the most popular running applications (4) shows that 50% of the male participants of 5-km events ran with an average speed between 9.1 (2.53 m·s−1) and 12.0 km·h−1 (3.33 m·s−1), 25% with a speed below 9.1 km·h−1, and 5% even below 7.0 km·h−1 (1.94 m·s−1). Surprisingly, distance running speeds are hardly dependent on the selected distance, except for very long distances such as marathons, in which the running speeds tend to be slower (4). This means that a considerable number of runners (>25% of the entire running population) run at slow speeds; that is, below 9.1 km·h−1, which is typically below the speeds investigated in biomechanical running research, as referred to by Schache et al. (5).

The error does not relate to results or interpretation of the conducted experiment in which grounded running was compared to normal running at the same speed. The main message of this paragraph is that “a considerable number of runners run at slow speeds” and correct determination of the running speeds results in slower running speeds which even reinforces this message.

Medicine & Science in Sports & Exercise. 51(8):1794, August 2019.

Distance running is one of the most popular moderate- to vigorous-intensity recreational activities (1), with associated health benefits given as one of the main reasons to go for a run. Indeed, recreational running has a significant effect on mental and physical health leading to increased life expectancy of about 3 yr on average (2). Another reason why running is so appealing is because of its ease of practice, making it one of the most accessible sports. However, potential disadvantages are running-related injuries (RRI) (3).

The running population is diverse, spanning a range from novice runners to elite athletes. Besides running experience, there is also a large range in fitness levels, running styles, distances run, and running speeds. To illustrate the latter, one of the most popular running applications (4) shows that 50% of the male participants of 5-km events ran with an average speed between 10.1 and 13.0 km·h−1 (2.8 to 3.6 m·s−1), 25% with a speed below 10.1 km·h−1, and 5% even below 8.1 km·h−1 (2.25 m·s−1). Surprisingly, distance running speeds are hardly dependent on the selected distance, except for very long distances such as marathons, in which the running speeds tend to be slower (4). This means that a considerable number of runners (>25% of the entire running population) run at slow speeds, i.e., below 10.1 km·h−1, which is typically below the speeds investigated in biomechanical running research, as referred to by Schache et al. (5).

To distinguish running from walking (WK), traditionally, four criteria are used (6): 1) the presence of a flight phase during which none of the feet make contact with the ground, 2) a vertical ground reaction force (vGRF) pattern showing a single maximum situated around midstance, 3) a maximal bent knee around midstance, and 4) in-phase fluctuations of kinetic and gravitational potential energy. Although in general these criteria coincide, a flight phase is sometimes lacking when people run slowly (7). According to the first mentioned distinction, the adopted locomotion pattern should be classified as WK. However, running without a flight phase seems to behave as a spring-mass model with a flexion–extension cycle of the supporting limb, making it very different from the typical WK inverted pendulum motion (7). Also, animals such as quails (8), ostriches (9), and gibbons (10) use a gait that shows the characteristics of a spring-mass model but do lack flight phases. On the basis of this observation, Vereecke et al. (10) suggest that the presence of a flight phase should not be used to distinguish between WK and running. As such, animal literature uses the term GR for this type of locomotion, but it seems that it also applies to human locomotion.

Recent observations by Shorten and Pisciotta (7) quantified the prevalence of GR at 11.25 km in the Portland marathon. They found that 16% of all participants performed GR. This observation also showed that GR mainly occurs when people run slowly. Approximately 50% of all grounded runners finished in 5 h 15 min to 7 h 45 min, which means they adopted an average speed between 5.5 and 8.0 km·h−1 (1.53 to 2.22 m·s−1). These observations made clear that GR is not a rare phenomenon and that GR mainly occurs at slow running speeds.

