Development of a comparative chimpanzee musculoskeletal glenohumeral model: implications for human function

Studies of evolution and biomechanics typically fall into three categories – comparative, experimental and modeling (Pontzer et al., 2009). For evolutionary science, comparative morphometric assessment persistently emerges as a primary historical method to quantify the physical abilities and locomotion of human relatives and ancestors. This often involves comparisons of single skeletal features, or a series of skeletal features from fossils for association of form with function with extant hominids such as humans and the great apes (Young, 2008). Experimental studies complement morphometric analyses, typically by quantifying and comparing locomotor and evolutionarily relevant tasks between different primate species (Bertram and Chang, 2001; Demes and Carlson, 2009; Larson, 1988; Larson and Stern, 2013; Stern and Larson, 2001). Experimental research quantifies differences between species in locomotor behavior, and provides clues as to probable adaptions following divergence from a common ancestor. However, both methods include problematic aspects. Although bone shape has been linked to function (Oxnard, 1969), the individual plasticity of skeletal features and the effect of external stimuli in altering morphological features reduce the correlation (Collard and Wood, 2000; Young, 2005). Comparative experimental work, while highly valuable, has been limited by subject availability and compliance, and procedural modifications for non-human subjects may reduce data precision and generalizability (Stevens and Carlson, 2008). Computational biomechanical models offer an alternative to morphological and experimental approaches by incorporating properties of musculoskeletal function and motor control dynamics using information obtained from musculoskeletal structure to simulate functionality (Hutchinson, 2012).

The modern human shoulder is primarily adapted for non-locomotor, below shoulder-height behaviors, despite a possible arboreal ancestry (Arias-Martorell, 2019; Larson, 2007; Lewis et al., 2001; Oxnard, 1969; Thorpe et al., 2007; Veeger and van der Helm, 2007; Young et al., 2015). The subacromial space is the area between the humeral head and the acromion of the scapula, through which the supraspinatus tendon passes (Bey et al., 2007). This space is narrow, as a result of a laterally projecting acromion that often slopes inferiorly (Voisin et al., 2014). But as the human arm elevates, the width of the subacromial space decreases further, reducing the space the supraspinatus tendon occupies. This increases the risk for impingement of the supraspinatus tendon and initiation of rotator cuff pathology (Bey et al., 2007; Graichen et al., 2001). The high musculoskeletal demands placed on the human shoulder during overhead tasks also make this a difficult posture to maintain without fatigue and fatigue-related disorders, particularly of the rotator cuff (Dickerson et al., 2015; Ebaugh et al., 2006a,b; Grieve and Dickerson, 2008; Rashedi et al., 2014). The human rotator cuff muscles fatigue rapidly in overhead postures, affecting muscular coordination at the glenohumeral joint and causing scapular and humeral dyskinesis (Chopp et al., 2010; Cote et al., 2009; Teyhen et al., 2008). Overhead postures become even more problematic as workload increases or the posture is sustained for longer periods (Ebaugh et al., 2006b). Humans who engage in climbing for sport or recreation experience an extremely high rate of upper extremity injury (Folkl, 2013; Nelson et al., 2017), with reports of rotator cuff tendonitis and impingement in climbing populations as high as 33% (Rooks, 1997). Conversely, closely related primates, like chimpanzees, regularly assume and maintain high force overhead climbing and suspensory postures without developing shoulder pathology (Potau et al., 2007; Stern and Larson, 2001). Despite humans having a likely arboreal common ancestor with chimpanzees (Kivell and Schmitt, 2009; Lovejoy et al., 2009; Thorpe et al., 2007, 2014), the human shoulder appears to have musculoskeletally devolved a capacity for overhead activities (Ebaugh et al., 2006a; Lewis et al., 2001; Oxnard, 1969; Punnett et al., 2000). To date, there have been few anthropological computational models of the upper extremity (Regnault and Pierce, 2018), and no upper extremity musculoskeletal model analyzing the human evolutionary path and evolutionary holdovers defining modern human musculoskeletal shoulder capacity.

As chimpanzees are the closest genetic living relative to humans, a chimpanzee shoulder model that parallels a human shoulder model (Dickerson et al., 2007) could provide novel insights into the history of human arborealism and its relationship to the form of the present human shoulder. Chimpanzees and humans share a similar shoulder structure and function (Young et al., 2015). Though specific differences exist which help to delineate the two, both chimpanzees and humans have shoulder bone shape and musculature that defines the great ape morphotype as distinct from other primates (Larson, 1998; Swindler and Wood, 1973; Young, 2003, 2008). Resultantly, the two species have a large amount of functional overlap at the shoulder. Chimpanzees possess a hybrid upper extremity that enables both arboreal and terrestrial quadrupedal locomotion, some suspensory brachiation and bipedalism, and non-locomotor behaviors (Cartmill and Smith, 2009). In contrast, humans have an upper extremity that has devolved any locomotor utility in favor of primarily non-locomotor, non-weight bearing, below shoulder height behaviors such as carrying, reaching, tool making and use, and throwing (Cartmill and Smith, 2009; Lewis et al., 2001; Veeger and van der Helm, 2007). While humans can perform locomotor behaviors such as climbing, and may have ancestral ties to them, the modern efficiency, comfort and sustainability is limited (Folkl, 2013; Nelson et al., 2017). As the chimpanzee represents a similar musculoskeletal system to humans that shares general functional shoulder ability, but with a greater ancestral arboreal capacity, it represents a useful comparative model. Comparative human and chimpanzee musculoskeletal shoulder models will aid in determining what morphological features functionally distinguish the species. These models can also provide insights into the evolutionary form and function relationship of the modern human shoulder.

