Journal Logo


Elite Sprinting: Are Athletes Individually Step-Frequency or Step-Length Reliant?


Author Information
Medicine & Science in Sports & Exercise: June 2011 - Volume 43 - Issue 6 - p 1055-1062
doi: 10.1249/MSS.0b013e318201f6f8
  • Free


An athlete's running velocity is the product of step frequency (SF) and step length (SL)-a step being from one foot contact to the next contact of the contralateral foot. The term stride is also used in the literature, which is equal to two consecutive steps. Although the equation of velocity equals SF multiplied by SL is very straightforward and simple in theory, athletes face problems in practice because the relationship between SF and SL is generally an inverse relationship at maximum effort. Thus, an increase in one parameter could typically lead to a decrease in the other. This is due to the negative interaction apparent in the production of these variables (11). Consequently, this relationship has attracted attention in the biomechanics literature.

Luhtanen and Komi (16) were among the first to comprehensively analyze the relationship between SF and SL and presented the development of SF and SL in track athletes when running velocity was increased from jogging at 3.9 m·s−1 to sprinting at 9.3 m·s−1. However, this study is not directly relevant to elite sprint athletes, who always need to run at very high individual velocities in competition. In a study of 28 sprint-related sportsmen (background, e.g., in athletics, soccer, touch rugby), Hunter et al. (11) found that at the group-level SL was significantly related to running velocity, whereas SF was not. However, at the individual level, the subjects performed with a significantly higher SF in their fastest trial in comparison with their third fastest trial. SL did not reveal significant differences in the individual analysis (11). The authors offered a potential explanation for these differences between individual and group analysis by stating that SF may be the more important factor in the short term, whereas longer steps may require the development of strength and power during a longer period. Hunter et al. (11) also offered further detailed explanations of the technique issues that were behind the aforementioned negative interaction between SL and SF. The sprinting velocities, however, ranged from 7.44 to 8.80 m·s−1 (11) and were measured only 16 m into the sprint. Thus, whereas the article provides general information about step characteristics and can be helpful to developing athletes, it is not fully applicable to elite sprinters, whose running velocities are much higher.

To fully explore how elite athletes could fine-tune their performances, it would be necessary to understand how they perform in competition. Mann and Herman (17) analyzed the first-, second-, and eighth-place finishers in the 1984 Olympic men's 200-m final and highlighted the fact that the major difference between the three athletes (especially those in first and second) was SF. Interestingly, all three athletes increased velocity, SF, and SL between the nonfatigued (125-m mark) and fatigued (180-m mark) phases of the sprint.

Ae et al. (1) analyzed the final of the men's 100 m from the 1991 World Championships in Athletics. One of the key points highlighted by Ae et al. (1) in their conclusions was that the gold medalist generally exhibited a shorter SL and higher SF than the silver medalist, although this was not consistent throughout the whole race. A similar type of analysis over each 10 m was performed by Gajer et al. (8) from the semifinals and final of the men's 100 m at the 1996 French Championships. The six fastest (10.18 ± 0.05 s) and six slowest athletes (10.52 ± 0.08 s) were divided into separate groups. SL was consistently higher in the faster group and was significantly higher in 7 of 10 sections. On the other hand, SF was higher in the slower group in all but the last 10-m section, although it was significantly higher in only 1 of 10 sections. The authors (8) drew the conclusion from their results that SL was the more important factor at the highest level. Recent competition analysis from the World Championships in Helsinki 2005 (13) provided a similar trend to that of Gajer et al. (8). Eighteen male sprinters from the 100-m heats were divided into faster and slower groups (nine athletes in each group; high-performance group = 10.12-10.32 s, lower performance group = 10.40-10.90 s). In the full-stride phase (around 60 m), the longest SL was significantly longer (P < 0.003) by 0.12 ± 0.03 m for the faster group than the slower group, whereas there were no significant differences in SF.