In this study, we aimed to confirm GR as a running gait from a dynamic spring-mass perspective, despite the absence of a flight phase. We hypothesized that the ground reaction forces (GRF) of GR will show a single hump pattern with its maximum around midstance (11), that the timing of maximal knee flexion will also occur around midstance, and that gravitational potential energy will be in-phase with kinetic energy (hypothesis 1). Key differences between slow running at 2.10 m·s−1 with and without a flight phase within the same subjects will be identified as this might lead to hypotheses regarding the reasons why certain people opt for GR instead of running with a flight phase at slow speeds. We will first focus on biomechanical aspects. We hypothesized that compared with running with a flight phase, GR would show lower values for external power (a general measure for muscular loading) and maximal vGRF (a general measure for both muscular and skeletal loading), which combined give an idea of the general musculoskeletal loading (hypothesis 2); lower vertical instantaneous loading rates (VILR) and tibial accelerations (TA), which are measures for impact intensity (hypothesis 3); and lower muscle stresses and peak eccentric power in the ankle and knee extensors, which are more explicit measures of muscular loading (hypothesis 4). Besides these biomechanical parameters, we also measured energy expenditure. We expected this measure to be lower for GR compared with running with a flight phase at the same slow speed (hypothesis 5). In addition, running at the population average running (PAR) speed (4), which is 11.50 km·h−1 (3.20 m·s−1), will be incorporated in the study. This speed will serve as a reference for the comparison between slow running at 7.50 km·h−1 (2.10 m·s−1) with and without a flight phase, placing the differences in perspective to a reduction in running speed.



Thirty male subjects who were active in sports participated in the experiment. We selected athletic subjects as they were able to run for 1 h continuously, which ensured that they could finish the protocol. Likely, people who naturally perform GR would not have been able to fulfill the protocol, which required participants to run at a speed of 3.20 m·s−1, a speed well above their preferred running speed. Three subjects were removed from data analysis as they showed duty factors (DF: stance time divided by stride time; DF <50% indicates a flight phase) well below 50% (45.25% ± 1.33%) and flight times (0.038 ± 0.014 s) that were situated within these of slow aerial running (SAR) at 2.10 m·s−1 when asked to perform a GR. As such, 27 subjects were retained for further analysis (age = 23.0 ± 1.9 yr, mass = 74.9 ± 6.0 kg, length = 1.81 ± 0.05 m). The experimental protocol was approved by the ethical committee of the Ghent University Hospital, and written informed consent was obtained from all subjects.

Experimental design

All subjects wore a neutral running shoe (Mizuno Wave Rider 18) provided by the experimenters. The experiment was executed on a force-instrumented, split-belt treadmill (Bertec Corp., Columbus, OH). It started with 4 min standing rest to measure metabolic baseline values, followed by a habituation protocol to become familiar with the experimental setup and the experimental conditions. During the habituation protocol, subjects walked and ran for 5 min at speeds ranging from 1.40 to 3.20 m·s−1. Then, subjects were instructed to perform GR at 2.10 m·s−1 for 5 min based on a short demonstration by the researchers and the simple verbal instruction “run without a flight phase.”

During the actual experiment, three conditions were performed in random order. Each condition lasted for 5 min with 3 min of rest in between: GR (at 2.10 m·s−1), SAR (at 2.10 m·s−1) and aerial running at the (PAR at 3.20 m·s−1).


Subjects performed SAR and PAR single belt and GR split-belt (see Videos, Supplemental Digital Contents 1–3, which demonstrate a GR, SAR, and PAR locomotion pattern respectively,,, and, which allowed GRF recordings of both legs separately. When performing GR, subjects were not instructed to place their right foot on the right belt and their left foot on the left belt but ran split-belt by focusing on an extension of the midline between the two belts drawn on the floor in front of the treadmill. Foot contacts that hit both the right and the left belt of the treadmill simultaneously were eliminated from data analysis.

GRF was sampled at 1000 Hz. Axial TA was measured at 1000 Hz using an accelerometer (Noraxon DTS accelerometer; Noraxon USA Inc., Scottsdale, AZ) that was taped to the medial border of the tibia 0.08 m above the medial malleolus of the right leg (see Figure, Supplemental Digital Content 4, which illustrates the attachment of the accelerometer to the medial border of the tibia, Subjects’ skin was shaved and prestretched with tape to avoid excessive skin movement (12). Oxygen consumption (V˙O2) and carbon dioxide production (V˙CO2) were measured breath by breath (Cosmed K4b2; Cosmed, Rome, Italy). GRF and TA were measured for 10 s in the last min of each condition, whereas metabolic measurements were taken continuously.

Data analysis

Breath by breath metabolic measurements were reduced to 30-s averages for further analysis. To check if steady state was obtained, the metabolic rate in minute 4 was compared with minute 5 of the respective condition. One out of 81 conditions showed a difference of more than 10% and was removed from analysis, retaining only steady state metabolic measurements (see Figure, Supplemental Digital Content 5, which presents the individual metabolic measurements to check if steady state was obtained, Net V˙O2 was obtained by subtracting V˙O2 (mL·min−1) values during standing rest and were normalized to body mass (mL·min−1·kg−1). METs were calculated based on the V˙O2 of each condition versus the V˙O2 during standing rest.

