In vivo measurement of the series elasticity release curve of human triceps surae muscle A.L. Hof* Department of Medical Physiology, University of Groningen, Bloemsingel 10, NL-9712 KZ Groningen, The Netherlands Received in final form 17 April 1998

Abstract The force-extension characteristic of the series-elastic component of the human triceps surae muscle has been measured in vivo by means of a hydraulic controlled-release ergometer in 12 subjects. The SEC characteristic can be described by a linear relation between muscle moment and extension, with a stiffness K , preceded by a quadratic ‘toe’ region at low moments. Stiffness K increases with the 1 1 level of activation of the muscle. At an ankle moment of 100 N m, values range between 250 and 400 N m rad~1. The elastic stretch corresponds then to about 30°, which is considerable compared with the total ankle movement range. At moments of 150 and 180 N m, which correspond to the peak moments in walking and running, respectively, the elastic energy stored in the SEC was 23—37 and 31—57 J. These energy values are amply sufficient to enable the conservation of negative (eccentric) work into a subsequent phase of positive (concentric) work in both walking and running. ( 1998 Elsevier Science Ltd. All rights reserved. Keywords: Muscle stiffness; Tendon compliance; Elastic energy; Series elasticity; Controlled release

1. Introduction Human running has been compared to the bouncing of a rubber ball (Cavagna et al., 1964). The theory is that potential and kinetic energy of the trunk are stored briefly as elastic energy in the leg extensor muscles during the first half of stance, to be released again in the second half (Alexander and Bennett-Clark, 1977). A major problem in this theory is that as yet only indirect estimates for the elastic properties of human muscle are available (Hof, 1990; van Ingen Schenau et al., 1997). In this paper we present measurements of the elasticity curves of the human triceps surae muscle group, the main extensor (plantarflexor) of the foot, obtained by quick-release experiments on a specially developed ergometer (Hof, 1997a). Two elastic components can be identified in a single muscle—tendon complex: the series-elastic component (SEC) in series with the contractile component (CC), and the parallel elastic component (PEC1), parallel to the CC (Fig. 1). When measurements are done on the joint of an

*Tel.: 0031 50 363 2645; fax: 0031 50 363 2751; e-mail: a.l.hof @med.rug.nl 0021-9290/98/$19.00 ( 1998 Elsevier Science Ltd. All rights reserved. PII: S0021-9290(98)00062-1

intact animal or human, an antagonist muscle group is always present as well. Even when it is kept passive, the force due to its PEC is still present (PEC2). Plantarflexion is mainly exerted by the triceps surae group, which consists of the muscles soleus, gastrocnemius medial head, and gastrocnemius lateral head. The other plantarflexors, like m. tibialis posterior and the peronei, are relatively small and have much shorter moment arms. All three triceps surae muscles have only short contractile fibres of 4—6 cm. Then the SEC can largely be identified with the Achilles tendon and with the extensive aponeurosis of this muscle group, which span the rest of the total muscle length of some 40 cm (Alexander and Bennett-Clark, 1977). PEC1 is probably due to elastic properties of the connective tissue between the muscle fibres, endo- and epimysium. Antagonist muscles are the foot dorsiflexors, m. tibialis anterior, m. extensor hallucis longus and the other toe extensors. In this paper triceps surae force and length will be expressed as the moment about and the angle of the ankle, with 90°"1.57 rad for the foot in the neutral position and plantarflexion positive. From Fig. 1 it follows then for the ‘length’ of the SEC: / "/ !/ % #

(1)