Gajer et al. (8) also reanalyzed the data of Ae et al. (1) by splitting the eight finalists into two groups: the first- to fourth-place and fifth- to eighth-place finishers. The four fastest athletes had a higher average SL in 9 of 10 intervals, whereas the four slowest athletes had a higher average SF in 7 of the intervals. This was presented by Gajer et al. (8) as further evidence to support their own conclusions. Thus, at the group level the finding was opposite to the conclusion of Ae et al. (1) regarding the first- and second-place finishers. It seems, however, that the results are very dependent on the grouping. The grouping used by Gajer et al. (8) for the data from Ae et al. (1) meant that the groups were equal in number, each containing four athletes. When the finishing times for the eight athletes were examined, a different method of grouping could be justified. The first six finishers all completed the race in times in a close range of 9.86-9.96 s. The last two finishers were considerably slower, finishing with times of 10.12 and 10.14 s. New calculations reveal the opposite trend to that presented by Gajer et al. (8). With the modified groupings based on the absolute level of performance, the six fastest athletes recorded a higher SF in 9 of the 10 intervals, whereas the slowest two athletes had a higher SL in 7 of the intervals. This change occurred because the fifth- and sixth-place athletes typically displayed short SL and high SF values when compared with the other six athletes. This example shows that an average group-based analysis can actually mask important issues at the individual level. Because of this problem, Dixon and Kerwin (6) called for a multiple single-subject approach in studies where important individual differences that are not visible in general trends of a group analysis may be present. This might be even more important for individual elite athletes because any improvement in their performance may give them an advantage over the competitors. Thus, when elite sprinters try to improve their performance by seeking to cut hundredths of a second from their race time, it is very important to understand the individual performance and step characteristics issues rather than analyze them at the group level. Recently, in track-and-field biomechanics, there have been single-subject analyses published in sprinting (3) and sprint hurdling (22).

It is clear from the results presented on elite athletes in a competition situation that there is no consensus of opinion over which factor, SF or SL, is the more important at this level of competition. These are important findings, nonetheless, because they give a good insight into the performance of the very best athletes in a competitive situation, something that a laboratory- or training-based study is not capable of doing. There is, however, a lack of consideration for the possibility that individual athletes may adopt differing strategies from one another, about optimizing SF and SL. Further insight could be realized if the same elite athletes were analyzed over several runs. Such analysis is clearly missing from the current biomechanics literature. Thus, the aim of this study was to investigate the step characteristics among the very best 100-m sprinters in the world to understand whether the elite athletes are individually more reliant on SF or SL.


A total of 52 male elite-level 100-m races were recorded from publicly available television broadcasts. The competitions included several Olympic, World, and European Championships; International Association of Athletics Federations (IAAF) Grand Prix series competitions; European Cups; and some National Championships. The summary of competitions is in Table 1. Data were collected from semifinals and finals of the major championships and heats and finals of individual Grand Prix series competitions. Official race times were recorded from the IAAF Web site (12). A similar approach of analyzing publicly available data from sport competitions for research purposes has been carried out by Stewart and Hopkins (23), who analyzed the consistency of performance between swimming strokes, race distances, and two competitions across 221 swimmers. In the current study, athletes' individual races were analyzed if the athlete ran fully through the finish line. Thus, individual races in which an athlete clearly eased off before the finish line (e.g., in some heats or semifinals), sustained an injury, or in any way was deemed not to perform normally, were disregarded from the analysis. Consequently, the worst individual race time analyzed was 10.39 s. It is clear that not every analyzed athlete was involved in each competition. All athletes who performed in 10 or more races were taken for this analysis yielding a total of 11 athletes. The number of races per athlete is listed in Table 2. Of these 11 athletes, 9 ran under 10.00 s at least once in these competitions.

Summary of analyzed competitions.
Number of races with the mean official race time (SD) for each individual athlete in those races.