Spatiotemporal parameters were based on the vGRF. Initial foot contact and toe-off were determined using a threshold of 20 N (13). Stance time was calculated as the time between initial foot contact and toe-off, swing time as the time between toe-off and initial foot contact, step time as the time between initial foot contacts of consecutive feet, stride time as the time between initial foot contacts of the same foot, step frequency as the inverse of step time, and step length as treadmill speed divided by step frequency. DF (%) was obtained by dividing stance time by stride time. DF higher than 50% indicate the absence of a flight phase.

GRF was low-pass filtered at 50 Hz using a fourth-order zero-lag Butterworth filter. vGRF was time normalized to average stance time, and maximal vGRF was calculated as the maximal vGRF during stance. Maximal VILR was calculated as the maximum of the first derivative of the vGRF, which was determined during the first 100 ms of stance. TA was filtered using a fourth-order zero-lag Butterworth low-pass filter with a cutoff frequency of 60 Hz. To compensate for gravity, 1g (9.81 m·s−2) was subtracted from the signal. Maximal TA was calculated in the first 100 ms of stance.

Whole body center-of-mass (bCOM) dynamics, i.e., gravitational potential energy, kinetic energy, and total energy, were determined from the GRF using the method of Saibene and Minetti (14). External power was obtained by dividing the summation of the positive increments in total energy within one stride-by-stride time. Percentage recovery (recovery), a measure indicating the magnitude of the exchange between forward kinetic and gravitational potential energy, was calculated according to Cavagna et al. (15).

To calculate population means, all measures were first averaged over consecutive strides of the same leg, then averaged between left and right and then averaged across subjects.

In-depth biomechanical analysis on subsample of 10 subjects

Whereas maximal vGRF (a general measure for both muscular and skeletal loading) and external power (a general measure for muscular loading) combined give a general idea of the musculoskeletal loading (16–18), more explicit measures of muscular loading refine the differences between GR and SAR. We opted to determine peak eccentric power, which is related to delayed onset muscle soreness (19), and muscle stress of the ankle and knee extensors (20) as more specific measures. The ankle and knee extensors were selected as these muscles generate the highest forces (21) during running.

To examine the differences in peak eccentric powers and muscle stresses between GR and SAR, a representative subsample from the main experiment (n = 10, age = 23.8 ± 2.4 yr, mass = 72.5 ± 6.1 kg, length = 1.83 ± 0.04 m, DFmain experiment = 51.19% ± 2.35%, DFin-depth experiment = 49.67% ± 2.07%) participated in a more profound experiment, completed on a separate occasion, in which detailed kinematics and kinetics were recorded. Subjects provided written informed consent, and the experimental protocol was approved by the ethical committee of the Ghent University Hospital.

The experimental protocol was similar to the protocol used in the main experiment with the exception that the PAR condition was not incorporated. Three-dimensional kinematics were recorded at 250 Hz using motion capture (Qualisys Oqus cameras, Göteborg, Sweden). The kinematic marker set consisted of 44 reflective markers (see Data File, Supplemental Digital Content 6, which describes the used marker set and kinematic model, Markers and GRF were filtered at 15 Hz using a fourth-order zero-lag Butterworth low-pass filter. A seven-segment, 42 DOF kinematic model was constructed using Visual3D (C-motion, Germantown, MD) to determine joint angles, joint angular velocities, joint moments, and joint powers. Maximal knee flexion angle during stance was subtracted from the knee angle at initial contact to obtain the amount of knee flexion in the beginning of stance (Δkneeflex). Timing of maximal knee flexion in stance (tkneeflex, max) was also determined. As an indicator of the vertical momentum of the lower leg and the foot at the end of the swing phase, the vertical velocity of the lateral malleolus of the left foot was determined just before initial contact of that foot. Muscle stress in the ankle and knee extensors was calculated according to Thorpe and colleagues (20). Their method combines published data of physiological cross-sectional areas and moment arms of the ankle and knee extensor muscles with experimental data (joint angles and joint moments) to calculate the minimal stresses in the ankle and knee extensors.


