The human iliotibial band is specialized for elastic energy storage compared with the chimp fascia lata

Bipedalism appears to be the initial derived feature that set the human lineage on a separate evolutionary trajectory from the African great apes (Darwin, 1871; Haile-Selassie, 2001; Zollikofer et al., 2005). There are many hypotheses about the selective pressures that favored the origin of hominin bipedalism, but one of the most widely accepted is that natural selection favored bipedal walking over more costly quadrupedal gaits, such as knuckle-walking, to reduce the energetic costs of travel as climatic shifts increased the distances between food sources (Rodman and McHenry, 1980; Sockol et al., 2007). This idea is supported by evidence that many adaptations in the lower limb associated with bipedalism, such as relatively longer limbs and shorter toes, also benefit locomotor economy (Pontzer, 2007; Rolian et al., 2009).

The iliotibial band (ITB) is a unique structure in the human lower limb, derived from the fascia lata (FL) of the thigh, which may contribute to locomotor economy (Fig. 1). The ITB is not present in other apes and thus almost certainly evolved independently in hominins, but its role in human locomotion is not well understood. Although the most common view of the ITB's function is to stabilize the pelvis in the frontal plane (Inman, 1947; Kaplan, 1958; Stern, 1972; Gottschalk et al., 1989), we recently created a musculoskeletal model of the ITB to investigate whether forces generated by the tensor fascia lata (TFL) or gluteus maximus (GMax) substantially stretch the ITB during running, storing elastic energy that is recovered later in the stride (Eng et al., 2015). We estimated that the anterior ITB stores about 1 J of energy per stride during the late stance and early swing phases, whereas the posterior ITB stores about 6 J per stride during the late swing phase at fast running speeds (5 m s⁻¹). Here, we interpret these findings within a broader comparative context by examining whether the chimp FL also stores elastic energy or whether the capacity of the human ITB to store energy during locomotion is unique to humans among apes. This study also examines whether the human ITB has a greater capacity than the chimp FL to transmit moments that stabilize the pelvis in the frontal plane during walking.

Chimpanzees (Hominidae: Pan) are a key comparative species for interpreting the derived nature of the ITB because chimps are the extant sister taxon to Homo (Ruvolo, 1994; Satta et al., 2000). Given the many morphological similarities between chimps and gorillas, with many differences likely resulting from the effects of size (Shea, 1985; Berge and Penin, 2004), it is most parsimonious to infer that the last common ancestor (LCA) of Pan and humans resembled Pan in morphology and was a knuckle-walking ape with no ITB (Pilbeam, 1996; Wrangham and Pilbeam, 2001; but see Sayers and Lovejoy, 2008; Lovejoy, 2009; Almécija et al., 2013). It is also possible that the ITB was convergently lost in both gorillas and chimpanzees, but given the lack of an ITB in other apes (Swindler and Wood, 1973; Sigmon, 1975), it is reasonable to assume that the ITB is a human autapomorphy. Therefore, this study compares the function of the human ITB with that of the chimp FL using musculoskeletal models developed from detailed anatomical experiments.

The ability of fascial connective tissues to store elastic energy depends in part on the muscular forces these tissues transmit, which likely differ between chimps and humans. Estimating these forces during walking requires information about the activation patterns, force–length properties and excursions of the inserting muscles. However, descriptions of relevant chimp anatomy in the literature vary considerably and are limited by small sample sizes (Stern, 1972; Swindler and Wood, 1973; Sigmon, 1975; Sigmon and Farslow, 1986). According to most studies, the TFL inserts on the chimp FL (Sigmon, 1974,, 1975; Sigmon and Farslow, 1986), as it does on the human ITB. However, compared with the human TFL, the chimp TFL is thought to be smaller and partially fused with the anteriormost portion of the cranial GMax (Sigmon, 1974,, 1975; Sigmon and Farslow, 1986). When referring to the GMax in chimps, we distinguish between the cranial portion (GMaxCr; also called gluteus maximus proprius) and the caudal portion (GMaxCd; also called ischiofemoralis). There is general agreement that GMaxCr is thinner and less massive than the homologous GMax in humans (Stern, 1972; Swindler and Wood, 1973; Sigmon, 1975; Lieberman et al., 2006), but there is a lack of consensus about where GMaxCr inserts. Some studies report that GMaxCr inserts on the FL (Swindler and Wood, 1973; Sigmon, 1974,, 1975), but others report that insertion on the FL is rare (Preuschoft, 1961; Stern, 1972). The GMaxCd is thought to insert not into the chimp FL, but along the femoral shaft from the gluteal tuberosity to the lateral epicondyle; this muscle is absent in humans (Stern, 1972; Sigmon, 1974,, 1975). Thus, characterizing the energy storage capacity of the chimp FL required us to identify more definitively the portions of TFL, GMaxCr and GMaxCd that insert on the chimp FL and to characterize the muscle architecture of these muscles.