For each run of each athlete, the average SL and SF over the whole 100-m distance were analyzed as follows. The total number of steps taken in the race by each of the athletes of interest was counted by viewing the race in slow motion on a normal television and using a video player (AG-7550; Panasonic, Osaka, Japan), which yielded 50 video fields per second. Because the athlete did not necessarily complete a step exactly at 100 m, the displacement of the last step (SLS) was defined. This was the overall displacement from the start line to the toe of the ground foot in the step closest to the finish line (either side). The displacement estimation was based on using the track markings, the length of the foot (approximately 0.3 m), and the expected values for SL. Because the first step out from the starting blocks does not cover as much ground as all subsequent steps and it clearly takes the longest time, this step was disregarded from the calculations. To do that, a pilot test was set up. Four national-level athletes (who provided informed consent) were videotaped with a high-speed video camera (Motionscope 500C; Redlake Imaging Corp., Alameda, CA) at 250 Hz to estimate the length of the first step both as a distance and time (from the start signal to the instance of the first contact). On the basis of these four athletes' performances over 16 runs (4 each), a distance of 0.55 m and a time of 0.52 s were subtracted from the calculations. Average SL throughout the race was therefore calculated as follows:


is average step length, SLS is displacement of the last step, and nS is number of full steps. The total number of steps that were taken over the exact 100 m (nS100) was then calculated as follows (this provides the last step as a fraction):

From this, the average SF for the race was calculated as


is the average step frequency and tr is the official race time.

Statistical analysis.

All athletes were analyzed individually. SF, SL, and race time data were natural log-transformed before analysis to normalize distributions and stabilize variance. To determine any SF or SL reliance for an individual athlete, the 90% confidence interval (CI) for the difference between the SF-time versus SL-time relationships was derived using a criterion nonparametric bootstrapping technique (7) (Resampling Stats 4.0.7; Resampling Stats, Inc., Arlington, VA). Briefly, for each set of n races for each individual athlete, 10,000 resamples with replacement (of n cases) of the race time, SF and SL variables were taken (maintaining case correspondence). On each bootstrap resample, the SF-time and SL-time correlations (Pearson r) were derived, and the difference between these correlations was calculated and stored (SF minus SL). The 90% CI (10) for the difference between the SF and SL correlations was obtained using a simple percentile method, from the 5th and 95th percentiles of the distribution of 10,000 differences. The threshold for a practically important difference between SF and SL correlations (in either direction) was set at a value of 0.1-a "small" effect size for the correlation coefficient (4). An athlete was declared SF reliant if the lower limit of the 90% CI was at or beyond the threshold of −0.1, with the upper limit <+0.1 (precluding SL reliance). Conversely, an athlete was declared SL reliant if the frequency−length correlation difference was positive (favoring length), with the 90% CI precluding frequency reliance (≤−0.1). An effect was deemed "unclear" if the 90% CI simultaneously extended into regions, suggesting both SF and SL reliance; the athlete could be SF reliant or SL reliant, or there could be a trivial difference favoring neither step characteristic. In addition, to investigate whether the elite athletes were more reliant on SF or SL or whether height influenced this reliance, three further Pearson correlations were carried out: the point difference between the SF-time versus SL-time correlation values from above was further correlated with the individual mean race times as well as with the athletes' personal best times and heights, both of which were obtained from the athletes' biographical information on the IAAF Web pages (12). These data were not natural log-transformed because the point difference yielded negative values that cannot be log-transformed. The 90% CI values were calculated, and the threshold for a practically important difference was set at 0.1 as above.


Table 3 provides the correlation coefficients for each athlete between the independent variables and the race time. The correlation values between SF and race time varied between 0.16 and −0.79. Contrary to SF, all athletes yielded a negative correlation between SL and race time. The range of correlation values for SL varied from −0.16 to −0.89.

Correlation values for natural log-transformed SF and SL versus time.

Figure 1 provides the difference between correlations for SF-time and SL-time, together with its 90% CI. Athlete A10 yielded the highest positive difference with a value of 1.05 (with the CI ranging from 0.50 to 1.53). The largest negative difference occurred for athlete A11 as -0.60 with the CI ranging from -1.20 to 0.03. The area of ±0.1 to indicate the smallest practically worthwhile difference between correlations is also shown in Figure 1.

Ther difference (diamonds) with 90% CI (bars) for each athlete A1 to A11. The area of ±0.1 from zero in the middle demonstrates the trivial (nonreliant) effect.