When data were available for the PAR conditions, repeated-measures ANOVA with post hoc tests was conducted with Bonferroni correction for multiple testing to compare GR, SAR, and PAR. One sample t-tests were conducted to compare GR with WK data from literature. Paired sample t-tests were performed to compare muscle stresses and peak eccentric powers between GR and SAR. Significance level was set at P < 0.05.


Significant differences were found between GR, SAR, and PAR for DF (F = 353.407, P < 0.001), stance time (F = 386.747, P < 0.001), swing time (F = 129.176, P < 0.001), step frequency (F = 23.417, P < 0.001), and step length (F = 1472.279, P < 0.001) (Table 1). DF was 23.3% higher for GR compared with SAR (P < 0.001). Compared with SAR, stance time was 21.9% longer (P < 0.001) and swing time was 15.6% shorter (P < 0.001) in GR. There were no significant differences in step frequency and step length between GR and SAR. Stance and swing time showed lower (P < 0.001) and higher (P < 0.001) values, respectively, for PAR compared with SAR, resulting in a significantly lower DF (P < 0.001) for PAR.

Spatiotemporal parameters (mean ± SD) for GR at 2.10 m·s−1, SAR at 2.10 m·s−1, and PAR at 3.2 m·s−1 (n = 27).

Parameters commonly used to classify gaits as WK or running were significantly different when GR was compared with WK data from literature (Table 2) and are situated closer to SAR reference values (Table 2; no statistics were performed between GR and SAR as these results are purely reference values to compare GR with SAR). The vGRF pattern showed a single hump with its maximum around midstance (Fig. 1, Table 2). The amount of knee flexion in the beginning of stance was 77% greater in GR than in WK (P < 0.001). Maximal knee flexion in GR was reached near midstance, which is typical for a running gait and much later compared to WK (P < 0.001). Recovery was 78% smaller for GR compared with WK (P < 0.001), indicating that only a small amount of energy is exchanged between kinetic and gravitational potential energy.

Parameters (mean ± SD) classifying GR as a running gait.
Ensemble averages of vGRF with SD at the maximum for GR at 2.10 m·s−1 (black line), SAR at 2.10 m·s−1 (blue line), and PAR at 3.2 m·s−1 (red line). The dashed line indicates initial contact of the contralateral foot in GR.

There were significant differences in maximal vGRF (F = 182.826, P < 0.001) and external power (F = 68.084, P < 0.001) between GR, SAR, and PAR (Table 3). They were 19.6% and 20.3% lower in GR compared with SAR, respectively (P < 0.001). Maximal TA (F = 106.294, P < 0.001) and maximal VILR (F = 134.941, P < 0.001) showed significant differences between GR, SAR, and PAR. Maximal TA and maximal VILR were 35.0% (P < 0.001) and 30.5% (P < 0.001) lower for GR than for SAR, respectively. The vertical velocity of the lateral malleolus of the left foot was 0.36 ± 0.08 m·s−1 for GR, which is 32% smaller than the 0.53 ± 0.11 m·s−1 for SAR (P < 0.001). Values of maximal vGRF (P < 0.001), external power (P < 0.001), maximal TA (P < 0.001), and maximal VILR (P < 0.001) were higher in PAR than in SAR. Significant differences in peak eccentric ankle (t = 4.240, P = 0.002) and knee power (t = 4.048, P = 0.003) between GR and SAR were identified. For both variables, GR showed lower values (respectively 34.0% and 25.6%) compared with SAR. Also, muscle stress at the ankle (t = −5.088, P < 0.001) and knee extensors (t = −5.142, P < 0.001) were lower for GR compared with SAR (19.6% and 17.2%, respectively).

Musculoskeletal loading parameters (mean ± SD) for GR at 2.10 m·s−1, SAR at 2.10 m·s−1, and PAR at 3.2 m·s−1.

Significant differences were found in metabolic measurements between GR, SAR, and PAR. Net V˙O2 (F = 573.846, P < 0.001), net V˙CO2 (F = 305.043, P < 0.001), and MET (F = 196.967, P < 0.001) were 4.8% (P = 0.003), 8.7% (P < 0.001), and 3.8% (P = 0.004) higher, respectively, in GR compared with SAR (Table 4). Values of net V˙O2 (P < 0.001), net V˙CO2 (P < 0.001), and MET (P < 0.001) were higher in PAR than in SAR.

Metabolic energy parameters (mean ± SD) for GR at 2.10 m·s−1, SAR at 2.10 m·s−1, and PAR at 3.2 m·s−1 (n = 27).