The amount of elastic energy stored by the chimp FL and the human ITB during walking also depends on the length changes of these fascial tissues, which depend on the length changes of the muscle–tendon units (MTUs). These length changes depend on the moment arms of the MTUs about the hip and knee and the changes in hip and knee angles during walking. Because moment arms generally vary as a function of joint angle (An et al., 1984; Hoy et al., 1990; Spoor et al., 1990), published descriptions of chimp moment arms based solely on bone morphology may not accurately predict changes in MTU length during walking. Thus, characterizing the energy storage capacity of the chimp FL also required us to measure the moment arms of TFL, GMaxCr and GMaxCd in chimpanzee cadavers over the ranges of hip and knee angles corresponding to walking. Although moment arm data are available for several chimp muscles (Thorpe et al., 1999; Payne et al., 2006; Holowka and O'Neill, 2013), and a detailed model of the chimp lower limb has been developed based on these data (O'Neill et al., 2013), the moment arms of TFL and the portions of GMax that insert on the FL have not previously been reported.

It is commonly thought that the human ITB functions to stabilize the pelvis in the frontal plane when tensed by the inserting muscles (Inman, 1947; Kaplan, 1958; Stern, 1972; Gottschalk et al., 1989). Unlike quadrupedal apes, bipedal hominins must stabilize the body's center of mass over a small area of support. Thus, selection for increased abduction moment capacity may have acted on the ITB or other structures to help stabilize the pelvis in the frontal plane during walking. In this study, we used our musculoskeletal models to compare the frontal plane moments transmitted by the chimp FL and the human ITB during midstance.

In summary, we measured the masses, fascicle lengths, pennation angles, and moment arms of the TFL and GMax in four chimpanzee cadavers. These data were used to refine a 3D musculoskeletal model of the chimp hindlimb (O'Neill et al., 2013) that includes the attachments and force–length properties of the FL and the inserting muscles. The model was used with kinematic descriptions of the chimp's hip and knee angles during bipedal walking to compute the forces and corresponding strains in the chimp FL and to predict its capacity for elastic energy storage and recovery (Fig. 2). Results were compared with analogous data from a model of the human ITB (Eng et al., 2015) to test four specific hypotheses: (H1) the muscles inserting on the human ITB have a greater force-generating capacity than the muscles inserting on the chimp FL, accounting for differences in body mass; (H2) the human ITB undergoes greater strains than the chimp FL during typical bipedal kinematics; (H3) the human ITB has a substantially greater potential to store elastic energy per unit body mass than the chimp FL during bipedal walking; (H4) the human ITB transmits substantially larger moments about the hip in the frontal plane than the chimp FL during midstance, which would provide some evidence that the human ITB may be specialized for frontal plane stability.

Dissections of the chimp TFL, GMaxCr and GMaxCd revealed that portions of all three muscles insert directly on the FL (Fig. 3). All TFL muscle fibers insert into the anterior FL (TFL–FL_ant). Consistent with findings from some previous studies (Sigmon, 1974,, 1975; Sigmon and Farslow, 1986), our dissections confirmed that the anterior portion of the chimp GMaxCr is fused distally with the TFL and inserts into the anterior FL (Fig. 1A). However, this portion of GMaxCr (GMaxCr–FL_ant) is relatively small, constituting only about 5% of the muscle's total mass (Table 1). Our dissections also revealed that about 25% of the chimp GMaxCd mass (GMaxCd–FL_post) inserts into the posterior FL, consistent with most other reports of chimpanzee anatomy (Champneys, 1871; Sigmon and Farslow, 1986) except Stern (1972).