Owing to the large variation shown in r-difference values, three athletes' data are specifically shown in Figure 2 to illustrate the athletes' times as a function of SL and SF. On the basis of data in Figure 1, athlete A10 (Fig. 2, A and B) had the largest SL reliance, athlete A4 (Fig. 2, C and D) did not yield any reliance either on SF or on SL, and athlete A11 (Fig. 2, E and F) was the only athlete who was clearly SF reliant. The minimum, maximum, and mean of average SL and SF values for each athlete are presented in Table 4, showing that the lowest range for the average SL was 0.06 m, whereas the largest range was 0.14 m. The respective values for the average SF range were 0.07 and 0.30 Hz.

Three athletes' race times as a function of SF and SL: athlete A10 (A and B) showed SL reliance, athlete A4 (C and D) did not yield either reliance, and athlete A11 (E and F) was SF-reliant. Please note thaty axes have been inverted because quicker times demonstrate improved performance. Points on the figures with trend lines are from the original data; r values are from the log-transformed data (Table 3). Because of inverted yaxes, the signs of r values do not match the visual impression.
Minimum (Min), maximum (Max), and mean of average SL and SF values for each individual athlete.

The SF-SL reliance (as correlation difference) did not show a meaningful relationship with the athletes' mean race time (CI for r, −0.27 to 0.71), personal best times (−0.63 to 0.40), or height (−0.13 to 0.77).


This study was designed to increase our understanding of the SF and SL characteristics of elite athletes in major competitions. The main results showed that these characteristics vary considerably between the athletes. Previous studies have generally identified only one of these variables to be the main reason for faster running velocities, and the results have given a somewhat confusing picture. Kuitunen et al. (15) showed that SF was the dominant factor when running velocity increased from 70% to 100%. Higher SF also seemed to be the major difference between three Olympic 200-m finalists (17). On the other hand, Gajer et al. (8) found that better 100-m sprinters in their study had longer SL than slower athletes, and Hunter et al. (11) showed that the SL was significantly related to running velocity at the group level (whereas SF was not). The results of Hunter et al. (11), however, showed that, within individuals, SF was higher in the fastest trials. None of these studies, however, have looked at elite athletes across different races to determine how an individual athlete performs. From an elite athlete's point of view, the group-level data do not provide appropriate information to improve individual performance. For example, by executing an average performance of 100-m Olympic finalists, the athlete would not win the race. In fact, often, if an athlete were to achieve the average performance of all the finalists, that would not be sufficient to even place that athlete on the podium. Thus, it is important to look at each elite athlete individually. To the best of our knowledge, this is the first study that has looked at step characteristics of elite athletes individually and longitudinally across multiple competitive races. In addition, a novel aspect of the current study is the use of a criterion bootstrapping method, together with a criterion for practical significance, to elucidate the within-athlete differences between SF and SL.

Average SF multiplied by average SL provides the average running velocity, which, in turn, has an inverse relationship with the race time. This means that both SF and SL are inversely linked with the race time and strongly related to each other. This collinearity between SF and SL makes it impossible to properly separate the independent influence of these predictors on race time, if both are entered together as predictors in a multiple regression model. Therefore, a novel approach was sought to understand any reliance on particular step characteristics by athletes. Consequently, it was decided that the most appropriate approach was to adopt a bootstrapping technique to calculate the 90% CI for the difference between SF-time versus SL-time correlations to inform how practically meaningful this effect was. The interpretation of the results in Figure 1 follows the recommendations by Batterham and Hopkins (2). The effect is considered reliant if the 90% CI is fully on either the SF or SL side or if one end reaches only to the area of a trivial effect in the middle. If the CI extends to include both frequency and length reliance, then the effect is considered unclear.