GR should be considered as a running gait (confirming hypothesis 1) as it is a symmetrical gait in which three out of four criteria, commonly used to discern running from WK, are present (Table 2, Fig. 2): (i) the pattern of the vGRF shows a single hump with its maximum around midstance, (ii) knee flexion during stance reaches its maximum value near midstance, and (iii) the kinetic and gravitational potential energy are in phase, resulting in a limited amount of exchange between them. These arguments agree with the spring-mass model of running (24) and justify its classification as a running gait although the temporal criterion (i.e., DF < 50%) is not met (Table 1). GR is running without a flight phase.

Ensemble averages of (A) vGRF, (B) knee angles, (C) kinetic energy, (D) gravitational potential energy, and (E) support fractions during the contact phase of the left foot for GR at 2.10 m·s−1 (black lines; own data) and walking at preferred walking speed (WK; green lines; data from Farley and Ferris [23] and Winter [22]). These variables are typically used to discern running from walking and identify GR as a running gait except for the support fraction as a period of double support is present.

All subjects participating in this experiment spontaneously ran with a flight phase at a slow speed of 2.10 m·s−1. Upon the simple instruction to “run without a flight phase,” they successfully performed GR at the same slow speed by increasing their DF from 41.5% to 51.2% on average. Subjects realized this by prolonging their absolute stance times by 21.9% while keeping stride frequency and step length constant (Table 1). To do so, the vertical movement of the bCOM (cf. vGRF impulses; Fig. 1) decreased, and the bCOM traveled over a longer horizontal distance during stance. Although GR remains a deviation from their natural running pattern, our subjects successfully performed GR upon a simple instruction. In natural slow runners, observations showed that a significant amount of them ran grounded spontaneously (7). Also, in an athletic population (11,25) and in animals (8–10), GR emerges spontaneously as long as the adopted running speed is slow enough. Gazendam and Hof (11) showed that approximately one third of an athletic population prefers GR over running with a flight phase at a speed of 2.00 m·s−1, a speed just below the speed imposed in this experiment.

Changing running style toward GR at a speed of 2.10 m·s−1 has two major consequences. The first consequence relates to the experienced musculoskeletal loading, which drastically decreases when DF is increased. General proxies of musculoskeletal loading were approximately 20% lower for GR than for SAR at the same slow speed (Table 3, Fig. 1) (confirming hypothesis 2). The impulse–momentum relationship explains this lower general loading. Over one stride, the average vGRF equals body weight. Consequently, prolonging stances without altering stride frequencies results in lower average vGRF over stance and thus lower maximal vGRF and external power (26). These reductions are in the same order of magnitude as the reductions that were evoked by purely a speed decrease from 3.20 m·s−1 (PAR) to 2.10 m·s−1 (SAR). As such, large differences in musculoskeletal loading are induced by changing a SAR pattern to a GR pattern.

A more specific aspect of musculoskeletal loading concerns impact intensity (maximal VILR and TA), which is approximately 31% to 35% lower in GR compared with SAR (confirming hypothesis 3). Impact intensity depends on the sudden deceleration of distal parts (foot and leg) on top of the slower deceleration of the proximal parts of the body (27). In GR, the proximal parts will indeed be less decelerated compared with SAR as the vGRF impulse is smaller (Fig. 1). In addition, the vertical momentum of the foot and the lower leg upon initial contact and thus the following deceleration during the initial impact phase is also smaller, resulting in lower impact intensity for GR compared with SAR (Table 3). Lower impacts might have implications for the occurrence of RRI. A recent model of Bertelsen and colleagues (28) proposes a conceptual framework explaining that RRI occurs when the musculoskeletal load capacity, i.e., the load that can be sustained before a RRI occurs, is exceeded by a cumulative load, i.e., the repetition of stride-specific mechanical loads that musculoskeletal structures are exposed to during a running session. Compared with SAR, the magnitude of each impact is smaller in GR while stride frequency and stride length are not altered, resulting in a lower cumulative load, thereby potentially reducing the risk of RRI.