794

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800

Fig. 1. Model of the elastic and contractile components in a muscle—tendon complex, with an inactive antagonist. SEC: serieselastic component, CC: contractile component, PEC1: parallel elastic component of agonist, PEC2: parallel elastic component of antagonist. Moment around and angle of the joint are given to represent muscle force and length: M" total moment, M "moment acting on SEC, % M and M "moments due to PEC1 and PEC2, respectively. 11 12 /"ankle angle (neutral position is 1.57 rad "90°) / "equivalent % length of SEC in angular units, / "equivalent length of CC. #

with / the angle corresponding to the length of the CC, # and / the measured ankle angle. When it is assumed that the antagonists are passive, for the moment exerted on the SEC holds: M "M!M . (2) % 12 In a controlled release a muscle is shortened at a high but constant speed. As long as the speed is constant, inertial forces are zero. The movement should be completed before the first reflexes arrive, for a release of triceps surae within 70 ms (Allum et al., 1982). The speed should be above the maximum shortening speed of the muscle, estimated at 10 rad s~1 (Hof and van den Berg, 1981). By recording the decline in force as a function of shortening, the SEC characteristic can be obtained, in principle from a single release (Bressler and Clinch, 1974; Hill, 1950).

2. Methods 2.1. Ergometer The hydraulic controlled-release ergometer has been described in an earlier paper (Hof, 1997a). The main component is a foot pedal that can be moved at a high but constant angular velocity, for the present experiments set at about 14 rad s~1. The initial angular acceleration was limited to 2000 rad s~2. After 20 ms and a rotation of 0.25 rad the final velocity was reached and for the remainder of the release (ca. 0.5 rad) the acceleration was negligible. At the end of the movement the deceleration was unlimited, but the inertial transients in this phase do not influence the measurement. An advantage of a hydraulic system is that it is very stiff: 6000 N m rad~1. The

foot of the subject was enclosed in a block of hard polyurethane foam, formed after a plaster of Paris cast, as described by Weiss et al. (1984). The block was mounted on the pedal such that the anatomical axis of the ankle coincided as well as possible with the axis of the pedal. The ankle moment was measured by strain gauges on the pedal, ankle angle by a precision potentiometer, and angular acceleration by means of a piezo-electric accelerometer. Moment and acceleration signals were both low-pass filtered (Krohn-Hite 3343, 500 Hz, 48 dB/ octave). Surface electromyograms were recorded from the soleus and (antagonist) tibialis anterior muscles. The signals were A/D converted at 2000 samples s~1, and processed on a personal computer. 2.2. Correction for inertia The angular acceleration was recorded to enable a correction for the inertial transients, based on a model consisting of two masses, foot and pedal plus mounting, with compliant connections. This led to a filtering method for moment and acceleration that gave the true muscle moment within a r.m.s. error of 2.5 N m and a peak error of about 8 N m (Hof, 1997b). The ankle angle was also corrected, for the finite stiffness of the tissues and material between foot and pedal, of the order of 2000 N m rad~1. 2.3. Correction for CC shortening In the release muscle, force decreases below the isometric value, and as a consequence the CC starts to shorten. The speed of shortening can be calculated from the Hill’s force velocity relation (Bressler and Clinch, 1974; Hill, 1950): M !M #, /Q "b 0 (3) # M #nM # 0 where /Q is the CC shortening speed, M is the moment # # exerted by the CC, obtained from the measured moment after correction for the inertial effects and after subtracting the PEC1 and PEC2 moments. M is the isometric 0 moment, as recorded before the release. For parameters b and n values of 1.2 rad s~1 and 0.12, respectively, have been adopted from Hof and van den Berg (1981). The shortening speed was integrated to obtain CC shortening (in angular units). With the adopted angular convention, where an increasing ankle angle corresponds to plantarflexion and muscle shortening, this CC shortening angle should be subtracted from the measured ankle angle. 2.4. PEC elasticity The moments of the parallel elastic components PEC1 and PEC2, M and M can be evaluated with passive 11 12