The purpose of this study was to develop a novel model of the chimpanzee shoulder that parallels the human Shoulder Loading and Assessment Modules (SLAM) model created by Dickerson and colleagues (2007). The model was evaluated using electromyographical data on chimpanzees from the Stony Brook Primate Locomotion Laboratory (Larson and Stern, 1986; Larson et al., 1991; S. Larson, Stony Brook University, unpublished data). The muscular and subacromial space width outputs of this model were compared with those from the existing human shoulder model in an attempt to better delineate those musculoskeletal features that inhibit human performance of overhead behaviors. It was hypothesized that the human model would predict higher muscle forces across all muscles as a percentage of maximal force-producing capability determined by muscle cross-sectional area, but particularly in the rotator cuff muscles, as well as a narrower subacromial space, compared with the chimpanzee model predictions. This would be due to humans having a lower muscle mass and PCSA, and a more laterally projecting acromion, respectively.

The chimpanzee model is designed to perform comparator analyses with parallel models of other species, such as humans. Outputted predictions from the chimpanzee model are not intended to produce standalone musculoskeletal predictions of chimpanzee shoulder function. The model provides insight into the differences in musculoskeletal function that can manifest between two species models in an analogous computational platform based on differences in geometric musculoskeletal form. Geometric musculoskeletal differences that can be examined include muscle moment arms, muscle force predictions, total muscle force, joint forces and moments, and glenohumeral joint stability.

This study required assembly of a novel chimpanzee glenohumeral musculoskeletal model of the right arm, using the template of an extant human SLAM model (Fig. 1) (Dickerson et al., 2007). Like the SLAM model, the chimpanzee glenohumeral model is composed of three inter-connected modules: (1) a musculoskeletal geometry module; (2) an external dynamic moment module; and (3) an internal muscle force prediction module. The primary inputs for these modules include average species anthropometric data, motion data and task-specific data. Motion data are kinematic motion capture marker positions, while task-specific data are the task-specific hand forces. The outputs of the geometry and moment modules produce the necessary inputs for the force prediction module, which uses an optimization routine to solve for muscle forces. Most differences between the two species-specific models were implemented in the geometry module, but all three modules were modified to represent the chimpanzee.

The task analyzed in this study was a single overhead horizontal bimanual arm suspension cycle. This task was chosen as it is considered to be a common ancestral behavior in the two species, and it is completed with very different levels of capability in modern chimpanzees and humans. Chimpanzees still habitually climb and suspend, while humans appear to no longer have a weight-bearing upper extremity suitable for locomotive purposes (Wood and Richmond, 2000).

The model was used to assess six different instances within the horizontal bimanual arm suspension cycle. (1) Early right support – double support phase. (2) Mid right support – left swing, single support phase. (3) Late right support – double support phase. (4) Early right swing – down phase of arm swing. (5) Mid swing – beginning of reach phase of arm swing. (6) Late swing – pre-contact with support rung.

These six instances represent distinct time points of the suspension cycle that require different levels of muscular support to move and stabilize the body and upper extremity (Larson and Stern, 1986; Larson et al., 1991). These six instances, along with the task-specific hand forces applied at each instance, can be seen in Fig. 2.

The original SLAM geometry module uses human representative bone scan data and postural motion coordinate data as inputs to determine boney orientations and positions of each segment, and subsequent lines of action and moment arms for each muscle element (Dickerson et al., 2007). The chimpanzee module paralleled this structure. When original chimpanzee musculoskeletal data were available, they were used in the geometric module. If original quantitative chimpanzee data were not available, musculoskeletal geometry was mathematically and iteratively fitted to the model to provide an appropriate representation of chimpanzee musculoskeletal geometry, as has been done in previous models (O'Neill et al., 2013). There are five different parts of the musculoskeletal geometry module.

The first part of the geometric module is a segment parameter definition. The model has five segments – the torso, and the right-side clavicle, scapula, humerus and forearm. The glenohumeral joint is modeled as a spherical joint with three degrees of rotational freedom. As the utility of the model is for shoulder and glenohumeral analysis, the forearm is visually modeled as a simplified, single radial/ulnar link with the elbow having one degree of freedom (flexion/extension) (Fig. 3). Pronation and supination of the forearm were still accounted for in the model in the external moment module (see ‘External dynamic moment module’, below) to calculate the three-dimensional elbow angles, forces and moments, which were used to drive the final internal force prediction module. The dimensions of each of the segments for the constructed chimpanzee model were determined from existing data on average bone dimensions in the chimpanzee upper extremity (Larson, 1998; Schoonaert et al., 2007; Thorpe et al., 1999; Young, 2003).