Another specific aspect of musculoskeletal loading concerns muscular loading. Muscle stress and peak eccentric power of the ankle and knee extensors are more explicit measures of muscular loading than the general proxies of musculoskeletal loading. As hypothesized (confirming hypothesis 4), they are between 16.5% and 34.0% lower for GR compared with SAR (Table 3). This indicates that the dynamics of GR engage the muscles involved in body weight support and propulsion less intensely compared with SAR, which may postpone local fatigue in these muscles. However, it might be that other muscle groups experience an increase in loading. Simplified bipedal models indeed show that running with a higher DF reduces the peak external force but that it also involves less time to swing the leg forward. This could increase the load on the swing-related structures (29).

The question remains, however, if and to what extent it is necessary to decrease the musculoskeletal loading with running style adaptation(s) as this loading is already low in SAR especially compared with PAR. Mainly specific groups of runners with low loading capacities or for whom the load at average speed would already be very high, such as overweight or older runners, might benefit from adopting a GR pattern. These runners run slowly (30,31) perhaps to keep musculoskeletal load low. A shift toward a GR pattern will further reduce the load, thereby enabling them to cover longer distances or reducing the risk of injuries.

The second consequence of deliberately changing a running pattern concerns the associated energetic cost. Changing from SAR to GR increased the energy expenditure by approximately 5%. Initially, this was a surprising result as a change from SAR to GR involves a reduction in external power and musculoskeletal loading in the large leg extensors, which we expected to result in a decrease in energy expenditure. However, it actually may be a logical outcome as our young and athletic subjects performed instructed GR at a speed they would normally perform SAR. Although all subjects successfully performed a GR pattern, it remained a deviation from their natural running pattern at that speed. Studies investigating spatiotemporal interventions also found an increase in energy expenditure when runners did not use their spontaneously selected spatiotemporal coordination (29,32). It is an interesting question whether or not the energy expenditure during GR would still remain higher compared with SAR if our subjects would become trained in GR or when subjects that spontaneously use a GR pattern at the imposed speed were tested. Nevertheless, placing energy expenditure of GR and SAR in perspective to these of PAR shows that a manifest decrease in speed from PAR to SAR results in a 30.9% reduction. Slowing down enables less fit runners to run for a longer time as they will run at a lower percentage of their maximal energetic capacity. The increase when performing GR instead of SAR is much smaller (5%) compared with the reduction realized by decreasing the speed from PAR to SAR (31%). The benefit of a large decrease in musculoskeletal load likely supersedes the relatively small and maybe only temporary increase in energy expenditure. From the perspective of health-related physical activity, it might even be good that energy expenditure does not further decrease when performing GR. A similar or larger energy expenditure combined with a reduction in musculoskeletal load likely leads to an increase in training effectiveness for most runners. On the basis of the MET values (33,34), GR remains a moderate to vigorous form of physical activity contributing to fulfillment of the recommendations concerning physical activity and public health (35).

The key differences between GR and SAR identified in this study cause two interesting research questions: (i) Which determinants cause GR to emerge spontaneously in selected populations? (ii) Does GR produce fewer RRI than SAR?

Answering the first research question requires hypotheses regarding the determinants of the spontaneous occurrence of GR in selected populations. These determinants likely relate to the onset of fatigue or the occurrence of RRI, which are both determined by the balance between the experienced load and the individual’s load capacity. As load capacity differs between individuals, specific determinants of GR probably also differ. The present study suggests musculoskeletal loading as the main determinant to adopt a GR pattern. Individuals with reduced leg extensor force capacities (e.g., obese people) and/or with decreased load capacity (e.g., people with knee arthritis) are expected to spontaneously adopt a GR pattern when running at slow speeds. In-depth analysis of impact forces and loading of muscle–tendon units in modeling or intervention studies might give more insight whether people indeed spontaneously perform GR to keep musculoskeletal loading low. Another determinant that should be further investigated is energy expenditure. The present study suggests to exclude energy expenditure as a determinant as it was higher during GR compared with SAR. However, this result might be an artifact of the selected protocol and population. We expect that the energy expenditure of runners who spontaneously run grounded will be optimized, meaning that it will be the lowest for GR compared with all other running patterns at that speed. For these people, non-GR patterns are a deviation from their natural gait, which very likely will increase energy expenditure.

To answer the research question whether GR has a lower injury incidence compared with SAR, an intervention study needs to be executed. In this intervention study, runners will be trained to perform GR at speeds at which they normally would run with a flight phase. Once entirely familiarized to the GR pattern, a long-term follow-up is necessary to document the retention and the prevalence of injuries, which should be compared with a control group that did not receive the intervention program. If this study shows a lower injury incidence when running grounded instead of running with a flight phase, running programs aiming for health benefits should include GR training sessions.