Moment arm measurements revealed that muscles inserting in the anterior FL flex and abduct the hip, whereas muscles inserting in the posterior FL extend and adduct the hip (Fig. 4). In particular, the chimp TFL–FL_ant has a relatively large hip flexion moment arm. The GMaxCr–FL_ant has a large hip abduction moment arm, but a very small hip flexion moment arm. The chimp GMaxCd–FL_post has a small hip adduction moment arm, but a relatively large hip extension moment arm that increases with hip extension. The moment arms predicted by our model are consistent with results from our tendon excursion measurements (Fig. 4).

Biomechanically, we found that the chimp GMaxCd–Fl_post is more comparable to the portion of the human GMax that inserts on the ITB (GMax–ITB_post) than is the homologous chimp GMaxCr–FL_ant. The chimp GMaxCr–FL_ant has a smaller mass and a more anterior insertion than the human GMax–ITB_post. By contrast, large portions of both the chimp GMaxCd and the human GMax insert posteriorly on the FL and ITB, respectively, and these muscles have similar hip extension moment arms. Therefore, to test our hypotheses, we compared the force-generating capacity and energy storage potential of the human GMax–ITB_post with predictions generated by our model of the chimp GMaxCd–FL_post.

Muscles inserting on the human ITB have the potential to transmit substantially larger forces to the ITB than the muscles inserting on the chimp FL transmit to the FL (Fig. 5). The force-generating capacity of the human TFL is significantly larger than that of the chimp TFL, even after accounting for differences in body mass. In particular, the normalized physiological cross-sectional area (PCSA) of the human TFL (0.24±0.07 cm² kg^−2/3) is more than three times greater than that of the chimp TFL (0.07±0.03 cm² kg^−2/3; P<0.05). The normalized PCSA of the human GMax–ITB_post is more than double that of the chimp GMaxCd–FL_post (0.72±0.13 cm² kg^−2/3 versus 0.30±0.04 cm² kg^−2/3; P<0.05). These data suggest that the human ITB transmits substantially larger muscle forces than the chimp FL.

The ITB MTUs in the human model undergo substantially greater length changes than the FL MTUs in the chimp model during bipedal walking (Fig. 6). These larger MTU excursions are not simply a result of larger human limbs. Rather, these data reflect differences between the moment arms of the chimp FL MTUs (Fig. 4) and the moment arms of the human ITB MTUs (Eng et al., 2015), as well as measured differences in hip and knee angles during walking.

In humans, the TFL stretches when it is active in late stance, similar to the TFL in chimps (Fig. 7). The GMax also stretches when it is active; this lengthening occurs during late swing in the human GMax–ITB_post and during early stance in the chimp GMaxCd–FL_post (Fig. 7). However, because humans extend their hips more than chimps during bipedal walking (∼45 deg range in humans, from −20 deg extension to 25 deg flexion, versus ∼25 deg range in chimps, from 25 to 50 deg flexion; O'Neill et al., 2015), human ITB MTUs undergo substantially greater length changes than chimp FL MTUs throughout the gait cycle (Fig. 6). The human TFL–ITB_ant stretches more than the chimp TFL–FL_ant in late stance because of a larger hip flexion moment arm (slope of the line in Fig. 6B versus A) and the greater hip flexion/extension excursion in humans. The human GMax–ITB_post also stretches more than the chimp GMaxCd–FL_post because of the greater hip flexion excursion in humans.

Because human ITB MTUs undergo relatively larger length changes during bipedal walking, and because the force-generating capacity of human muscles inserting on the ITB is larger relative to the force-generating capacity of chimp muscles inserting on the FL, the human ITB has a greater potential to store energy than the chimp FL (Fig. 8). Activating GMax at 20% during late swing in the human model generates force that stretches the posterior ITB by about 4% and stores 0.0133 J kg⁻¹ of elastic energy per stride (Fig. 8A,C). Activating GMaxCd at 20% during midstance in the chimp model strains the posterior FL by a similar amount, but stores only 0.0009 J kg⁻¹ of elastic energy per stride. Activating TFL at 20% during late stance in the human model stretches the anterior ITB by 2% and stores 0.0018 J kg⁻¹ of elastic energy per stride (Fig. 8B,D). Activating TFL at 20% during late stance in the chimp model stretches the anterior TFL less, storing only 0.0001 J kg⁻¹ of elastic energy per stride. Energy storage increases with muscle activation level (Fig. 8). For example, activating the human GMax and chimp GMaxCd at 60% stores 0.0616 J kg⁻¹ of elastic energy per stride in the human ITB and 0.0038 J kg⁻¹ in the chimp FL (Fig. 8C), which is over four times the energy stored with 20% activation. With 60% activation of the TFL, the human ITB stores 0.0089 J kg⁻¹ and the chimp FL stores 0.0006 J kg⁻¹ (Fig. 8D).

