royalsocietypublishing.org/journal/rsos
+Research
+ Cite this article: van Bijlert PA, van Soest AJK, Schulp AS. 2021 Natural Frequency
+ Method: estimating the preferred walking speed of
Received: 12 August 2020
+Accepted: 15 March 2021
+Subject Category:
+ Locomotor energetics are an important determinant of an animal’ s ecological niche.
+ It is commonly assumed that animals minimize locomotor energy expenditure by
+ selecting gait kinematics tuned to the natural frequencies of relevant body parts.
+ We demonstrate that this allows estimation of the preferred step frequency and
+ walking speed of
Organismal and evolutionary biology
+Subject Areas:
+biomechanics/palaeontology/biomechanics
+Keywords:
+locomotion, cost of transport, theropoda, tetanurae, optimal walking speed
+Author for correspondence:
+Pasha A. van Bijlert e-mail: pasha.vanbijlert@naturalis.nl
+Electronic supplementary material is available online at + https://doi.org/10.6084/m9.figshare.c.
+5369033.
+Natural
+Frequency
+Method:
+estimating
+the
+preferredwalking
+speed
+of
+
+
based
+on tail
+natural
+frequency
+Pasha A. van Bijlert1,3, A. J.‘Knoek’van Soest1,4
+and Anne S. Schulp2,3,5
+1Department of Human Movement Sciences, Faculty of Behavioural and Movement Sciences, + and2Department of Earth Sciences, Faculty of Science, Vrije Universiteit + Amsterdam, Amsterdam, The Netherlands
+3Naturalis Biodiversity Center, Leiden, The Netherlands
+4Research Institute Amsterdam Movement Sciences, Amsterdam, The Netherlands
+5Faculty of Geosciences, Utrecht University, Utrecht, The Netherlands
+PAvB, 0000-0002-5567-7022 ; AJKvS, 0000-0002-1959-1061 ;
+AS, 0000-0001-9389-1540
+© 2021 The Authors. Published by the Royal Society under the terms of the Creative + Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which + permits unrestricted use, provided the original author and source are + credited.
+1. Introduction
+Animals display a variety of gaits and walking speeds. Two locomotor extremes are of + particular interest: the maximal speed, and the energetically optimal walking + speed, at which the metabolic cost of transport (MCOT, in J m −1 kg −1 ) is + minimal for walking gait. In the absence of external task constraints, optimal + walking speed is very close to preferred walking speed (PWS) in humans [1,2], + ratites [ 3 ], horses [ 4 ] and elephants [ 5 ]. It is likely that they are + related in all terrestrial animals, and that animals would tend to forage at this + speed. During locomotion at preferred speed, animals tend to make use of resonance + by matching locomotor frequencies to the (undamped) natural frequencies of their + relevant body parts, sometimes tuned through muscular contractions [ 6 – 13 ]. + Indeed, it has been demonstrated that mammalian quadrupedal taxa have convergent + natural frequencies between fore- and hindlimbs despite differing limb + morphologies [14,15]. An important advantage of moving a body part close to its + natural frequency is that this reduces mechanical work [16,17]. Based on the + commonly adopted idea that PWS is chosen to minimize MCOT [ 1 – 5 ], it has been + argued that preferred step frequencies are close to the natural frequencies of + relevant body parts [7,11,12]. Determination of the natural frequency of a body + part relevant to locomotion can, therefore, constrain preferred (and presumably, + optimal) step frequencies and walking speeds of extinct taxa [ 10 ].
+ Theropod dinosaurs had tails that were actively involved in locomotion [18,19], but
+ were passively supported through the caudal interspinous ligaments [ 20 – 22 ],
+ thus forming a mass-spring system.
In recent biomechanical models of non-avian dinosaurs, the focus has been on
+ detailed hindlimb muscular reconstructions, while the tail was simplified to a
+ single rigid structure (for instance: [30,41]). It is our view, however, that tail
+ flexibility is an essential aspect of the locomotion of nonavian dinosaurs. We,
+ therefore, chose to focus on the tail, because investigating the dynamic effects
+ of a compliant tail may provide interesting insights into locomotor capabilities.