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800

muscle. In that case M "0, and M"M #M (see # 11 12 Fig. 1). In agreement with Esteki and Mansour (1996) and Hof and van den Berg (1981) these were modelled with exponential functions. The moment due to PEC1, from triceps surae, increases with dorsiflexion: M "M e~ a1(# (4) 11 110 The moment of PEC2, from the antagonist muscles, is negative and becomes more negative with plantarflexion: (5) M "!M e a2( 12 120 Next to this, a constant offset moment M was pres0&&4%5 ent, probably due to the vertical force exerted by the subject and the imperfectly aligned axes of the ankle and pedal. 2.5. Subjects and procedure There were 12 subjects, 10 male and 2 female, without a recent history of ankle or foot problems. Subject data have been given in Table 1. During the experiments the subjects stood erect, knees straight, with their left foot on the floor and the right foot in the ergometer. They could hold to a set of handle bars. In the experiments the guidelines for human experimentation of the Groningen Faculty of Medical Sciences have been followed. Firstly, some recordings were made with the muscles passive, at slow speed (0.1 rad s~1 ). This gives the parameters of PEC1, M and a , and PEC2, M and a . 110 1 p20 2 Secondly, recordings are made at high speed (14 rad s~1), some with the ergometer empty and some with the foot mounted, but passive muscles. A parameter estimation process then gives the moments of inertia and the transfer functions that enable the inertia correction (Hof, 1997b). Finally, a series of fast (14 rad s~1 ) recordings was made at different levels of triceps surae contraction, in

Table 1 Subject data. M (max) is the maximum isometric moment, recorded 0 just before the release Subject

Age (yr)

Stature (m)

Body mass (kg)

M (max) 0 (Nm)

1 2 3 4 5 6 7 8 9 10 11 (f) 12 (f)

34 21 27 27 27 35 60 26 49 35 30 31

1.78 1.73 1.85 1.80 1.87 1.83 1.82 1.95 1.92 1.89 1.72 1.69

72 59 76 88 71 78 80 89 88 79 72 64

121 102 123 173 116 95 106 109 119 92 112 97

Mean

34.1

1.82

76.3

113.8

795

steps of about 20 N m up to the maximum voluntary contraction. The release started with the foot dorsiflexed, at about 75° (1.3 rad). Subjects could monitor their own exerted moment on a bar display. Tibialis anterior EMG was made audible to help the subjects prevent antagonist cocontraction. In processing the recorded data (Fig. 2a) first the inertia correction was applied to the measured moment, using the acceleration signal (Fig. 2b). Then, the correction for CC shortening, Eq. (3), was applied to the angle recording (Fig. 2c). Finally, the PEC2 moment (Fig. 2d) was subtracted from the total moment, according to Eq. (2). This gives the SEC moment M as a function of % the SEC release (Fig. 3). The SEC ‘length’ / can be found % by subtracting this corrected / from the unknown, but constant, CC length at the start of the release, Eq. (1). In practice, this was done by shifting the measured curve horizontally such that the initial (highest) value of M was on the fitted curve (Fig. 4). % 3. Results A series of elastic curves, moment as a function of ankle angle, obtained from releases at various levels of active contraction are shown in Fig. 3. A registration with passive muscles has been included. These curves, and those with the lowest preloads approach the zero moment level completely. This shows that the tendon can become completely slack after a sufficient release. A good fit for all SEC curves was obtained with a linear function, with a constant elasticity K , preceded by 1 a quadratic curve in the ‘toe’ region. Using M as the b demarcation between the quadratic and the linear region, M can be fitted as: e (6a) M "(K2/4M ) /2 for 0(M (M , % " 1 " e % M "K / !M for M *M . (6b) % 1 % " % " The values for K were obtained by linear regression of 1 M vs / for M 'M . Then M was determined by % % % " " iteration, minimizing the mean square error of the complete curve. An example of the fit of this function on an experimental curve is shown in Fig. 4. It can be seen from Fig. 3 that K increases with the 1 initial moment M : the SEC becomes stiffer at higher 0 moments. In Fig. 5, K has been plotted as a function of 1 M . On the basis of a simple model (see Discussion) the 0 points have been fitted as 1 1 1 " # . K K c M #K 1 0 , 0 1

(7)

The numerical values for K , K and c have been given 0 1 , in Table 2. With Eq. (7) the measured values of K could 1 be predicted within 20 N m radv1 .