When we maximally activated muscles inserting in the ITB and FL at midstance in our models, we found that the frontal plane moment transmitted via the human ITB is not substantially greater than the moment transmitted via the chimp FL (Fig. 9) after normalizing the moments by body weight and hemi-pelvis width. In both models, the TFL and anterior GMax MTUs have the capacity to generate small moments about the hip that help support the pelvis. By contrast, the MTUs with the greatest force-generating capacity, GMax3,4–ITB_post in the human and GMaxCd–FL_post in the chimp, both generate an opposing moment at the hip that pulls the pelvis inferiorly. If we ignore GMax3,4–ITB_post in the human model, we estimate that the ITB transmits about 10% of the total frontal plane moment-generating capacity of all muscles inserting on the ITB and femur (Fig. 9), which is greater than our estimate of the percentage transmitted by the chimp FL (3%). However, if GMax3,4 is activated with other portions of GMax during walking, the summed contribution of the human ITB MTUs to the frontal plane moment is negligible. In summary, this analysis provides little evidence that the human ITB is specialized to transmit forces in the frontal plane to stabilize the pelvis and support the torso against gravity during walking.

This study tested whether the human ITB is specialized for elastic energy storage relative to the chimp FL. We conducted detailed anatomical experiments on the largest sample of chimp lower extremities to date, and we analyzed musculoskeletal models of both humans and chimps to test four hypotheses.

First, we asked whether the muscles inserting on the human ITB have a greater force-generating capacity than the muscles inserting on the chimp FL, after accounting for body mass (H1). We found that, in total, the force-generating capacity of the muscles inserting on the ITB is three times greater than the force-generating capacity of the muscles inserting on the FL, suggesting substantially greater forces are transmitted via the ITB compared with the FL. This greater capacity for force primarily stems from the fact that only about 10% of the chimp TFL, GMaxCr and GMaxCd mass inserts in the FL, whereas nearly 60% of the human TFL and GMax mass inserts in the ITB.

Second, we hypothesized that the human ITB undergoes greater strains than the chimp FL during typical bipedal walking kinematics (H2). We found that the greater MTU length changes and greater mass-specific force-generating capacity of the human TFL result in greater peak strains in the human anterior ITB than the chimp anterior FL. The anterior ITB in humans stretches more than the anterior FL in chimps because humans walk with greater hip flexion/extension excursion than chimps (O'Neill et al., 2015). Contrary to our hypothesis, peak strains in the posterior ITB and posterior FL are similar in our models. However, consistent with our third hypothesis that the human ITB has a substantially greater potential to store elastic energy, per unit body mass, than the chimp FL during bipedal walking (H3), the larger forces transmitted to the posterior ITB result in substantially greater energy storage. Thus, differences in both anatomy and locomotor mechanics between chimpanzees and humans determine the human ITB's greater elastic energy storage capacity compared with the chimp FL.

We also tested the prevailing hypothesis that the ITB functions to stabilize the pelvis in the frontal plane (Inman, 1947; Kaplan, 1958; Stern, 1972; Gottschalk et al., 1989) by estimating the maximum frontal plane moments that could be transmitted via the human ITB and chimp FL at midstance. These analyses indicate, contrary to our hypothesis (H4), that the human ITB does not have a substantially greater capacity to transmit frontal plane moments about the hip than the chimp FL. The human ITB transmits only about 10% of the total hip frontal plane moment at midstance and its contribution to the total frontal plane moment is negligible when GMax3,4–ITB_post is included. Thus, these results do not suggest that the ITB is primarily specialized for frontal plane stability of the pelvis. Although the human ITB does provide a frontal plane moment, our results indicate that the human ITB was likely selected to store elastic strain energy for improved locomotor economy.