A potential limitation of this study is the use of a split-belt treadmill when examining GR because it might induce an increase in step width. Arellano and Kram (36) indeed found that imposing step widths increased energetic cost and altered gait characteristics. In the present study, contrary to studies investigating the effect of step width on gait parameters (e.g., 36,37), subjects did not receive any instruction on step width. Nonetheless, it remains possible that subjects altered their step width to run split-belt, influencing their gait characteristics. A second limitation is the exclusive use of trained male subjects who were unfamiliar with the GR pattern. To gain more insight into the (habitual) GR locomotion pattern, future research should focus on habitual slow (grounded) runners and include several speed steps to compare a GR pattern with an SAR pattern in this habitual GR population.

In conclusion, GR contains all characteristics of a spring-mass model and should therefore be categorized as running although flight phases are absent. Upon the simple instruction “run without a flight phase,” athletic subjects successfully altered their normal running pattern into a GR pattern at a slow speed of 2.10 m·s−1. The deliberate change in running pattern decreased musculoskeletal loading drastically with a decrease in impact loading of 31% to 35% and in muscular loading of 17% to 34%. These large reductions might have implications for RRI and the onset of local muscle fatigue as a limiting factor of distance running. However, it also increased energy expenditure by approximately 5%, a small increase compared with the large decreases in musculoskeletal loading. We also believe that the increase in energy expenditure might disappear when subjects become familiarized with the new running pattern whereas the musculoskeletal load reductions will likely be retained as they are consequences of the altered running mechanics.

Mizuno Corporation provided financial and product support for this study. The authors acknowledge Rud Derie for his help with data collection and data processing and Davy Spiessens and Joeri Gerlo for their technical support.

All authors declare no conflict of interest. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of the present study do not constitute endorsement by the American College of Sports Medicine.