The second part of the geometric module is an algorithm for shoulder rhythm. Shoulder rhythm is the closed-chain kinematic interaction between the bones and joints of the shoulder (Inman et al., 1944). Measuring this kinematic interaction can be difficult, as movement of the scapula and clavicle is particularly hard to acquire with skin surface marker motion capture methods, the most commonly used approach for quantifying three-dimensional kinematics (Karduna et al., 2001; van Andel et al., 2009). Assessment of shoulder rhythm is also difficult because of individual variation. Measured shoulder rhythm has been shown to be highly variable across individuals and populations (Grewal and Dickerson, 2013; Grewal et al., 2017; Ludewig and Cook, 2000; Ludewig et al., 2009; McClure et al., 2001; Tsai et al., 2003). However, several invasive studies have demonstrated a predictable kinematic pattern in scapular and clavicular three-dimensional motion with respect to the more easily acquired humeral and thoracic motion (Högfors et al., 1991; Karduna et al., 2001; van Andel et al., 2009). Therefore, mathematical equations are used to predict three-dimensional scapular and clavicular rotations from the measured three-dimensional kinematics of thoracohumeral rotations. Mathematically predicted shoulder rhythm is characterized by a total of six equations representing three clavicular rotations and three scapular rotations. Each shoulder rhythm equation contains mean value coefficients for the population and they do not change between individuals. Thus, two individuals with the same three-dimensional thoracohumeral orientation will have the same equation-predicted scapular and clavicular three-dimensional orientations, though individual variation may result in different true scapular and clavicular orientations. A number of human shoulder rhythm equations have been published to estimate clavicular and scapular kinematics from thoracohumeral skin surface motion capture. Those produced by Grewal and Dickerson (2013) were used in the human model, as they account for a large range of humeral elevation, much like that assumed during bimanual horizontal climbing (Grewal and Dickerson, 2013). No shoulder rhythm equations currently exist for chimpanzees.

Eqns 1–6 predict the chimpanzee scapular retraction/protraction (γ_S), scapular lateral/medial rotation (β_S), scapular anterior/posterior tilt (α_S), clavicular elevation/depression (γ_C), clavicular retraction/protraction (β_C), and clavicular forward/backward rotation (α_C). The three thoracohumeral rotations used to predict the scapular and clavicular rotations are represented as plane of elevation (γ_TH0), elevation (β_TH), and internal/external rotation (γ_TH1).

The third and fourth parts of the geometry module are the definition of right-side muscle elements and their corresponding lines of action (Fig. 3). Fourteen separate upper extremity muscles were modeled. In both the chimpanzee and human geometry module, five of these muscles – biceps, triceps, infraspinatus, supraspinatus and deltoids – were modeled with multiple mechanical elements to represent their multiple attachments for a total of 20 muscle elements. Chimpanzees share most of the same muscular anatomy with humans. The exception is that chimpanzees have an additional muscle, the dorsoepitrochlearis (Ashton and Oxnard, 1963; Diogo et al., 2013; Swindler and Wood, 1973). The dorsoepitrochlearis muscle typically arises from the latissimus dorsi or coracoid process and attaches on the distal humerus (Ashton and Oxnard, 1963). Muscle attachment sites were determined using published data (Ashton and Oxnard, 1963; Ashton et al., 1976; Carlson, 2006; Diogo et al., 2013; Swindler and Wood, 1973; Thorpe et al., 1999). Precise three-dimensional locations were not available for chimpanzee muscle origins and insertions. Estimations were made iteratively following muscle footprints provided in literature sources (Swindler and Wood, 1973). Muscle lines of action were defined as strings, partially using spherical and cylindrical geometric muscle wrapping techniques around orthopedic surfaces that generate more physiologically representative lines of action and paths about the glenohumeral joint in chimpanzees (Fig. 3) (Dickerson et al., 2007; van der Helm, 1994). While modeling muscles as strings is an oversimplification of muscle physiology and architecture, ‘overfitted’ muscle models can result in erroneous muscle force predictions and generally have limited predictive capabilities (Buchanan et al., 2004; Scott and Winter, 1991). The fifth and final part of the module created contact force application sites between the scapula and ribcage, as well as ligament placements. Ligaments were not included as contributing elements in either model.

The postural motion data inputs for the chimpanzee model geometric module were derived from human motion capture files. Quantitative three-dimensional data, sufficient as inputs to the geometry module, do not currently exist of chimpanzees performing brachiation tasks, including horizontal bimanual arm suspension (Demes and Carlson, 2009; Reghem et al., 2013; Stern and Larson, 2001; Usherwood et al., 2003). In the absence of three-dimensional chimpanzee kinematics, human arm suspension kinematics were systematically modified to anthropometrically represent chimpanzee segment lengths and joint positions (MacLean and Dickerson, 2019). Appropriate chimpanzee arm lengths were achieved by translating the wrist and elbow joint centers, each of which were defined by human skin surface marker positions, to modify the forearm and upper arm length, respectively, to those reported in the literature (Schoonaert et al., 2007). As chimpanzees also have a more superiorly and medially positioned scapula, it was also necessary to shift the human acromion marker both superiorly and medially by a standardized distance based on the chimpanzee x-rays (Thompson et al., 2020). As the glenohumeral joint position is estimated from the position of the acromion marker, this automatically shifted the chimpanzee glenohumeral joint. Once joint centers were translated, these data were used as a surrogate for chimpanzee arm suspension kinematics. For the purpose of this analysis, the kinematic data of a single experienced climber were used for the simulation. Outputs of an average single subject delineate mean differences between species, and provide an initial indication of the realism of the comparative models that could be washed out through population means. Multi-subject analyses can be performed within each model, with appropriate subject kinematic, anthropometric and task data to further explore the influence of performance variation.

The external dynamic moment module uses motion capture data to derive external forces and moments using inverse dynamics (Dickerson et al., 2007). The moment module was driven by the same modified human motion capture data inputs used to drive the geometric module. An inverse dynamics approach was used to calculate joint forces and moments (Vaughan et al., 1992).