It is important to acknowledge the limitations of this analysis. First, in our chimp model, we scaled the muscle PCSAs by the same specific tension of 31.5 N cm⁻² used in O'Neill et al. (2013), which is higher than the typically reported range of 17–25 N cm⁻² (Bodine et al., 1987; Lucas et al., 1987; Greaser et al., 1988). Scaling the chimpanzee muscle PCSAs by a lower specific tension value would decrease our estimates of FL energy storage and further amplify the difference in energy storage between the chimp FL and human ITB. Because the human model's muscle parameters were based on data from elderly cadaveric specimens whose atrophied muscles likely underestimate muscle PCSAs for healthy human subjects, muscle PCSAs were scaled by a specific tension of 61 N cm⁻² (Arnold et al., 2010). In the model of Arnold and colleagues, this value of specific tension predicts hip, knee and ankle joint moments that are consistent with the moments measured in healthy human subjects. If we use the higher specific tension value for both the human and chimp models, chimp FL energy storage values are doubled, but human ITB energy storage is still substantially greater.

Second, because we lack data describing chimp FL material properties, we assumed its elastic modulus was similar to the human ITB, taking into account the lower FL cross-sectional area when calculating normalized FL stiffness. If the chimp FL has a lower elastic modulus than the human ITB, we may have underestimated the capacity of the chimp FL for energy storage during walking. However, even if we make the chimp FL twice as compliant, peak mass-specific energy storage at 20% muscle activation is still nearly an order of magnitude lower in the chimp FL than in the human ITB.

Third, chimps may require greater hip extensor muscle activation to maintain their typical bent hip and bent knee posture, so the chimp GMaxCd may have greater activation than the human GMax during walking. Higher activations and muscle forces in the chimp GMaxCd would reduce differences in human ITB and chimp FL energy storage. However, even when GMaxCd is activated at 60% in our model, the chimp FL_post stores 70% less mass-specific elastic energy than the human ITB_post with 20% GMax activation.

Fourth, we estimated the force generated by each muscle at joint angles corresponding to bipedal walking and ignored the muscle force–velocity properties. If these muscles do not operate isometrically during walking, then we may have overestimated force and energy storage. However, muscles acting in series with long elastic tissues often operate isometrically over much of force development (e.g. Biewener and Roberts, 2000). Consistent with this, ultrasound-based studies of human gastrocnemius function show limited shortening until push-off (Lichtwark et al., 2007; Farris and Sawicki, 2012), and Arnold et al.'s simulations of human walking and running (2013) showed that the GMax muscle fiber velocities were low during walking. In Arnold et al.'s model, GMax inserts on the femur via a relatively stiff tendon, and this assumption likely amplifies the muscle's estimated shortening velocity. It is plausible that GMax would change length even less in the model if it were more accurately represented as inserting via the ITB.

A final limitation of our analyses is that primates other than chimpanzees may be under greater selective pressure to increase locomotor economy through energy storage and recovery. While the last common ancestor of humans and other apes was probably a knuckle-walker that resembled chimpanzees in some respects, compromises between adaptations for walking and for climbing probably explain why the chimpanzee's cost of locomotion is unusually high among mammals (Taylor and Rowntree, 1973; Taylor et al., 1982; Sockol et al., 2007; Pontzer et al., 2014). To better understand whether evolved changes in FL anatomy are indeed related to minimizing the cost of terrestrial locomotion, studies that examine FL function in additional primates are vital.

Our results show that the human ITB is specialized for elastic energy storage compared with the chimp FL, suggesting that modifications to the FL and surrounding muscles in hominins were selected to increase the economy of bipedal locomotion. The chimpanzee FL is likely to be a good model for the ancestral condition from which the derived human ITB evolved. Although the chimp FL stores substantially less elastic energy than the human ITB during walking, it resembles the human ITB in being relatively thick and fibrous compared with other deep fascia in the lower limb. Both the chimp FL and human ITB receive muscle fibers from GMaxCr and TFL, and both insert anteriorly on the tibia. Assuming this anatomy was also present in the last common ancestor of chimps and humans, it is reasonable to hypothesize that the human ITB was elaborated from a chimp-like FL by increasing the sizes of the GMaxCr and TFL muscles inserting in the ITB and by increasing ITB thickness, thereby augmenting the potential of the ITB for energy storage.