796

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800

Fig. 2. Moment as a function of ankle angle. (a) Moment and angle recorded by the ergometer. (b) Moment corrected for inertia effects and angle for finite stiffness of the ergometer and the foot. (c) Angle corrected for the shortening of the contractile component during the release, according to Eq. (3). (d) The passive moment of the antagonists M , Eq. (5). SEC moment M is the difference between the curves 12 % c and d, Eq. (2).

Fig. 3. SEC moment M as a function of ankle angle / (corrected as % in Fig. 2) for five recordings from releases starting at isometric moments of 10, 19, 47, 75, and 114 N m (same as in Fig. 2), from bottom to top. Subject 5.

Although the curve fit was not very sensitive to this parameter, it was found that M could best be taken " proportional to M : 0 M "c M . (8) " " 0 Some subjects had fairly straight curves, with low c , like " the subject presented in Figs. 2—5, while others had more convex curves, with higher c (Table 2). " The parameters determined in this way, the average error of the curve fit to all recordings of a subject (Table 2) was close to the estimated error of measurement of 2.5 N m r.m.s. (Hof, 1997b).

4. Discussion The amount of elastic stretch in a loaded human triceps surae is remarkable: about 0.56 rad at a moment

of 114 N m (Figs. 3 and 4). Assuming a moment arm of 50 mm and a total length of tendon and aponeurosis of 350 mm, this amounts to a muscle-tendon length change of 28 mm or 8%. In running considerably higher moments than the maximum voluntary isometric value are attained, in sprinting up to 210 N m (3.0 times body weight, Jacobs and van Ingen Schenau, 1992). Because of the increasing stiffness at higher moments, the stretch then does not reach improbable values: extrapolating for the same subject 5 with Eqs. (6) and (7) this would result in a stretch of 0.73 rad, 36 mm or 10%, respectively. This comes close to the stretch limit before breaking, observed in tendon material in vitro, about 15% (Schechtman and Bader, 1997). The human ankle plantarflexors seem to belong to the ‘type iii’ muscles (Alexander and Ker, 1990), with short muscle fibres and relatively slender and highly stressed tendons, optimized for elastic energy storage. 4.1. Increase of stiffness with force The increase of stiffness at higher moments can be explained by a simple model, see Fig. 6. The SEC is then thought as the serial connection of a common part with stiffness K and a part consisting of several parallel 0 divisions. These parallel parts, anatomically one may think of parallel strands in the aponeurosis, are only effective if the muscle fibres inserting on it (CC) are active. The stiffness of this part would thus be proportional to the amount of recruited muscle fibres, i.e. to M , the 0 isometric muscle moment at optimal length. One parallel part, stiffness K , is connected to PEC1, and thus always 1 engaged. It is assumed that the length of PEC1 is equal to the CC length (i.e. the length of the muscle fibres), so that on release only K is measured. This model leads to Eq. 1 (7). It is seen that in some subjects this stiffness increase is

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800

797

Fig. 4. SEC moment M as a function of SEC stretch / , together with the fit according to eqs. (6a and b). K "288 N m radv1, M "46 Nm % % 1 " calculated from the regression data in Table 2. Measured curve was shifted such that the rightmost point (o) was on the calculated line. R.m.s. error of this recording was 2.7 N m. (Subject 5, same recording as in Fig. 2 and 3, top).

Fig. 5. K as a function of M for subject 5. Regression line according to Eq. (7), with c "2.8 radv1, K "80 Nm radv1 and K "1030 N m radv1. The 1 0 , 1 0 coefficient of correlation was 0.94.