The moment module has four steps. The first was the description of segment properties. Segment properties of the chimpanzee upper extremity (upper arm, forearm and hand) were determined using anthropometric data on segment mass, length and moments of inertia (Thorpe et al., 1999; Schoonaert et al., 2007; Zihlman, 1992). The modified human motion data were used to estimate the center of rotation for the glenohumeral, elbow and wrist joints, which were also used to determine locations of the segmental centers of mass (Dickerson et al., 2007). Local coordinate systems for each segment were then defined.

The second and third parts of the module are the calculation of linear and angular kinematics. These were determined from differentiating filtered motion data and the Euler angle decomposition method employed by Dickerson et al. (2007). For the purposes of the current static analysis, this step was not utilized. However, it is present in the module for the possibility of dynamic analyses.

The fourth step is the calculation of external joint forces and net moments. Forces and moments were calculated using Newtonian laws of motion (Vaughan et al., 1992). Gravity was the only external force applied, with the reaction force acting at the hand. In the swing phase of the suspension cycle, this translated to an external force equivalent to the gravitational force produced by the masses of the upper body segments. In the mid-support phase of the suspension cycle, when only the right hand provided support, the external force acting at the hand was assumed to be equivalent to total body mass multiplied by gravity. At early and late support, the external force was assumed to be equally shared by the two limbs and was half of body mass multiplied by gravity, and applied at the right hand.

The outputs of both the geometry and moment modules provide inputs to the muscle force prediction module (Fig. 1). The high number of muscles that contribute to glenohumeral motion constitute an indeterminate system, with more muscles than mechanical equations to define the system (Dickerson et al., 2007). Thus, an optimization approach was used to generate muscle force predictions using muscular and mechanical constraints. The optimization routine consists of five interconnected parts that delimit potential force prediction solutions, enhancing physiological feasibility.

First, a series of mechanical constraints were defined for the three-dimensional angular and linear equilibrium of the glenohumeral joint, composed of muscle forces, joint contact forces and external forces. An additional mechanical constraint was enforced for elbow joint flexion/extension moment equilibrium.

Second, muscle force bounds were defined. The lower bound for all muscles was 0, while the upper bound was proportional to the absolute physiological cross-sectional area (PCSA) of each chimpanzee muscle, based on published data (Table 1) (Carlson, 2006; Kikuchi, 2010; Michilsens et al., 2009; Mathewson et al., 2014; Oishi et al., 2009; Thorpe et al., 1999; Ward et al., 2006), multiplied by a specific tension value. As no data exist for baseline muscle tension in chimpanzees, the previously used specific tension for humans of 88 N cm⁻² was applied to determine muscle force upper bounds (Wood et al., 1989). Chimpanzee subscapularis and infraspinatus PCSA data were only provided as whole muscle, and not the three and two respective mechanical elements of each muscle. To determine appropriate mechanical muscle element PCSAs for subscapularis and infraspinatus, the percentage breakdown of PCSA for the elements of these muscles in humans was used to assume a PCSA of the mechanical muscle elements for chimpanzees.

Third, another constraint, glenohumeral contact force, was applied. Derived glenohumeral stability force ratios were implemented to determine force thresholds in eight directions perpendicular to the surface of the glenoid (Lippitt et al., 1993). As glenohumeral stability force ratios were unknown in chimpanzees, they were estimated through known structural differences between species in glenoid shape and depth (Larson, 1998; Lippitt et al., 1993; Macias and Churchill, 2015; Matsen et al., 1994; Young, 2003). The human glenoid fossa is approximately 4.8 mm deep, inclusive of the labrum, whereas the chimpanzee fossa was determined to be approximately 6 mm deep. According to Matsen and colleagues (1994), stability ratios increase 10.9% for every 1 mm increase in depth. As the chimpanzee glenoid fossa is of a similar shape to that of humans, it was assumed to have a proportionally similar increase of 13.13% in stability force ratios in all eight directions as a result of increased depth.

Fifth, the solution methodology was defined. The optimization routine has a standardized scheme made up of the previous parts of the prediction module. The methodology solves for the indeterminacy of the mechanical system in a sequential manner. Each solution is used to inform the next sequential solution.

The chimpanzee glenohumeral model was evaluated using a concordance analysis to compare computational model chimpanzee muscle force predictions with experimentally collected chimpanzee electromyographical (EMG) data for a subset of muscles included in the model. Evaluation of the chimpanzee model presented challenges unfamiliar to human modeling efforts. As novel experimental data on chimpanzees cannot be readily acquired because of new legislation and the lack of experimental facilities, model evaluation was limited to comparisons with previously collected EMG data on chimpanzee muscle activity. The tissue loading predicted by the chimpanzee model was assessed through comparison with published and unpublished experimentally acquired EMG data (Larson, 1988; Larson et al., 1991; Larson and Stern, 1992, 2013) using concordance analysis (Dickerson et al., 2008). This analysis determines timing concordance in muscle activity and inactivity between EMG and predicted model muscle forces. If both the EMG and predicted muscle forces predict muscle activity above defined thresholds, there is concordance. If one indicates activity and the other does not, there is discordance (Dickerson et al., 2008). A concordance analysis is appropriate in this scenario, as instantaneous relative EMG amplitudes are highly variable with postures and movements, and normalization methods, and typically show weak relationships with predicted muscle forces (Makhsous, 1999; van der Helm, 1994). As little data exist that include EMG of chimpanzees brachiating, this method also prevents biased evaluation of the predicted muscle forces via a limited EMG dataset. The concordance analysis was used to assess each of the six simulated instances of the suspension cycle, representing six static points of a full horizontal bimanual arm suspension cycle of the right arm.