+ The primary goal of this study was to develop a reductionist method to estimate
+ PWS for non-avian dinosaurs. To this end, we performed a detailed morphological
+ reconstruction of the tail, including the caudal interspinous ligaments. By
+ incorporating their spring-like properties in a biomechanical model, we
+ subsequently estimated the natural frequency of the vertical swaying of the tail.
+ Using this as an analogue for step frequency, we then combined the tail natural
+ frequency with trackway data to estimate the PWS. We will refer to this approach
+ as the Natural Frequency Method (NFM), and demonstrate its application by
+ determining the PWS for
2. Material and methods
+ We estimated tail natural frequency of
Inertial parameters of this model were based on a 3D volumetric musculoskeletal
+ reconstruction. This reconstruction was performed on adult
Taking the morphological reconstruction as starting point, we then defined a
+ simplified biomechanical model consisting of five rigid bodies, connected in hinge
+ joints, with a nonlinear rotational spring at each joint ( and for angle
+ definitions). During walking, the most prominent movements and forces occur in the
+ sagittal plane, and the moment arm of the CFL is also the largest in this plane.
+ Furthermore, the exchange between gravitational and elastic potential energy is
+ most meaningfully studied in this plane. Therefore, similar to previous
+ researchers [27,30], we elected to perform a sagittal plane analysis. Inertial
+ properties of the rigid bodies were acquired from the corresponding segments of
+ the morphological reconstruction. Each of the rotational springs was assumed to
+ generate a torque that increased quadratically with joint angle when stretched [
+ 42 ]. As a result, each joint spring has two parameters: joint angle at which the
+ spring torque is zero, and a stiffness parameter. To assign values to these
+ parameters, the biomechanical model was aligned with the morphological
+ reconstruction in two different postures in which the spring torques were known (
+ ). We acquired the first posture by defining passive horizontal equilibrium, which
+ implies that the ligaments are strained in horizontal posture to counteract
+ gravity ( ). It has been shown that the interspinous ligaments could generate
+ ample force to maintain this pose [ 20 ], which our models confirm (see the
+ electronic supplementary material). In the second posture, all interspinous
+ ligaments were at resting length, and thus, all spring torques equalled zero. We
+ determined this posture on the basis of an assumption regarding the ligament
+ strain in horizontal equilibrium. Tendons and ligaments can be roughly divided
+ into low-stress and high-stress varieties, with the high-stress varieties
+ providing more energy savings at the cost of lower safety factors [ 39 ]. We
+ imposed an intermediate 4% strain in horizontal equilibrium ( εhor = 0.04), and
+ used this to find a pose where the ligaments are not strained ( ). This is
+ analogous to passive equilibrium in the absence of gravity, or a
In the absence of damping, the resonance of the tail would occur at the undamped + natural frequency ( fn ) of the system [ 43 ]. In reality, damping is omnipresent. + In that case, resonance occurs at a forcing frequency of fn · (1 – 2 β2 ) 0.5 , + with β being the damping ratio [ 43 ]. Ligaments (like tendons) display relatively + little energy loss during work-loop experiments [39,44], so they have low damping + ratios. However, damping of dorsoventral oscillations would be dependent on more + than just the ligaments, and a structure-based estimate of damping would be + difficult, if not impossible. Therefore, we will use the undamped natural + frequency of the biomechanical model as a proxy for the resonance frequency. To + calculate the natural frequency of the biomechanical model, we first linearized + the equations of motion in horizontal equilibrium using standard numerical + methods. Next, the five pairs of conjugated purely imaginary eigenvalues of the + system matrix were determined using the ‘eig’ function in MATLAB. Subsequently, + the natural frequency was calculated as the oscillation frequency pertaining to + the fundamental resonance mode, i.e. the mode with the lowest eigenfrequency. + Finally, walking speed was estimated by multiplying the natural frequency of each + model by step length, defined as half the distance between two consecutive + footfalls of the same foot. We determined a step length of 1.94 m by scaling a + large tyrannosaurid trackway [ 45 ] based on footprint length of RGM.792000 (see + the electronic supplementary material for details).