Table 2 Elasticity parameters according to eqs. (6)—(8). The right column gives the average r.m.s. error between measured and calculated curves, from 11 to 18 per subject Subject

c , (rad~1)

K 0 (Nm rad~1)

K p (Nm rad~1)

c "

K at 100 1 Nm(Nm rad~1)

R.m.s. error(Nm)

1 2 3 4 5 6 7 8 9 10 11 12

7.9 4.5 4.1 4.8 2.8 5.6 5.0 14 7.3 6.8 4.2 3.7

470 780 650 1050 1030 510 770 390 500 450 900 670

110 40 70 180 80 80 110 110 110 270 50 50

0.55 0.70 0.55 0.40 0.40 0.55 0.60 0.60 0.55 0.30 0.50 0.45

309 301 276 405 267 283 340 309 313 305 309 258

2.7 2.7 4.6 5.0 3.0 2.6 3.8 3.6 3.2 2.7 4.5 2.0

Mean S.D.

5.9 3.0

681 227

105 64

0.51 0.11

306 39

3.37 0.94

798

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800

ness values at the maximal muscle moment. If this maximum is estimated at 250 N m, K values in our subjects 1 are between 350 and 600 (average 430) N m radv1. Given the uncertainty in the estimates, this is a surprisingly good agreement. The general form of the SEC curve is also in good agreement with the data on isolated tendons, e.g. of Schechtman and Bader (1997) on human EDL or Scott and Loeb (1995) on cat soleus: a toe region up to M "0.4M followed by a linear part. " 0 Fig. 6. Model of the SEC and CC to explain variation of effective stiffness with initial muscle force. The SEC is divided into a common part, with stiffness K , and a number of parallel subdivisions, all 0 connected to a part of the muscle fibres. One part, with stiffness K is p connected to PEC1, and always engaged. See discussion.

4.3. Earlier data

very evident, i.e. in those with low c and high K , while , 0 in others the effect is much less. The K values themsel1 ves, e.g. those at a moment of 100 N m (Table 2) are of the same order of magnitude: 300$40 N m rad~1. Subject 4, with the highest K , had also an exceptionally high 1 voluntary moment (Table 1). The difference should be noted between this increase of stiffness due to the engagement of more aponeurosis strands, and the slope of a single SEC release curve. The SEC curve can thus not be found by integrating stiffness K . It is often stated (e.g. Proske and Morgan, 1987) that 1 an elasticity component proportional to muscle force should be located in the muscle fibres (crossbridges). For human triceps surae this is improbable because of the short muscle fibres (Alexander and Bennett-Clark, 1977). Assuming that crossbridges are completely released at a shortening of 2%, crossbridge stiffness would be equivalent to c "50 rad~1, and form only a small part of , total muscle—tendon compliance.

The results presented here seem to be the first data on human muscle elasticity, in which the complete release curve has been determined. A precursor is work on controlled stretches and releases of the elbow flexors by Cavagna et al. (1968). Most earlier work has been directed at parameter estimation of muscle stiffness at a number of values for the moment. Of these the ‘elastic bounce’ method of Cavagna (1970) is the simplest, it calculates plantarflexor muscle stiffness from the frequency of the body bouncing on the feet in tiptoe. Blanpied and Smidt (1991) have applied stretches at a much lower speed (3 rad s~1 ), but were careful to evaluate only the first 62 ms after the stretch, an extension of about 0.08 rad. Their results thus give only the stiffness parameter K , and only a small 1 part of the SEC curve. Their average results of c "4.3 rad~1 and K "81 Nm rad~1 are in excellent , 1 agreement with our results in Table 2. With a pseudorandom movement and a more complicated parameter identification method, Weiss et al. (1988) found values of K between 50 and 110 N m and c between 8 and 11. 1 , The latter values seem considerably higher than ours. A possibility is that their method did not sufficiently differentiate between pre- and post-reflex responses.