Overall, the results in Figure 1 revealed that there is a large variation of performance patterns among the elite athletes. There were clearly athletes at the highest elite level of 100-m sprinting who were SL reliant (athletes A10, A9, and A5), whereas only athlete A11 was clearly SF reliant. All other athletes did not have clear reliance on either side, although there were trends implying that, for example, athlete A7 was most likely to be SL reliant and athlete A2 was most likely to be SF reliant. When looking at the results in further detail, athlete A10 yielded a 90% CI (0.50-1.53), which did not even cross over the ±0.1 trivial effect region (Fig. 1). Thus, athlete A10 performed best when he was able to produce long steps (within his own range; Fig. 2B). Such reliance of SL meant that if the athlete was not able to produce long steps, he could not compensate the performance enough with high SF to produce fast 100-m times. On the contrary, athlete A11 performed his best times when he was capable of producing high step frequencies (within his own range; Fig. 2E). The 90% CI (−1.20 to 0.03) crosses over into the trivial effect region from the SF reliance but does not reach to a SL-reliant effect (Fig. 1). This meant that, if the athlete could not produce high step frequencies (e.g., if the nervous system was not ready to fire quickly enough to have a fast turnover of the steps), the SL had not compensated the running velocity enough. The athletes whose 90% CI reached over all three different zones in Figure 1 were such that they produced the best times sometimes with slightly higher SL (and lower SF) and sometimes with slightly higher SF (and lower SL; see an example of athlete A4 in Fig. 2, C and D).

When looking at the individual SL and SF values within athletes and across the races, the three examples in Figure 2 provided a very similar range of values. SL range was 0.08 m for A11, 0.11 m for A10, and 0.13 m for A4. The respective SF ranges were 0.22, 0.11, and 0.30 Hz. Athlete A4 had the largest range in both SL and SF from all athletes (Table 4). As the range of values on SF and SL were quite similar for all athletes, it reinforced that SF or SL reliance occurred within the normal range of that variable in individual athletes, and it was not due to some clear outliers in occasional runs.

The average within-athlete SF in this study ranged from 4.43 to 5.19 Hz, whereas the average SL ranged from 2.01 to 2.34 m (Table 4). It is clear that the average SL over the full 100 m in this study were less than those found in the maximum velocity phase in the literature because data in the current study also contain steps at the start of the run, which are shorter than later in the run. Ae et al. (1) reported SL from 2.29 up to 2.71 m for the World Championships' finalists in the maximum velocity phase. SL values reported by Gajer et al. (8) fell within the range provided by Ae et al. (1). SF values in the current study match more closely to those at maximum velocity because SF does not alter largely during the race. This is due to the fact that when early contact phases are generally longer, the flight phases are shorter. This ratio gradually changes; however, the total step time (and thus frequency) does not drastically change, as visible in the data of the first four steps out of the blocks in a study by Salo et al. (21). Step frequencies in the maximum velocity phase provided by Ae et al. (1) and Gajer et al. (8) generally matched the range seen in the current study.

At the group level, SF-SL reliance did not yield meaningful relationships with athletes' mean race times (CI for r, −0.27 to 0.71) or the personal best times (−0.63 to 0.40). This means that, for example, SL-reliant athletes were not any faster than SF-reliant athletes. Thus, it is possible to reach the absolute top level of sprinting in the world (run under 10.00 s) with widely varying patterns of SF and SL reliance. The results also showed an unclear (trivial) effect (i.e., there was no relationship) between the height of the athletes and SF-SL reliance (CI for r, −0.13 to 0.77). This means that taller athletes within this group were not SL reliant (against the general perception) or that shorter athletes were not SF reliant and vice versa. Overall, these three results support the idea that either SF or SL reliance is a highly individual occurrence.

The wind has been shown to influence the finishing time in sprinting. For example, the theoretical calculation by Ward-Smith (24) showed that a 2-m·s−1 following wind improves a 100-m result at the elite level (10.00-s runner) by 0.10 s, whereas the same head wind would slow the runner down by 0.13 s. However, the situation in the current study was different from that of an individual race because data were collected during a long period and across numerous races. It is clear that athletes train and target some major competitions, and thus, they potentially run faster in these races regardless of the wind speed in comparison with races perhaps earlier in a season. There were no clear trends that the faster times were set with better wind conditions. In addition, regardless of the wind, the race time was performed with that specific SF and SL combination found in the analysis, and it is this specific SF-SL pattern (reliance) that is the interest in the current study.

Because the SF or SL reliance varied considerably between the athletes, it is proposed here that this should be taken into account in their training, especially in the preparations for the most important competitions. The effect of different types of training on athletes' performance is difficult to prove owing to two factors: first, there is an inherent problem in getting elite athletes to participate in training studies (14), and second, it is practically impossible to isolate the training influence of one specific type of exercise or mode of exercise. However, some indirect conclusions can be drawn from the literature and theory of specificity in training.