+As was done in previous studies on dinosaur dimensions [27,33,46,47], we subjected + the major inputs to a sensitivity analysis. Since we hope to see future studies of + dinosaur locomotion incorporate our tail model, we intentionally chose wide bounds + for the sensitivity analysis. This serves to inform researchers which individual + steps in the modelling process induce the largest variation in the result. These + steps could then be the focus of future research in an attempt to reduce the + uncertainty.
+To account for uncertainty in inertial estimates, we individually varied length (and + thus moment of inertia) and mass. Because we assumed 4% ligament strain in + horizontal equilibrium, we investigated the effects of low- and high-stress + ligaments on the result, by imposing 3% and 5% strain in horizontal equilibrium, + respectively [ 39 ]. Although each ligament is well-bounded by the skeleton, the + effective point of force application of each ligament was unknown. In the baseline + model, we based the moment arms on the area centroids of the ligaments. We used + the ligament reconstruction to determine bounds for the moment arms (electronic + supplementary material, figures S5 and S7). Lastly, whereas the proximal tail + shows pronounced articulations between the zygapophyses, after approximately the + 13th caudal this is no longer the case (electronic supplementary material, figure + S5). This shifts the rotational axes in the distal tail ventrally, towards the + vertebral centra, with the vertical position dependent on whether the tail is in + flexion or extension (electronic supplementary material, figure S6 shows the + baseline and two extreme possibilities). We, therefore, bounded the rotational + axes in tail segments 3 and 4, which affected joint springs 3–5. The dorsal axes + are unrealistically high, ensuring a wide bound for the sensitivity analysis. The + ventral axes are located at the vertebral articulation with the chevrons, and + represent their mechanical effect when the tail is flexed. In total, the + aforementioned bounds led to 11 different models.
+To encourage the adoption of our compliant tail model for dinosaur locomotion, we + have provided in-depth descriptions of reconstruction and biomechanical modelling + steps, mathematical derivations, scaling relationships and supplementary results + in a combined document titled ‘Supplementary texts’ as electronic supplementary + material. Reconstructions, 3D modelling and measurements were all done in + Rhinoceros 6 (McNeel and Associates, Seattle, WA, USA). The biomechanical model + was constructed using custom-code written in MATLAB 2019b (MathWorks, Natick, MA, + USA). A 3D model of the iliosacrum and caudal skeleton of RGM.792000, and all + custom MATLAB scripts are provided as electronic supplementary material.
+3. Results
+ Natural frequencies and walking speeds determined in this study, as well as key
+ outcomes of the sensitivity analysis, are reported in table 1 . The natural
+ frequency of our baseline model of the tail was 0.66 s −1 , which we interpreted
+ as the preferred step frequency for
The effects of changes in length (and thus inertia), mass and strain are + mathematically predictable in our model, and we have thus derived general scaling + laws for these parameters in the electronic supplementary material. Mass had zero + effect on the natural frequency. Length ( L ) and strain in the horizontal + position ( εhor ) are related to the natural frequency ( fn ) as follows: fn ∝ + L−0.5 and fn ∝ 1 hor— 0:5 . These relations can also be applied to the altered + moment arms or distal axes models. For instance, a 6% longer tail, with 10% more + strain in horizontal equilibrium (4.4% strain instead of 4%), using the maximal + moment arms, would lead to a natural frequency of 0.79 × 1.06 −0.5 × 1.1 −0.5 = + 0.73 s −1 . The a These models only changed the moment arms within the 3rd – 5th + segments; segments 1 and 2 remained unchanged from the baseline model. moment arms + models cannot be combined with the distal axes models in this way, because these + models are based on measurements of morphological traits, instead of systematic + parameter variations. Their combinations are reported in electronic supplementary + material, table S2, and the above proportionality relations can be applied to + those combination effects as well.