Studies that have analyzed muscle activity in primates have not conducted maximal voluntary contractions to normalize EMG produced during activity, as it is not logistically possible. EMG from primates is often normalized to the maximal EMG signal produced during the task of interest (Larson et al., 1991; Usherwood et al., 2003). To determine the ‘active’ or ‘inactive’ state of a muscle, a predicted model muscle force was considered ‘active’ if it was greater than 5% of its maximal force producing capacity (Dickerson et al., 2008). Because of possible noise and spurious predictions, chimpanzee EMG signal was considered ‘active’ if it was above approximately 5% of the maximal produced signal. Only select muscles were included in the concordance analysis, owing to availability of experimental data. These included published data on anterior deltoid, middle deltoid, posterior deltoid, supraspinatus, infraspinatus, subscapularis, teres minor, triceps brachii, teres major and coracobrachialis (Larson, 1988; Larson et al., 1991; Larson and Stern, 1992, 2013). Unpublished data were also retrieved from University of Stony Brook, New York, USA, courtesy of the Department of Anatomy and used in the concordance analysis. These muscles included triceps brachii, coracobrachialis and middle deltoid. Muscles that were modeled as multiple muscle elements were combined for the concordance analysis, as they were experimentally analyzed as a single muscle element.

Once the chimpanzee glenohumeral model development was complete, the novel model was run for evaluation analysis and for comparative analysis with the human SLAM model. Anthropometric, hand force and kinematic postural inputs for both models were applied for model operation, and selected muscle force and subacromial space width outputs were compared between species.

Anthropometrics were used to approximate segment parameters for determining joint forces and moments in the external dynamic torque module. Representative average healthy human (mass: 72 kg; height: 1.8 m) and chimpanzee (mass: 45 kg; height: 1.32 m) males were used as the criteria subjects within each species-specific glenohumeral model.

Applied hand forces depended on the suspension phase, and were used in the external dynamic torque module to predict joint forces and moments. As the model was run statically, gravity was assumed to be the only external force, and acting at the hand.

The motion data of a single experienced male climber were used as the static postural input for both the human and chimpanzee glenohumeral models. This motion data were modified to be more representative of the chimpanzee shoulder structure for the chimpanzee model.

The two models used the same human kinematic inputs, but with systematic joint position and segment length modifications in the chimpanzee model. Joint orientation was not altered. Therefore, the chimp model and the SLAM model used the same overall static upper body postures as inputs (Fig. 4). The arm is most horizontally adducted and extended forward in late swing and early support. In late support, the arm is horizontally abducted and most elevated, and resultantly also positioned closest to the torso. Arm elevation decreases with the beginning of swing phase as the hand is released from the support rung and begins to horizontally adduct (Fig. 2).

Chimpanzee model outputs from the suspension task were compared with those produced by the human SLAM model while conducting the same functional task of horizontal bimanual arm suspension. The comparison between species was made for the six instances of a single right arm suspension cycle. The human SLAM was executed using experimentally measured bimanual suspension kinematics (MacLean and Dickerson, 2019), subject anthropometrics and estimated external hand forces to determine resultant human glenohumeral muscle forces and subacromial space. The chimpanzee model was subsequently executed using the geometrically modified human kinematics to determine subsequent chimpanzee glenohumeral muscle force and subacromial space.

Output-dependent variables compared between human and chimpanzee models included individual muscle force, average normalized muscle force and subacromial space. Individual muscle forces compared between humans and chimpanzees included the rotator cuff (supraspinatus, infraspinatus, subscapularis, teres minor), anterior deltoid, middle deltoid, posterior deltoid, teres major, biceps brachii, triceps brachii, coracobrachialis, brachialis and brachioradialis. Each individual muscle's maximal muscle force was determined by dividing the muscle force prediction in Newtons by each muscle's PCSA and specific tension. Muscle forces were normalized to a maximal muscle force, and presented as a percentage of the maximal muscle force, for each muscle for both the human and chimpanzee model. As it is only extant in chimpanzees, analysis of the dorsoepitrochlearis muscle extended only to examination of muscle force sharing predictions in the chimpanzee model. An average normalized muscle force was also reported. All 19 or 20 normalized muscle forces from the human and chimpanzee model, respectively, were also summed and divided by the total number of muscle elements observed in each phasic instance of the suspension cycle to give an indication of the average normalized requirement from the glenohumeral musculature as a percentage, for each species. Subacromial space was the distance between the inferior portion of the acromion and the most superior point of the humeral head.

For this initial analysis, each model simulation was run once to produce single values for each of the dependent variables for an average human and chimpanzee. This precluded the use of typical statistical analyses for determining significant differences between species in shoulder biomechanics. Differences between species are thus presented as an observation of differences between average chimpanzees and humans in shoulder function and physical capability.

To evaluate the chimpanzee glenohumeral model, muscle force predictions were compared with chimpanzee experimental electromyographical muscle activations while performing the same task – horizontal bimanual arm suspension across all six cycle instances. When both model and experimental data showed activity, concordance was indicated. A total of 12 muscles at 6 discrete time points were used to determine concordance between predicted and observed muscle activity, for a total of 72 data points. Concordance occurred – both model and EMG predicting on or off – in 46 of 72 data points, or 63.8%.