In addition to differences in muscle morphology, differences in bipedal gait patterns influence the energy storage potential of the human ITB and chimp FL. Chimps, like other primates, walk with a crouched posture and limited hip flexion/extension excursion (Schmitt, 1999; Sockol et al., 2007), reducing stretch and energy storage in the chimp FL compared with the human ITB. Since the last common ancestor of humans and apes almost certainly used a crouched gait, the FL of this species was likely to be stretched a similar amount during walking. FL energy storage would have increased in hominins with more upright postures and larger GMax muscles. However, because the FL does not fossilize and no skeletal markers of this trait have been identified, it is not possible to identify when the ITB evolved in hominin evolution beyond inferences drawn from extant taxa.

A number of skeletal features in australopiths and Homo suggest an increased capacity for FL energy storage relative to a chimp-like last common ancestor. Features indicating an extended lower limb posture in hominins include a posteriorly oriented ischium (Robinson, 1972), a large femoral bicondylar angle (valgus knee) (Heiple and Lovejoy, 1971; Stern and Susman, 1983), anteroposteriorly flattened femoral condyles (Heiple and Lovejoy, 1971; Ward, 2002), a more perpendicular orientation of the tibia's distal articular surface relative to the long axis of the tibial shaft (Latimer et al., 1987) and longitudinally oriented trabecular struts in the distal tibia (Barak et al., 2013). Expanded muscle attachment areas on the ilium for GMax and TFL suggest an increase in the force-generating capacity of these muscles.

When did these features first appear in the fossil record? The oldest known hominin species for which we have pelvic material, Ardipithecus ramidus, probably had inferiorly oriented ischia like chimpanzees (Lovejoy et al., 2009), but in Australopithecus afarensis, the posteriorly oriented ischia (Robinson, 1972; Stern and Susman, 1983), the large bicondylar angle (Stern and Susman, 1983) and the perpendicular orientation of the tibia's distal articular surface relative to the tibial shaft (Latimer et al., 1987) suggest a more upright posture than chimpanzees. The distally flattened femoral condyles (Heiple and Lovejoy, 1971) and longitudinally oriented trabecular struts in the distal tibia (Barak et al., 2013) in Australopithecus africanus provide additional evidence that hominins by 2- to 3-million years ago had a human-like extended limb posture. Thus, fossil evidence suggests that australopiths walked with a human-like gait in terms of hip, knee and ankle angles (Latimer et al., 1987; Tardieu and Trinkaus, 1994; Crompton et al., 1998; Ward, 2002; Carey and Crompton, 2005; Barak et al., 2013), likely resulting in increased FL energy storage during walking compared with a chimp-like last common ancestor. However, without increases in the mass of muscles inserting in the FL, which fossil evidence indicates occurred later in the genus Homo, FL energy storage would not have reached human-like magnitudes. Pelvises from Homo erectus indicate that the GMaxCr attachment surface is broadened and human like (Day, 1973; Rose, 1984; Lieberman et al., 2006), suggesting increased GMaxCr mass and force transmission to the FL.

In summary, our synthesis of experimental data from cadaveric specimens and energy calculations from musculoskeletal models has revealed a novel function for the uniquely human ITB and has identified biomechanical features that increase the potential for energy storage in the human ITB relative to the chimp FL. Although the upright posture of australopiths would have increased FL energy storage during bipedal walking to some extent, only in Homo is there evidence of an increase in the size of muscles transmitting force through the FL. The forces transmitted by the human ITB during walking and especially during running, are substantial: we have previously estimated that the ITB stores about 14% as much energy as the Achilles tendon during fast running (up to 7 J; Eng et al., 2015), which suggests that the ITB may be specialized to increase the endurance running capabilities in Homo. In future studies, comparative analyses of a broader range of musculoskeletal models, based on data derived from chimpanzees and other higher primate species, offer much potential to advance our understanding of locomotor shifts in the fossil record.

Detailed measurements of muscle moment arms and anatomy were collected from four fresh-frozen chimpanzee [Pan troglodytes (Blumenbach 1775)] cadaveric pelvises (Table 2) obtained from the Texas Biomedical Research Institute (San Antonio, TX). Moment arms of the portions of TFL, GMaxCr and GMAXCd inserting on the FL were determined for hip flexion/extension, hip rotation, hip adduction/abduction and knee flexion/extension using the tendon excursion method (Brand et al., 1975; An et al., 1984).