+modelrelative change in tail natural frequency from baseline (%)tail natural + frequency (s−1)corresponding walking speed (m s−1)baselinen.a.0.661.28mass +8% + mass −8% length +8% length −8% high-strain (+25%) low-strain (−25%)0 0 −3.8 +4.3 + −10.6 +15.50.66 0.66 0.63 0.69 0.59 0.761.28 1.28 1.23 1.33 1.14 1.48max. moment + arms min. moment arms ventral axesadorsal axesa+20.6 −14.3 +6.7 −13.30.79 0.56 + 0.70 0.571.54 1.09 1.36 1.11
+Varying the moment arms had the largest overall effect on tail natural frequency, + followed by the models accounting for strain magnitude. Together, these models + accounted for uncertainty in the effective point of force application and + composition of the caudal interspinous ligaments. We have supplied a spreadsheet + as electronic supplementary material that interactively calculates natural + frequency and walking speed, depending on parameter changes input by the user.
+4. Discussion
+ The primary goal of this study was to develop a reductionist method to estimate the
+ tail natural frequency of
Our model predicts several scaling laws which provide insight into the locomotion of
+
It has previously been proposed that the distal tail may act as a ‘dynamic + stabilizer’ [ 19 ]. Varying the distal rotational axes revealed that even though + the distal three segments accounted for only 14.5% of the total caudal mass, they + still had a meaningful effect on the natural frequency ( table 1 ). Our results + support the notion that the medio-distal tail plays an important role in overall + tail dynamics. Furthermore, the ventral axes model suggests that the chevrons + increase locomotor speed when the tail is in slight flexion, by increasing the + moment arm of the interspinous ligaments. Combining altered moment arms with + alternative rotational axes strengthened this effect (electronic supplementary + material, table S2). However, this table should be interpreted with caution + because the distal axes models do not represent axes that would have been used + throughout the whole step cycle. Instead, the rotational axes would migrate + ventrally during flexion, and dorsally during extension of the tail, which could + potentially be influenced by contraction of the epaxial and hypaxial musculature. + Thus, the bounds in electronic supplementary material, table S2 are + unrealistically wide, due to the assumptions underlying this table. Yet, to + acknowledge the uncertainties in dinosaur gait reconstruction, we used these as + the reported range in our result.
+Varying the amount of strain imposed on the ligaments in the horizontal posture ( + εhor , expressed as a fraction) had a similar effect as varying the length: fn ∝ 1 + hor— 0:5 . This is analogous to increasing or decreasing the stiffness at a + constant moment of inertia. Ligaments that are subject to high levels of strain + have increased energy-saving properties [ 39 ]. Our high-strain model displayed a + lower natural frequency, implying that selective pressures for higher PWS may + result in less energy-efficient locomotion, and vice versa.
+ Our baseline model suggests a PWS of 1.28 m s −1 for
Very little work has been done to directly investigate PWS of dinosaurs. Limb
+ natural frequencies have been used to estimate ‘comfortable walking speeds’ of
+ several sauropod taxa [ 10 ], and dynamic simulations were used to estimate PWS of
+
Recent amendments have been proposed to improve DS-based estimations, either by
+ focusing on swing phase dynamics [ 52 ], or by incorporating taxon-specific
+ morphological parameters to reduce uncertainty [ 53 ]. The latter are of course
+ difficult to obtain from a fossilized trackway. Our method can help in this
+ regard, by providing a range of plausible step frequencies for any given taxon.