On the basis of animal research, Heglund and Taylor (9) concluded that the increased stride length in various animals primarily required higher average muscle force production pointing toward the association between muscle strength and stride (step) length. Studying humans' sprinting performance, Weyand et al. (25) concluded that the faster running speeds were achieved by greater vertical ground reaction forces rather than more rapid leg movements. The higher average force production during the contact (i.e., strength) resulted in considerably higher stride lengths. The regression analysis showed that a 1.8-fold increase in top running speed was achieved with 1.69 times longer strides (and with an average vertical force production that was 0.5 times body weight larger). It is acknowledged that, in the same study, higher stride frequency was also associated with increased force production. This was because a higher vertical force production allowed athletes to produce the required impulse in a shorter contact time. However, the regression analysis showed only a 1.16-times increase in stride frequency across the same range of top speeds as above. On the other hand, Mero and Komi (18) found that only well-trained athletes (as opposed to less-trained athletes) were able to increase SF when towed to supramaximal velocities. Ross et al. (20) concluded that this ability to increase SF may have been caused by neural adaptations of training. Furthermore, Heglund and Taylor (9) stated that higher stride frequencies in animals require faster production of cross-bridges owing to faster force generation demands pointing toward the association between SF and neural conditioning.

Hunter et al. (11) hypothesized, based on their results and the literature, that developing longer SL requires long-term development of strength and power, especially to increase horizontal ground reaction impulse. Cronin et al. (5) studied how two types of resistive training (sled towing and weighted vest) acutely influenced step characteristics over the first 30 m in comparison with unresisted sprinting. Because relative strength due to additional weights was reduced, the decrease in performance was mainly due to lower SL with only small decreases in SF. Moir et al. (19) had a slightly different approach when the authors studied the influence of 8 wk of resistance training on step characteristics. Although these were analyzed only over the first three steps after the start, and thus may not be fully applicable to the current article, the results gave indications of how such training affects these step variables. Increased maximum and explosive strength was associated with increased SL and reduced SF over these first three steps (19).

Thus, overall, it is reasonable to conclude that SL is related more to increased force production, and SF is associated with faster force production during the contact and quick leg turnover requiring neural adaptations. Higher SF requires cross bridges within the muscles to be built at high rates, and thus, these need a high rate of neural activation. Consequently, it is proposed that the SF-reliant athletes are required to concentrate on neural activation in their final preparations for the major races and have a nervous system ready such that they can produce the quick turnover of the legs. On the other hand, the SL-reliant athletes need to keep their strength levels up throughout the season and have the required flexibility in the hip area to produce long steps. Naturally, athletes cannot totally forget the nonreliant variable because any disproportionate reductions in one variable cannot be generally compensated for by the other variable. The athletes who cross over the ±0.1 trivial effect region should perhaps focus equally on SL and SF in their training. It is also good to remember that the current study was based on the average step variables over the whole 100 m, whereas SL, especially, varies throughout the race. It is estimated that the accuracy of our measurements is about 0.01 m for the SL and 0.06 Hz for the SF. The final caution is that this article was able to provide only results based on how people have performed not how an ideal performance could be created. Thus, to further understand how SF and SL influence each other and interact to produce the velocity of the individual athlete, these same variables should be analyzed individually at the maximum velocity phase and longitudinally in training. Some questions about the SF, SL, and velocity relationships could probably be best answered by adopting a modeling approach.


This study analyzed SL and SF of world elite male 100-m sprinters over multiple competitions. Because group-level analysis could mask personal differences, this study concentrated on analyzing each athlete separately. Individually, some athletes' performances were more reliant on SL, one athlete was clearly SF reliant and some athletes used combinations that showed a reliance on neither. It is proposed that athletes should take this reliance into account in their training, with SF-reliant athletes needing to keep their neural system ready for fast leg turnover and SL-reliant athletes requiring more concentration on maintaining strength levels.

This study has been partly funded by UK Athletics Ltd. and the Leverhulme Trust, United Kingdom.

The authors thank the BBC Sports Library and its staff for access to broadcast archives, and the Research Institute for Olympic Sports, Finland, and their staff are thanked for their assistance in the pilot test for the "first-step" evaluations.