Predicted muscle forces were very different between species. The human infraspinatus lower muscle element was recruited to maximal force and the teres minor to nearly maximal force in early support (Fig. 5A,D). The human model predicted no supraspinatus and a very low late support phase subscapularis contribution in humans (Fig. 5B,C). Chimpanzees were predicted to have more evenly dispersed rotator cuff forces. Humans were predicted to have greater muscle force contributions from all portions of the deltoid than chimpanzees in the support phase (Fig. 5E). The anterior deltoid was predicted to not contribute to the suspension task in either support or swing phase in the chimpanzee model. The teres major, coracobrachialis and dorsoepitrochlearis were limited contributors to completing the suspension task. Only in the human simulation was the coracobrachialis predicted to contribute to the glenohumeral joint equilibrium in early swing and minimally in late swing (Fig. 5K). The triceps long head and lateral heads were active in chimpanzees during mid-support (Fig. 5H), whereas the triceps long head was a large contributor to the glenohumeral and elbow joint force for support phase in the human model (Fig. 5H). The biceps were more active in the human model, contributing a very high percentage of maximal force in early and mid-swing (Fig. 5G). At the elbow, along with the biceps, humans were predicted to rely mostly on the brachioradialis, with muscle force contributions as high as 70% of maximum force production capability (Fig. 5I). Chimpanzees utilized the brachialis more than humans in mid-support.

Average normalized muscle force was much higher in the human model than in the chimpanzee model during the support phase (Table 2). This difference is less magnified when muscle forces were normalized to body mass, but still continued in early and mid-swing (Table 2). Average normalized muscle force was predicted to be more than 3 times as great in humans in early support. The average normalized force in the swing phase was small for both species.

Chimpanzees had a considerably wider subacromial space than humans in all six phasic instances of the suspension cycle (Fig. 6). Differences between the two species were approximately 2 mm. The chimpanzee subacromial space was narrowest in late support when the arm was the most elevated.

This study developed a novel chimpanzee glenohumeral joint model for use in comparative musculoskeletal analyses. Comparisons with a parallel human glenohumeral joint model contrasted the influence of muscular and geometric differences between the humans and chimpanzees. The human glenohumeral model predicted higher muscle forces as a percentage of maximum force-producing capability for most muscles and a narrower subacromial space, mostly supporting the research hypotheses. These directional differences indicate musculoskeletal divergence that may associate functional differences with the evolutionary foundation of modern human rotator cuff function and pathology.

Evaluation of the novel chimpanzee glenohumeral model through concordance analysis provided evidence of the usefulness of the model. The results of the heuristic concordance analysis demonstrated an agreement between the timing of chimpanzee model predictions of muscle activity and experimental electromyographical measures of muscle activity for the same horizontal bimanual arm suspension task. With moderate agreement between measured and predicted muscle activity, the model can be considered sufficiently biologically realistic (Dickerson, 2005). This step provided the necessary evidence to allow plausible comparative analyses between models.

The concordance analysis did not have complete agreement between model predicted and measured muscle activity timing but the concordance value obtained was moderate. EMG is a very sensitive measurement technique (Basmajian and De Luca, 1985; De Luca, 1997). Thus, the present concordance analysis concentrated on muscle activity timing, as concordance analyses that include muscle amplitudes would be susceptible to EMG variability and Type II error (De Luca, 1997; Miller, 2006). A concordance value of 0.638 is considered satisfactory for an analysis of biological modeling (Dickerson, 2005). Realism in biological modeling is difficult as it requires the consideration of a variety of biological variables dependent on numerous parameters (Dickerson, 2005; Garner and Pandy, 2001).

Differences between the two models mostly represent biological differences in the modeled musculoskeletal systems of the geometric module. The lack of complete biological realism in the chimpanzee model – including some synergistic and antagonistic muscle action – falls within the range of similar biomechanical models (Cholewicki et al., 1995; Dickerson, 2005). The chimpanzee glenohumeral model developed in the present study was and is intended only for muscular comparison with a parallel representation of the extant human, and future evolutionary models. The model's ability to distinguish different musculoskeletal strategies – particularly muscle force sharing strategies – from other models is its primary objective. As mathematical computational representations of biology, both the human and chimpanzee models represent a simplification of the musculoskeletal system, including a shoulder rhythm to predict scapular and clavicular orientations, muscles modeled as strings, no ligamentous contribution, and an optimization routine for predicting muscle forces. Assumptions and limitations present in the chimpanzee glenohumeral model are structurally mirrored in the comparator model.

The results of the present evaluation demonstrate that the chimpanzee model has value as a comparative model, providing insight into shoulder function and evolution. Compromises and assumptions were essential to develop efficient, purposeful models, but predictions should be viewed in their context. The design of the model limits the generalizability of the present results beyond the above-stated usage and assumptions. However, for performing comparative analyses, these models reflect and highlight real musculoskeletal differences between each species in modeled geometry.