+ With this goal in mind, it would be most effective to estimate the natural
+ frequencies of taxa that are wellrepresented in the trackway record. That way, it
+ would be possible to combine step length data from several trackways, increasing
+ confidence that the average result should tend towards preferred (and therefore
+ optimal) gait. Unfortunately, this was not possible in the present study, because
+ large theropod trackways with reasonably close taxonomic proximity to
+
DS does not incorporate MCOT, and therefore cannot be used to estimate PWS of + dinosaurs. Therefore, when estimating dinosaur foraging costs, researchers have + used DS to scale the walking speed [ 29 ], or even kept it at a constant 2 m s −1 + for larger taxa [ 31 ]. However, methods to calculate MCOT are heavily dependent + on both speed and mass [ 28 ], so we suggest that any comparison of foraging costs + should use the PWS of the respective taxa as a starting point. This could be done + using the natural frequency of the legs [ 10 ], musculoskeletal simulations [ 49 + ], or NFM as we propose it.
+ To our knowledge, musculoskeletal simulation models have not been used to estimate
+ how MCOT varied with walking speed in
Our method for walking speed estimation is meant to be a reductionist analysis, so
+ we have only incorporated the interspinous ligaments, since we expect their
+ mechanical effect to dominate overall tail dynamics. Natural frequencies play an
+ important role in animal locomotion [ 6 – 13 ], and NFM, therefore, provides a
+ reasonable starting point for investigating how the resonant properties of
+ dinosaur tails affect overall locomotion. However, in reducing the analysis to the
+ mechanics of a single structure, many of the complex interactions are ignored. For
+ instance, animals tune the frequencies of their segments through muscular
+ contractions [6,7].
Tail musculature could also be employed to enforce beneficial phase relationships
+ between vertical hip and tail oscillations. During walking, vertical oscillations
+ between withers and heads in most ungulates are out of phase, which provides them
+ with an energetic benefit [36–38,55]. This is theorized to be modulated by natural
+ frequencies of the neck segment [37,38], but there is an active component to this
+ behaviour as well [ 36 ]. This phase relationship reduces the losses incurred
+ during the step-to-step transitions, which are identified as major sources of
+ energy losses in inverted pendulum models of walking [ 56 ]. The implications of
+ energy storage are also further complicated by the serial elasticity in
+ musculotendon complexes: tendon stiffness can substantially impact the mechanical
+ work done by the muscle fibres [ 57 ]. Finally, the cost of cyclical muscle
+ contractions could also affect metabolic optima, depending on the task
+ requirements [16,17]. Such complex interactions would undoubtedly affect PWS, but
+ require extensive reconstructions of muscular contractile properties. Muscular
+ parameters currently provide the biggest uncertainty in most musculoskeletal
+ simulations of dinosaurs [41,49,58], so any investigation of these interactions
+ would require careful consideration. Inclusion of our compliant tail model into
+ fully actuated hindlimb simulations of
We set out to develop a method that does not require assumptions regarding + contractile properties, so we accept that this limits the predictive power of our + method. Given the overall uncertainties we must deal with when investigating the + locomotion of extinct animals, we consider the simplicity of the method to be one + of its strengths. Provided the neck or tail posture was supported passively, our + method could be used to arrive at an estimate of the PWS of any sufficiently + complete dinosaur, without the need to estimate contractile properties.
+5. Conclusion
+ Gait reconstruction of dinosaurs has numerous inherent uncertainties, and therefore
+ it is important to compare results from different methods, in an attempt to find a
+ convergent point. We have proposed a method based on reconstructing the natural
+ frequency of the vertical swaying of the tail, which we refer to as the Natural
+ Frequency Method. This method requires relatively few assumptions, which are
+ furthermore different assumptions to other methods. Our results for preferred
+ walking speed of
Data accessibility. Supplementary texts, 7 and tables S1–S 2, video S1, a spreadsheet + to calculate all possible parameter effects of our models, custom MATLAB code + written for the analyses and 3D scans of the iliosacrum and caudal skeleton of + RGM.792000 are provided as electronic supplementary material.
+Authors’contributions. P.A.v.B., A.J.K.v.S. and A.S. conceived the project. + A.J.K.v.S. designed the biomechanical model. P.A.v.B. performed the + reconstructions, measurements, coding. P.A.v.B., A.J.K.v.S. and A.S. wrote the + manuscript.