In each specimen, skin and subcutaneous fat were dissected from the gluteal region and thigh to expose the muscle origins and insertions. The surface of the FL was cleaned of subcutaneous fat, and its insertion on the tibia was exposed distally. Each muscle was separated into portions based on origins and insertions (Table 3). Muscle portions were each represented by a Kevlar thread path for tendon excursion measurements. We anchored Kevlar thread to a screw eye located at the insertion of the MTU, routed the thread through plastic tubing to a screw eye located at the origin of the MTU, and attached it to one of two cable-extension position transducers (PTX101, Celesco, Canoga Park, CA) that measured length changes with an accuracy of ±0.32 mm while applying a tension of 1.4 or 2.8 N. The tubing ensured a repeatable path and decreased friction between the thread and underlying tissues. The 3D coordinates of each muscle path were digitized relative to segment coordinate systems, using a motion-tracking system (Polhemus Fastrak, Colchester, VT) and were used to accurately recreate the paths in the musculoskeletal model. We placed marker pairs in the anterior and posterior FL and tracked their locations with high-speed video (Photron USA Inc., San Diego, CA) while moving the limb through its ranges of hip and knee motion. These data were used to determine the hip and knee angles at which the anterior and posterior FL began to stretch.

Each specimen was mounted in a custom frame as described in Eng et al. (2015), which allowed independent control of hip rotation, hip flexion/extension, hip abduction/adduction and knee flexion/extension after proper specimen alignment (Fig. 10). Briefly, the pelvis was mounted on a table using threaded fixation half-pins (IMEX Veterinary, Inc., Longview, TX). The femur was fixed to the inner of two concentric rings mounted on a rotating cart. Rotating the inner ring relative to the outer ring rotated the femur. Rotating the cart flexed and extended the hip when the specimen was mounted for hip flexion and abducted and adducted the hip when the specimen was mounted for hip adduction. Vertical translation of the rings on the cart adducted or flexed the hip when mounted for hip flexion or adduction, respectively.

Electromagnetic receivers were rigidly attached to the pelvis, femur and tibia, and were used to track the segment's positions and orientations in space using a motion-tracking system. Joint angles were computed from receiver data using custom LabView software (National Instruments, Austin, TX). Anatomical landmarks on each segment were marked with screws and digitized to define each segment's coordinate system based on anatomical axes. To find the hip joint center, the coordinates of the knee joint center (the midpoint between the medial and lateral epicondyles) were recorded while moving the femur through a range of hip flex/extension and hip abuction/adduction angles. A sphere was fitted to this cloud of knee joint center points, and the center of this sphere was defined as the hip joint center and the origin of the femoral coordinate system.

Specimen alignment was performed with real-time feedback of the segment positions and orientations as described in Eng et al. (2015). After aligning the specimen, we monitored coupling of hip angles and ensured that hip adduction varied <2 deg and hip rotation <4 deg over a 75 deg range of flexion. When the specimen was aligned for hip abduction/adduction, we ensured that hip flexion varied <2 deg and hip rotation <4 deg over a 50 deg range of abduction/adduction.

Excursion and joint angle data were simultaneously sampled at 10 Hz using an A/D converter (National Instruments BNC-2090). While moment arms were measured about one joint axis, the other joints were secured at the approximate joint positions found in midstance during bipedal walking (hip flexion, 35 deg; hip rotation, 0 deg; hip adduction, −15 deg; knee flexion, 45 deg) (O'Neill et al., 2015). Lengthening excursion versus joint angle data were fitted with a fourth-order polynomial and the derivative of the polynomial was averaged across trials to estimate the moment arm. A minimum of five trials was collected for each condition from each muscle.

We measured muscle architecture of the chimp TFL, GMaxCr and GMaxCd to estimate each muscle's force-generating capacity. After measuring moment arms, each muscle was carefully dissected to its insertion on the femur or fascia lata (FL). Blunt dissection was used to identify and separate the muscle fibers inserting on the FL from those inserting in the femur. Insertion on the FL was confirmed by placing traction on the separated fibers and ensuring that force was transmitted distally through the FL and not to the tendon inserting in the femur. The portions of TFL and GMax inserting on the FL and femur were weighed, and these masses were summed to calculate the total mass of each muscle. Muscle masses were normalized by body mass and these data were compared with the relative masses of the TFL and GMax inserting on the human ITB reported in Eng et al. (2015).