Chimpanzees were predicted to execute the suspension task using an overall lower percentage of their muscular capacity than humans at all discrete stages of the suspension cycle, at both the elbow and glenohumeral joint. This was expected, as chimpanzees have a greater proportional muscle mass in their upper extremity (Walker, 2009). Humans and chimpanzees have approximately 9% and 16% of their total body mass relegated to their upper extremity, respectively (Zihlman, 1992). Despite having a lower average body mass by as much as 25 kg, the chimpanzee upper extremity muscle masses can be upward of twice that of analogous human muscles (Carlson, 2006; Mathewson et al., 2014; Thorpe et al., 1999; Walker, 2009). This translates into individual chimpanzee upper extremity muscles requiring a smaller percentage of their maximal muscle exertion to execute a task with the same posture and applied external force as humans.

Previous in vitro research on chimpanzees has shown them to have a greater absolute PCSA than humans for all muscles in the present glenohumeral model. Even when scaled to the same body mass, chimpanzees still have greater relative PCSA across all muscles (Table 1) (Thorpe et al., 1999). There are notable muscles with a PCSA that are much greater than the human muscle PCSA. These included the coracobrachialis, middle deltoid, subscapularis and supraspinatus. Increased PCSA increases the force production capabilities of a muscle by increasing muscle fiber content (Nigg and Herzog, 2007). That shoulder muscles in chimpanzees have a PCSA greater than that of humans demonstrates a large difference in force production capabilities between species (Thorpe et al., 1999). The most pronounced differences in PCSA may indicate the heightened importance of the subscapularis, supraspinatus and middle deltoid in producing rotational and stabilizing forces about the shoulder and glenohumeral joint in particular.

At the rotator cuff, muscular contribution from subscapularis and supraspinatus was almost exclusively only predicted in the chimpanzee glenohumeral model. While chimpanzees generally have a greater overall muscle mass and relative PCSA in their upper extremity than humans, the difference between species in each individual muscle also varies (Sonnabend and Young, 2009). Because of a wide breadth of upper extremity-inclusive locomotor behaviors, primates, including humans and chimpanzees, typically have a ‘subscapularis dominant’ rotator cuff (Mathewson et al., 2014). Yet, the human supraspinatus and subscapularis are relatively smaller than those of other primates that habitually use their upper extremity in a climbing or suspensory capacity (Inman et al., 1944; Larson, 2015; Mathewson et al., 2014; Sonnabend and Young, 2009). The greater absolute and relative PCSA of the chimpanzee supraspinatus and subscapularis indicates a greater capacity for producing forces to stabilize the glenohumeral joint (Larson and Stern, 1986). The smaller size of the human supraspinatus and subscapularis may have made the human shoulder less idealized for the strenuous tasks of climbing and suspension, and thus weak mechanical contributors.

Less multi-muscle contribution in the computational predictions of the rotator cuff could represent biological unsustainability of weight-bearing suspension and climbing in humans. Both the subscapularis and infraspinatus exert an inferior pull about the glenohumeral joint in chimpanzees and humans, countering the superior action of the deltoids and supraspinatus in many postures (Inman et al., 1944; Roberts, 1974), and maintaining the width of the subacromial space. It is surprising that the supraspinatus was not active in the human model. The supraspinatus is generally active during overhead activities to elevate the arm in synergy with the deltoids (Inman et al., 1944). As a result of the use of an optimization routine to predict muscle forces, the greater force-producing capacity of the deltoids was likely selected over the small supraspinatus in the human model to provide the required superior force. Unlike habitually climbing and suspensory chimpanzees, the entire modern human rotator cuff has evolved to be small, to reduce segment mass and inertial properties for increasingly non-weight-bearing behaviors that require less muscular effort (Larson et al., 2000; Raichlen, 2006; Schoonaert et al., 2007; Taylor et al., 1974). This has allowed a more energy efficient redirection of muscular effort toward non-locomotor modern behaviors, particularly those below shoulder height, such as tool making and manipulating, and hunting and throwing (Arias-Martorell, 2019; Mathewson et al., 2014; Roach et al., 2013; Sonnabend and Young, 2009; Young et al., 2015). As the subscapularis is particularly reduced in size and weakened in humans (Mathewson et al., 2014), the infraspinatus is the primary preventative means to superior migration of the humeral head and reduction of the subacromial space. The human model predicted high muscle forces from the infraspinatus to provide inferior forces about the glenohumeral joint, a scenario that would accelerate infraspinatus fatigue. Along with superior humeral head migration, fatigue of the infraspinatus can reduce the posterior tilt and lateral rotation of the scapula necessary to widen the subacromial space in overhead postures (Borstad et al., 2009; Ebaugh et al., 2006b; Tsai et al., 2003). As chimpanzees have a large infraspinatus and subscapularis, in part to counter deltoid action, this may guard against overload of a single muscle, as was predicted in the model.

Model-predicted activation of the deltoids differed between species in both amplitude and timing. The deltoids are active during the support phase in experimental studies on chimpanzees and humans, to raise the arm and counter traction at the glenohumeral joint from hanging and suspension (Larson and Stern, 1986; MacLean and Dickerson, 2019). Humans were predicted to activate all three deltoids to a much higher degree than chimpanzees. Chimpanzees have greater force-producing capacity in their upper extremity muscles than humans, including the deltoids, and lower body mass (Thorpe et al., 1999; Walker, 2009). As such, a lower percentage of their musculature was predicted to complete the same postural task as humans. The chimpanzee model predicted no contribution from the anterior deltoid in support phase, whereas the human model predicted the greatest contribution from the anterior deltoid. This may be due to differences in muscle lines of action, as computational models and optimization routines are often very sensitive to variation in muscle lines of action and subsequently selective about which muscles are active in specific postures (Latash, 2012; Nussbaum and Chaffin, 1996). Given the broad base of origin of the deltoid across the scapula and clavicle (Inman et al., 1944), the modeled anterior deltoid line of action was likely positioned as the most efficient prime mover in the human model for producing a superior force at the glenohumeral joint. The greater predicted activation of all three deltoids in humans could have repercussions for shoulder function, as activation of arm abductors decreases the subacromial space (Graichen et al., 2001).