+Competing interests. The authors declare no competing interests;
+Funding. Nothing to declare.
+Acknowledgements. We thank K. Lemaire, M. Bobbert and D. Kistemaker for their vital + role in early discussions. Further thanks to V. Vanhecke for 3D scanning, P. + Larson for providing excavation maps and D. Tanke for providing photographs. H. + Mallison and P. Manning are thanked for in-depth discussion and feedback on + earlier versions of this manuscript. Discussions with W. Sellers, K. Bates and F. + van Diggelen further helped us to contextualize our work. Lastly, we extend our + gratitude to two anonymous reviewers whose constructive feedback and insights + helped shape the final manuscript.
+References
+1. Browning RC , Baker EA , Herron JA , Kram R.2006Effects of obesity and sex on the + energetic cost and preferred speed of walking . J. Appl. Physiol.100 , 390 – 398 . + (doi: 10.1152/ japplphysiol.00767.2005 )
+2. Gast K , Kram R , Riemer R.2019Preferred walking speed on rough terrain: is it + all about energetics?J. Exp. Biol.222 , jeb185447. (doi: 10. 1242/jeb.185447 )
+3. Watson RR , Rubenson J , Coder L , Hoyt DF , Propert MWG , Marsh RL . + 2011Gait-specific energetics contributes to economical walking and running in emus + and ostriches . Proc. R. Soc. B278 , 2040 – 2046 . (doi: + 10.1098/rspb.2010.2022)
+4. Hoyt DF , Taylor R.1981Gait and the energetics of locomotion in horses . + Nature292 , 239 – 240 . (doi: 10.1038/292239a0)
+5. Langman VA , Roberts TJ , Black J , Maloiy GM , Heglund NC , Weber J-M , Kram R , + Taylor CR . 1995Moving cheaply: energetics of walking in the African elephant . J. + Exp. Biol.198 , 629 – 632 .
+6. Ahlborn BK , Blake RW . 2002Walking and running at resonance . Zoology105 , 165 – + 174 . (doi: 10.1078/0944-2006-00057)
+7. Ahlborn BK , Blake RW , Megill WM . 2006Frequency tuning in animal locomotion . + Zoology109 , 43 – 53 . (doi: 10.1016/j.zool.2005. 11.001 )
+8. Wagenaar RC , Van Emmerik REA . 2000Resonant frequencies of arms and legs + identify different walking patterns . J. Biomech.33 , 853 – 861 . (doi: + 10.1016/S0021-9290(00)00020-8)
+9. Huat OJ , Ghista DN , Beng NK , John TCC . 2004Optimal stride frequency computed + from the double-compound pendulum of the leg, and verified experimentally as the + preferred stride frequency of jogging . Int. J. Comput. Appl.Technol.21 , 46 – 51 + . (doi: 10.1504/ijcat.2004.005338)
+10. Christian A , Müller RHG , Christian G , Preuschoft H.1999Limb swinging in + elephants and giraffes and implications for the reconstruction of limb movements + and speed estimates in large dinosaurs . Foss. Rec.2 , 81 – 90 . (doi: + 10.5194/fr-2-81-1999)