4.2. Comparison with material data

4.4. A simpler model

The estimates for K can be obtained from Hooke’s 1 law. The SEC length was estimated as the total muscle length minus the muscle fibre length, giving about 350 mm. Recent data on human extensor digitorum longus (EDL) tendon in vitro (Schechtman and Bader, 1997) give a Young’s modulus of about 1.0 GPa. The data on a wide range of animals gave values between 1 and 2 GPa (Bennett et al., 1986). For the cross-sectional area of human Achilles tendon Voight (1994) found from MRI images values between 45 and 98 mm2 in five subjects, with an average of 65 mm2. Assuming that this area is constant over the whole length of tendon and aponeurosis (Scott and Loeb, 1995), and with an assumed moment arm of 50 mm, this would result in values for K between 320 and 700 (average 464) N m radv1. These 1 values should be compared with the extrapolated stiff-

In our opinion the model as presented here, with a linear elasticity curve preceded by a quadratic toe region, with stiffness increasing with the amount of active muscle fibres, gives the best representation of our findings in the whole group of subjects. For modelling work a simpler model may be adequate. This might consist of a quadratic curve with a constant coefficient: (9) M "K /2 . % 2 % The average values for K have been given in Table 3, 2 together with the r.m.s. error of this fit. It turns out that the simple representation of Eq. (9) is as good as the quadratic-plus-linear model of Eqs. (6a) and (6b) in six out of the 12 subjects. In the others, the deviations for individual curves could be considerable. Especially, the increase of stiffness with force, which the quadratic

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800 Table 3 Parameter K for simplified quadratic model, eq. (9). Right column 2 gives average r.m.s. error between measured and calculated curves Subject

K 2 (Nm rad~1)

R.m.s. error (Nm)

1 2 3 4 5 6 7 8 9 10 11 12

350 295 290 590 250 285 420 400 365 400 370 240

2.2! 2.5! 5.1 5.7 5.1 2.4! 3.8" 4.8 2.6! 2.7! 4.7" 2.4!

Mean S.D.

355 96

3.7 1.3

799

(slow running). This indicates that energy saving by elastic energy is very well possible in human walking and running.

Acknowledgements We thank Jan Peter van Zandwijk (Department of Human Movement Science, Free University, Amsterdam) and Jaap van der Leest for their collaboration.

References

! Good fit. " Fair fit.

model shows only to a limited extent, was not sufficiently represented. 4.5. Elastic energy For a long time there have been conjectures in literature about the possibility of energy storage in elastic structures (Alexander and Bennett-Clark, 1977; Cavagna et al. 1964; Hof, 1981, 1990; van Ingen Schenau et al., 1997). While it is well documented for animals such as kangaroos, horses and camels (Alexander, 1988), data were more speculative for humans. The elastic energy E can be calculated for a given moment M from Eqs. %% (6a) and (6b). This gives (for M 'M ): % "

A B A B

1 M2 1 M2 % # " . E " %- 2 K 6 K 1 1

(10)

(The second term is comparatively small.) In slow running about 45 J of energy is absorbed (‘negative work’) by triceps surae in the first half of stance, while 60 J of positive work is generated in push-off, at a peak moment of 180 N m ( Hof, 1990; Winter, 1983a). In walking at normal speed the corresponding values are 20 J of negative work and 30 J of positive work at 150 N m (Hof, 1981; Winter, 1983b). When the elastic energy at the peak moment is between the values for negative and positive work, saving of a substantial amount of energy is possible. In that case the whole of the negative work may be stored, and only the difference between positive and negative work needs to be provided by an energy consuming muscle fibre contraction. For our subjects the elastic energy was found to be between 23 and 37 J at 150 Nm (walking) and between 31 and 52 J at 180 N m