We modified paths of the TFL–FL and GMax–FL MTUs in the musculoskeletal model reported by O'Neill et al. (2013) to match our digitized muscle attachments, regional paths and moment arm data (Fig. 11). Using SIMM, we created two paths for TFL, one path for GMaxCr, and two paths for GMaxCd. MTUs were represented as line segments spanning from origin to insertion and were constrained by ‘via’ points (points through which a muscle is constrained to act) and wrap objects to simulate underlying structures and more accurately estimate changes in length with changes in joint angle (supplementary material Fig. S1). Via points and wrapping surfaces were iteratively adjusted so that the paths resembled the paths digitized during the experiments and the model's moment arms matched the moment arms determined experimentally. The experimentally measured moment arms were used to adjust the model's paths and verify their accuracy. Our refined model is available on SimTK (simtk.org).

Although multiple muscle paths were created and analyzed to gain insight into FL–MTU 3D anatomy, paths were combined to represent the three major paths of chimp FL force transmission that we observed in our experiments, including the two anterior FL paths (GMaxCr–FL_ant and TFL1 and TFL2 combined as TFL–FL_ant) and the posterior FL path (GMaxCd3,4; GMaxCd–FL_post). We iteratively made small adjustments to the via points and wrapping objects of the anterior and posterior FL paths to yield combined MTU paths with average moment arms.

We used a Hill-type muscle model (Zajac, 1989; Delp et al., 1990) to estimate isometric forces generated by TFL–FL_ant, GMaxCr–FL_ant and GMaxCd–FL_post. Active and passive force–length curves were scaled to each FL MTU path using two key parameters, maximum isometric force (F_max) and optimal fiber length (L_opt). We used measured fascicle lengths for L_opt, and we calculated F_max as the product of each muscle region PCSA and the muscle-specific tension of 31.5 N cm⁻² used in O'Neill et al. (2013). F_max and tendon slack length (L_TS) were used to scale each MTU's ‘tendon’ force–length curve. Tendon slack lengths were chosen for each MTU such that the FL began to stretch passively at hip and knee angles consistent with our experimental measurements. We verified that these L_TS values allow the muscles in our model to generate force over functional ranges of motion.

We used our model along with joint kinematics and EMG activations to estimate the ability of the FL to store and recover elastic energy during bipedal walking. First, we calculated the origin-to-insertion lengths of the MTUs at hip and knee angles typical of chimps during bipedal walking (O'Neill et al., 2015). We identified periods of the stride when MTUs were near maximum length and also likely active (Stern and Susman, 1981) and we assumed that peak strains in FL_ant or FL_post would occur at these times. Next, we separated the MTU lengths into FL lengths and muscle fiber lengths by independently activating each MTU in the model and solving for the lengths at which the muscle force and FL force were equivalent, accounting for pennation angle. We set each muscle's activation to 20%, 40% and 60% (of its maximum activation) to assess FL strains during walking. We estimated energy storage capacity at each activation level by integrating the FL and ITB force–length curves from L_TS to the peak FL or ITB length during walking. Peak energy storage in the anterior and posterior chimp FL was normalized by the chimp model's body mass of 55 kg and compared with similar estimates of mass-specific peak energy storage in the anterior and posterior human ITB (normalized by the human model's body mass of 83 kg) derived from the data reported in Eng et al. (2015).

To assess whether the human ITB is specialized to stabilize the pelvis in the frontal plane compared with the chimp FL, we examined the moment-generating capacity of the FL and ITB MTUs about an anterior–posterior axis through the hip joint center when the muscles are fully activated at the joint positions found in midstance. These frontal plane moments were compared with the total frontal plane moment generated by all of the other hip abductors when fully activated. Moments were normalized by body weight and hemi-pelvis width (the distance between the hip joint center and midline of the pelvis).

The authors dedicate this manuscript to Farish A. Jenkins Jr (1940–2012) who provided many stimulating and insightful discussions during C.M.E.'s dissertation research, which provided the basis for this study. The authors thank two anonymous reviewers for constructive comments that significantly improved this manuscript. The authors also gratefully acknowledge Delande Justinvil and Zachary Lewis for technical assistance during the moment arm experiments. We are grateful for the assistance of Casey Boyle and Yasmin Rawlins during pilot studies, and we thank Dr Andrew Mountcastle and Glenna Clifton for help with videography. We thank Dr Thomas Roberts for his helpful comments on this manuscript.