+11. Holt KG , Hamill J , Andres RO . 1990The forcedriven harmonic oscillator as a + model for human locomotion . Hum. Mov. Sci.9 , 55 – 68 . (doi: + 10.1016/0167-9457(90)90035-C)
+12. Holt KG , Hamill J , Andres RO . 1991Predicting the minimal energy costs of + human walking . Med. Sci. Sports Exerc.23 , 491 – 498 . (doi: 10. + 1249/00005768-199104000-00016 )
+13. Lee LF , Krovi VN . 2008Musculoskeletal simulation-based parametric study of + optimal gait frequency in biped locomotion . In Proc. of 2nd Biennial + IEEE/RAS-EMBS Int. Conf. on Biomedical Robotics and Biomechatronics, BioRob 2008, + Scottsdale, AZ, 19–22 October , pp. 354 – 359 . IEEE . (doi: 10.1109/BIOROB.2008. + 4762864 )
+14. Myers MJ , Steudel K.1997 Morphological conservation of limb natural pendular + period in the domestic dog ( Canis familiaris ): implications for locomotor + energetics . J. Morphol. 234 , 183 – 196 . (doi: 10.1002/(SICI)1097- + 4687(199711)234:2<183::AID-JMOR5>3.0.CO;2-D )
+15. Raichlen DA . 2004 Convergence of forelimb and hindlimb natural pendular period + in baboons ( Papio cynocephalus ) and its implication for the evolution of primate + quadrupedalism . J.Hum. Evol.46 , 719 – 738 . (doi: + 10.1016/j.jhevol.2004.04.002)
+16. Doke J , Donelan JM , Kuo AD . 2005Mechanics and energetics of swinging the + human leg . J.Exp.Biol.208 , 439 – 445 . (doi: 10.1242/jeb.01408)
+17. Dean JC , Kuo AD . 2011Energetic costs of producing muscle work and force in a + cyclical human bouncing task . J. Appl. Physiol.110 , 873 – 880 . (doi: + 10.1152/japplphysiol.00505.2010)
+18. Gatesy SM . 1990Caudofemoral musculature and the evolution of theropod + locomotion . Paleobiology16 , 170 – 186 .
+19. Farlow JO , Gatesy SM , Holtz TR , Hutchinson JR , Robinson JM . 2000Theropod + locomotion . Am. Zool.40 , 640 – 663 . (doi: 10.1093/icb/40.4.640)
+ 20. Hengst R.2004 Gravity and the
21. Carpenter K.2013 A closer look at the hypothesis of scavenging versus predation + by Tyrannosaurus rex . In Tyrannosaurid paleobiology (eds JM Parrish , RE Molnar , + PJ Currie , EB Koppelhus ), pp. 265 – 277 . Bloomington, Indiana: Indiana + University Press .
+22. Wilson JP , Woodruff DC , Gardner JD , Flora HM , Horner JR , Organ CL . + 2016Vertebral adaptations to large body size in theropod dinosaurs . PLoS ONE11 , + e 0158962 . (doi: 10.1371/journal.pone.0158962)
+23. Alexander RM . 1976Estimates of speeds of dinosaurs . Nature261 , 129 – 130 . + (doi: 10.1038/ 261129a0 )
+24. Thulborn RA . 1984Preferred gaits of bipedal dinosaurs . Alcheringa Australas J. + Palaeontol.8 , 243 – 252 . (doi: 10.1080/03115518408618947)
+25. Hutchinson JR , Garcia M.2002Tyrannosaurus was not a fast runner . Nature416 , + 1018 – 2021 .