Alexander, R.McN., Bennet-Clark, H.C., 1977. Storage of elastic strain energy in muscle and other tissues. Nature 265, 114—117. Alexander, R.McN., 1988. Elastic Mechanisms in Animal Movement. Cambridge University Press, Cambridge. Alexander, R.McN., Ker, R.F., 1990. The architecture of leg muscles. In Multiple Muscle Systems. Winters, J.L., Woo, S.-L. (Eds.), Springer, New York, pp. 568—577. Allum, J.H.J., Mauritz, K.-H., Vo¨gele, H., 1982. Stiffness regulation provided by short-latency reflexes in human triceps surae muscles. Brain Research 234, 159—164. Bennett, M.B., Ker, R.F., Dimery, N.J., Alexander, R.McN., 1986. Mechanical properties of various mammalian tendons. Journal of Zoology (London) 209, 537—548. Blanpied, P., Smidt, G.L., 1991. Human plantarflexor stiffness to multiple single-stretch trials. Journal of Biomechanics 25, 29—39. Bressler, B.H., Clinch, N.F., 1974. The compliance of contracting skeletal muscle Journal of Physiology 237, 477—493. Cavagna, G.A., Saibene, F.P., Margaria, R., 1964. Mechanical work in running. Journal of Applied Physiology 19, 249—256. Cavagna, G.A., Dusman, B., Margaria, R., 1968. Positive work done by a previously stretched muscle. Journal of Applied Physiology 24, 21—32. Cavagna, G.A., 1970. Elastic bounce of the body. Journal of Applied Physiology 29, 279—282. Esteki, A., Mansour, J.M., 1996. Experimentally based nonlinear viscoelastic model of joint passive moment. Journal of Biomechanics 29, 443—450. Hill, A.V., 1950. The Series Elastic Component of Muscle. Proceedings of the Royal Society of London B 137, 273—280. Hof, A.L., van den Berg, Jw., 1981. EMG to force processing II: Estimation of parameters of the Hill muscle model for the human triceps surae by means of a calf ergometer. Journal of Biomechanics 14, 759—770. Hof, A.L., 1990. Effects of muscle elasticity in walking and running. In: Winters, J.L., Woo, S.-L. (Eds.) Multiple Muscle Systems. Springer, New York, pp. 591—607. Hof, A.L., 1997a. A controlled-release ergometer for measuring the series-elasticity of the human ankle plantarflexors. Journal of Biomechanics 30, 203—206. Hof, A.L., 1997b. Correcting for limb inertia and compliance in fast ergometers. Journal of Biomechanics 30, 295—297. Ingen Schenau, van, G.J., Bobbert, M.F., Haan, A., 1997. Does elastic energy enhance work and efficiency in the stretch-shortening cycle? Journal of Applied Biomechanics 13, 389—415. Jacobs, R., van Ingen Schenau, G.J., 1992. Intermuscular coordination in a sprint push-off. Journal of Biomechanics 25, 953—965. Proske, U., Morgan, D.L., 1987. Tendon stiffness: methods of measurement and significance for the control of movement. A review. Journal of Biomechanics 20, 75—82. Schechtman, H., Bader, D.L., 1997. In vitro fatigue of human tendons. Journal of Biomechanics 30, 829—835.

800

A.L. Hof / Journal of Biomechanics 31 (1998) 793—800

Scott, S.H., Loeb, G.E., 1995. Mechanical properties of the aponeurosis and tendon of the cat soleus muscle durung whole-muscle isometric contractions. Journal of Morphology 224, 73—86. Voight, M., 1994. Muscle-tendon elasticity in vivo and its significance for human performance. Ph.D. Thesis, University of Copenhagen, Copenhagen. Weiss, P.L., Hunter, I.W., Kearney, R.E., 1984. Rigid polyurethane foam casts for the fixation of human limbs. Medical and Biological Engineering and Computing 22, 603—604.

Weiss, P.L., Hunter, I.W., Kearney, R.E., 1988. Human ankle joint stiffness over the full range of muscle activation levels. Journal of Biomechanics 21, 539—544. Winter, D.A., 1983a. Energy generation and absorption at the ankle and knee during fast, natural, and slow cadences. Clinical Orthopaedics and Related Research 175, 147—154. Winter, D.A., 1983b. Moments of force and mechanical power in jogging. Journal of Biomechanics 16, 91—97.