The authors declare that they have no conflict of interest and that the results of the present study do not constitute endorsement by the American College of Sports Medicine.


1. Ae M, Ito A, Suzuki M. The men's 100 metres. N Stud Athlet. 1992;71(1):47-52.
2. Batterham AM, Hopkins WG. Making meaningful inferences about magnitudes. Int J Sports Physiol Perform. 2006;1(1):50-7.
3. Bezodis IN, Kerwin DG, Salo AIT. Lower-limb mechanics during the support phase of maximum-velocity sprint running. Med Sci Sports Exerc. 2008;40(4):707-15.
4. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Hillsdale (NJ): Lawrence Erlbaum Associates; 1988. p. 567.
5. Cronin J, Hansen K, Kawamori N, McNair P. Effects of weighted vests and sled towing on sprint kinematics. Sports Biomech. 2008;7(2):160-72.
6. Dixon SJ, Kerwin DG. Variations in Achilles tendon loading with heel lift intervention in heel-toe runners. J Appl Biomech. 2002;18(4):321-31.
7. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. New York (NY): Chapman & Hall; 1993. p. 436.
8. Gajer B, Thépaut-Mathieu C, Lehénaff D. Evolution of stride and amplitude during course of the 100 m event in athletics. N Stud Athlet. 1999;14(1):43-50.
9. Heglund NC, Taylor CR. Speed, stride frequency and energy-cost per stride-how do they change with body size and gait. J Exp Biol. 1988;138:301-18.
10. Hopkins WG, Marshall SW, Batterham AM, Hanin J. Progressive statistics for studies in sports medicine and exercise science. Med Sci Sports Exerc. 2009;41(1):3-12.
11. Hunter JP, Marshall RN, McNair PJ. Interaction of step length and step rate during sprint running. Med Sci Sports Exerc. 2004;36(2):261-71.
12. International Association of Athletics Federations Web site [Internet]. Monaco: IAAF; [cited 2009 Sep 16]. Available from:
13. Ito A, Ishikawa M, Isolehto J, Komi PV. Changes in the step width, step length, and step frequency of the world's top sprinters during the 100 metres. N Stud Athlet. 2006;21(3):35-9.
14. Kearney JT. Sport performance enhancement: design and analysis of research. Med Sci Sports Exerc. 1999;31(5):755-6.
15. Kuitunen S, Komi PV, Kyröläinen H. Knee and ankle joint stiffness in sprint running. Med Sci Sports Exerc. 2002;34(1):166-73.
16. Luhtanen P, Komi PV. Mechanical factors influencing running speed. In: Asmussen E, Jørgensen K, editors. Biomechanics VI-B-International Series on Biomechanics. Baltimore (MD): University Park Press; 1978. p. 23-9.
17. Mann R, Herman J. Kinematic analysis of Olympic sprint performance: men's 200 meters. Int J Sport Biomech. 1985;1:151-62.
18. Mero A, Komi PV. Effects of supramaximal velocity on biomechanical variables in sprinting. Int J Sport Biomech. 1985;1(3):240-52.
19. Moir G, Sanders R, Button C, Glaister M. The effect of periodized resistance training on accelerative sprint performance. Sports Biomech. 2007;6(3):285-300.
20. Ross A, Leveritt M, Riek S. Neural influences on sprint running-training adaptations and acute responses. Sports Med. 2001;31(6):409-25.
21. Salo AIT, Keränen T, Viitasalo JT. Force production in the first four steps of sprint running. In: Wang Q, editor. Proceedings of XXIII International Symposium on Biomechanics in Sports. Beijing (China): The China Institute of Sport Science; 2005. p. 313-7.
22. Salo AIT, Scarborough S. Changes in technique within a sprint hurdle run. Sports Biomech. 2006;5(2):155-66.
23. Stewart AM, Hopkins WG. Consistency of swimming performance within and between competitions. Med Sci Sports Exerc. 2000;32(5):997-1001.
24. Ward-Smith AJ. New insights into the effect of wind assistance on sprinting performance. J Sports Sci. 1999;17(4):325-34.
25. 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.


©2011The American College of Sports Medicine