+26. Hutchinson JR . 2004 Biomechanical modeling and sensitivity analysis of bipedal + running ability. II. Extinct taxa. J. Morphol.262 , 441 – 461 . (doi: + 10.1002/jmor.10240 )
+27. Sellers WI , Manning PL . 2007Estimating dinosaur maximum running speeds using + evolutionary robotics . Proc. R. Soc. B274 , 2711 – 2716 . (doi: + 10.1098/rspb.2007.0846 )
+28. Pontzer H , Allen V , Hutchinson JR . 2009Biomechanics of running indicates + endothermy in bipedal dinosaurs . PLoS ONE4 , e 7783 . (doi: + 10.1371/journal.pone.0007783 )
+29. Kane A , Healy K , Ruxton GD , Jackson AL . 2016Body size as a driver of + scavenging in theropod dinosaurs . Am. Nat.187 , 706 – 716 . (doi: 10. 1086/686094 + )
+30. Sellers WI , Pond SB , Brassey CA , Manning PL , Bates KT . 2017 Investigating + the running abilities of Tyrannosaurus rex using stressconstrained multibody + dynamic analysis . PeerJ5 , 1 – 19 . (doi: 10.7717/peerj.3420 )
+31. Dececchi TA , Mloszewska AM , Holtz TR , Habib MB , Larsson HCE . 2020The fast + and the frugal: divergent locomotory strategies drive limb lengthening in theropod + dinosaurs . PLoS ONE15 , e 0223698 . (doi: 10.1371/journal.pone.0223698 )
+32. Ruiz J.2018 Comments on ‘A tyrannosaur trackway at Glenrock, Lance Formation + (Maastrichtian), Wyoming’ (Smith et al ., Cretaceous Research, v. 61, pp. 1 – 4, + 2016). Cretac. Res.82 , 81 – 82 . (doi: 10.1016/j.cretres. 2017.05.033 )
33. + Bates KT , Manning PL , Hodgetts D , Sellers WI . 2009Estimating mass + properties of dinosaurs using laser imaging and 3D computer modelling . PLoS + ONE4 , e 4532 . (doi: 10.1371/ journal.pone.0004532 )
+34. Bates KT , Benson RBJ , Falkingham PL . 2012A computational analysis of + locomotor anatomy and body mass evolution in Allosauroidea (Dinosauria: + Theropoda) . Paleobiology38 , 486 – 507 . (doi: 10.5061/dryad.09kf4g02 )
+35. Persons WS , Currie PJ . 2011 The tail of Tyrannosaurus : reassessing the + size and locomotive Importance of the M. caudofemoralis in non-avian + theropods . Anat. Rec.294 , 119 – 131 . (doi: 10.1002/ar.21290 )
+36. Gellman KS , Bertram JEA . 2002The equine nuchal ligament 2: passive + dynamic energy exchange in locomotion . Vet. Comp. Orthop. Traumatol.15 , 7 + – 14 . (doi: 10.1055/s-0038-1632706 )
+37. Loscher DM , Meyer F , Kracht K , Nyakatura JA . 2016Timing of head + movements is consistent with energy minimization in walking ungulates . + Proc. R. Soc. B283 , 20161908 . (doi: 10.1098/ rspb.2016.1908 )
+38. Basu C , Wilson AM , Hutchinson JR . 2019The locomotor kinematics and + ground reaction forces of walking giraffes . J. Exp. Biol.222 , jeb 159277 . + (doi: 10.1242/jeb.159277 )
+39. Biewener AA . 2008Tendons and ligaments: structure, mechanical behavior + and biological function . In Collagen (ed. P Fratzl ), pp. 269 – 284 . + Boston, MA: Springer . (doi: 10.1007/ 978-0-387-73906-9_10 )
+40. Manning PL . 2008Spring in the tail? In Grave secrets of dinosaurs: soft + tissues and hard science , pp. 292 – 295 . Washington: National Geographic + Society .
+41. Bishop PJ , Cuff AR , Hutchinson JR . 2020How to build a dinosaur: + musculoskeletal modeling and simulation of locomotor biomechanics in extinct + animals . Paleobiology47 , 1 – 38 . (doi: 10.1017/ pab.2020.46 )
+42. Scholz MN , Bobbert MF , van Soest AJ , Clark JR , van Heerden + J.2008Running biomechanics: shorter heels, better economy . J. Exp. Biol.211 + , 3266 – 3271 . (doi: 10.1242/jeb.018812 )
+43. Ruina A , Pratap R.2019Mechanics Toolset, Statics and Dynamics . See + http://ruina.tam. cornell.edu/Book/ .
+44. Alexander RM . 2006Principles of animal locomotion . Princeton, NJ: + Princeton University Press .
+45. McCrea RT , Buckley LG , Farlow JO , Lockley MG , Currie PJ , Matthews NA + , Pemberton SG . 2014 A ‘terror of tyrannosaurs’ : the first trackways of + tyrannosaurids and evidence of gregariousness and pathology in + Tyrannosauridae . PLoS ONE9 , e 103613 . (doi: 10.1371/journal.pone.0103613 + )
+