1. Tudor-Locke C, Johnson WD, Katzmarzyk PT. Frequently reported activities by intensity for U.S. adults: the American Time Use Survey. Am J Prev Med. 2010;39(4):e13–20.
2. Lee D, Brellenthin AG, Thompson PD, Sui X, Lee I-M, Lavie CJ. Running as a key lifestyle medicine for longevity. Prog Cardiovasc Dis. 2017;60(1):45–55.
3. Lopes AD, Hespanhol LC, Yeung SS, Costa LO. What are the main running-related musculoskeletal injuries? A systematic review. Sports Med. 2012;42(10):891–905.
4. Runkeeper Website [Internet]. Boston; [cited 2017 Aug 25]. Available from:
5. Schache AG, Blanch PD, Dorn TW, Brown N, Rosemond D, Pandy MG. Effect of running speed on lower limb joint kinetics. Med Sci Sports Exerc. 2011;43(7):1260–71.
6. Novacheck TF. The biomechanics of running. Gait Posture. 1998;7(1):77–95.
7. Shorten MR, Pisciotta E. Running biomechanics: what did we miss? Proceedings of the 35th Conference of the International Society of Biomechanics in Sport. 2017;34–7.
8. Andrada E, Rode C, Blickhan R. Grounded running in quails: simulations indicate benefits of observed fixed aperture angle between legs before touch-down. J Theor Biol. 2013;335:97–107.
9. Rubenson J, Heliams DB, Lloyd DG, Fournier PA. Gait selection in the ostrich: mechanical and metabolic characteristics of walking and running with and without an aerial phase. Proc Biol Sci. 2004;271(1543):1091–9.
10. Vereecke EE, Aerts P. The dynamics of hylobatid bipedalism: evidence for an energy-saving mechanism? J Exp Biol. 2008;211(23): 3661–70.
11. Gazendam MG, Hof AL. Averaged EMG profiles in jogging and running at different speeds. Gait Posture. 2007;25(4):604–14.
12. Digby CJ, Lake MJ, Lees A. High-speed non-invasive measurement of tibial rotation during the impact phase of running. Ergonomics. 2005;48(11–14):1623–37.
13. Watt JR, Franz JR, Jackson K, Dicharry J, Riley PO, Kerrigan DC. A three-dimensional kinematic and kinetic comparison of overground and treadmill walking in healthy elderly subjects. Clin Biomech (Bristol, Avon). 2010;25(5):444–9.
14. Saibene F, Minetti AE. Biomechanical and physiological aspects of legged locomotion in humans. Eur J Appl Physiol. 2003;88(4): 297–316.
15. Cavagna GA, Heglund NC, Taylor CR. Mechanical work in terrestrial locomotion: two basic mechanisms for minimizing energy expenditure. Am J Physiol. 1977;233(5):R243–61.
16. van der Worp H, Vrielink JW, Bredeweg SW. Do runners who suffer injuries have higher vertical ground reaction forces than those who remain injury-free? A systematic review and meta-analysis. Br J Sports Med. 2016;50(8):450–7.
17. Scott SH, Winter DA. Internal forces at chronic running injury sites. Med Sci Sports Exerc. 1990;22(3):357–69.
18. Cavagna GA, Thys H, Zamboni A. The sources of external work in level walking and running. J Physiol. 1976;262(3):639–57.
19. Eston RG, Mickleborough J, Baltzopoulos V. Eccentric activation and muscle damage: biomechanical and physiological considerations during downhill running. Br J Sports Med. 1995;29(2):89–94.
20. Thorpe SK, Li Y, Crompton RH, Alexander RM. Stresses in human leg muscles in running and jumping determined by force plate analysis and from published magnetic resonance images. J Exp Biol. 1998;201(Pt 1):63–70.
21. Hamner SR, Seth A, Delp SL. Muscle contributions to propulsion and support during running. J Biomech. 2010;43(14):2709–16.
22. Winter DA. The Biomechanics and Motor Control of Human Gait Waterloo (ON): University of Waterloo Press; 1987. pp. p21–7.
    23. Farley CT, Ferris DP. Biomechanics of walking and running, center of mass movements to muscle action. Exerc Sport Sci Rev. 1998;26:253–85.
    24. Blickhan R. The spring-mass model for running and hopping. J Biomech. 1989;22(11–12):1217–27.
    25. Hreljac A. Determinants of the gait transition speed during human locomotion: kinetic factors. Gait Posture. 1993;1(4):217–23.
    26. Weyand PG, Sternlight DB, Bellizzi MJ, Wright S. Faster top running speeds are achieved with greater ground forces not more rapid leg movements. J Appl Physiol. 2000;89(5):1991–9.
    27. Bobbert MF, Yeadon MR, Nigg BM. Mechanical analysis of the landing phase in heel-toe running. J Biomech. 1992;25(3):223–34.
    28. Bertelsen ML, Hulme A, Petersen J, Brund RK, Sørensen H, Finch CF. A framework for the etiology of running-related injuries. Scand J Med Sci Sports. 2017;27(11):1170–80.
    29. Minetti AE. Alexander RMN. A theory of metabolic costs for bipedal gaits. J Theor Biol. 1997;186(4):467–76.
    30. Zdziarski LA, Chen C, Horodyski M, Vincent KR, Vincent HK. Kinematic, cardiopulmonary, and metabolic responses of overweight runners while running at self-selected and standardized speeds. PM R 2016;8(2):152–60.
    31. Devita P, Fellin RE, Seay JF, Ip E, Stavro N, Messier SP. The relationships between age and running biomechanics. Med Sci Sports Exerc. 2016;48(1):98–106.
    32. Minetti AE, Capelli C, Zamparo P, Di Prampero PE, Saibene F. Effects of stride frequency on mechanical power and energy expenditure of walking. Med Sci Sports Exerc. 1995;27(8):1194–202.
    33. Haskell WL, Lee I-M, Pate RR, Powell KE, Blair SN, Franklin BA. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc. 2007;39(8):1423–34.
    34. Jetté M, Sidney K, Blümchen G. Metabolic equivalents (METS) in exercise testing, exercise prescription, and evaluation of functional capacity. Clin Cardiol. 1990;13(8):555–65.
    35. Samuelson G. Global strategy on diet, physical activity and health Scand. J Nutr. 2004;48(2):57–7.
    36. Arellano CJ, Kram R. The effects of step width and arm swing on energetic cost and lateral balance during running. J Biomech. 2011;44(7):1291–5.
    37. Donelan JM, Kram R, Kuo AD. Mechanical and metabolic determinants of the preferred step width in human walking. In: Proceedings of the Royal Society B: Biological Sciences, 2001 Oct 7: London (England). p 1985–92.


    Supplemental Digital Content

    Copyright © 2018 by the American College of Sports Medicine