The subacromial space was wider at all static instances of the suspension cycle in the chimpanzee model, a possible geometric mechanism for reduced subacromial impingement risk in the species. The difference between species was approximately 2 mm throughout the entire suspension cycle. The laterally projecting human acromion is typically sloped inferiorly, unlike in chimpanzees, which can reduce the width of the subacromional space across most postures (Voisin et al., 2014). The increased space between the humerus and acromion in chimpanzees would provide a wider berth for tissues in the subacromial space, such as the supraspinatus, throughout the range of shoulder elevation, reducing the risk for impingement of tissues (Lewis et al., 2001). While the reduction in size and absolute PCSA of the supraspinatus in humans brought about decreases in force production, it is likely related to the need to exist in a narrower subacromial space (Voisin et al., 2014).

Differences between species in the subacromial space are partly the result of geometric changes in scapular bone shape. Chimpanzees have a more superiorly oriented glenoid, scapular spine and acromion. This reorients the lines of action of the deltoids and rotator cuff to optimize overhead behaviors, particularly propulsive arm swinging motions (Larson, 2007; Larson and Stern, 1986; Roach et al., 2013). Humans have a laterally oriented glenoid, scapular spine and acromion. Humans also have an enlarged and widened acromion process, and lateral projection of the acromion over the glenohumeral joint (Schultz, 1968; Voisin et al., 2014). The laterally oriented human glenoid establishes the human range of motion as ideal for the use of the hands in front of the body and below the shoulder, while also optimizing the glenohumeral muscle lines of action for lateral motions such as throwing (Larson, 1988; Roach et al., 2013). The lateral projection of the acromion changes the mechanical leverage of the deltoids, shifting the muscle origin to be over the joint. This improves the deltoid moment arm in below-the-shoulder action and compensates for the reduced force production of the supraspinatus muscle (Lewis et al., 2001; Voisin et al., 2014). However, the lateral orientation and projection of the acromion also reduces the subacromial space, resulting in higher injury risk for impingement in humans (Lewis et al., 2001; Voisin et al., 2014).

The driving forces behind and stages of evolutionary change in the human subacromial space and rotator cuff, and the propensity for rotator cuff pathology remain unclear. Tissue mass and shape are modified by external stimuli and loading (Byron et al., 2011; Green et al., 2012; Robling et al., 2006; Ruff et al., 2006; Turner, 2007). Scapular shape and rotator cuff muscle mass and PCSA are influenced by exposure to the external forces experienced through typical upper extremity behaviors. Therefore, evolutionary adaptations to scapular shape and rotator cuff architecture have occurred in conjunction with evolutionary behavioral modifications. The increasing need to walk bipedally, hunt or throw were likely strong influences on the lateralization of the scapula (Larson, 2007; Lewis et al., 2001; Roach et al., 2013). Unlike arboreal behaviors, these behaviors maintain below-shoulder multiplanar ranges of motion. It has been hypothesized that the acromion became more lateralized to optimize leverage of the deltoids for below-shoulder behaviors, leading to a lowered mass and PCSA of the supraspinatus tendon (Voisin et al., 2014). This theory may imply that scapular shape adaptations toward modern human behaviors precede adaptations in rotator cuff muscle size and PCSA. However, these changes may have been simultaneous. Concurrent with increasing bipedalism, a parallel reduction in arborealism would have reduced the external forces on the shoulder from locomotion. Increasingly less arborealism would reduce the necessary contribution of the rotator cuff to the repetitive or sustained force-producing arm elevation and axial rotation (Larson and Stern, 2013; Larson, 2015; Sonnabend and Young, 2009). These beneficial evolutionary adaptations optimized the human shoulder for evolutionarily advantageous behaviors such as walking, throwing and tool manipulation, but consequently may have resulted in negative vestigial consequences, particularly for the modern, industrialized human existence.

Relative to the musculoskeletal diversity of life, small differences between closely related species in bone shape and orientation can greatly affect joint function and pathology. All primates share a similar shoulder organization (Pronk, 1991; Sonnabend and Young, 2009). While modern human climbing and overhead capacity is still present, it is greatly reduced compared with that of other primates as a result of these adaptations to the shoulder complex. The rotator cuff and deltoids form a series of force couples around the glenohumeral joint that act to center the humeral head in the glenoid (Larson and Stern, 1986). The superior unit of the force couples is composed of the deltoids and supraspinatus, which elevate the arm, while the inferior unit is composed of the rest of the rotator cuff, which depresses the arm (Inman et al., 1944). The reduced force-producing capability of the subscapularis in humans represents an evolutionary adaptation that has reduced its contribution to centering the humeral head in the glenoid against the superior pull of the deltoids (Potau et al., 2009). This has increased the susceptibility of humeral head superior migration and subsequent subacromial space impingement. Deleterious consequences of superior migration of the humeral head in humans are compounded by the more lateral projection of the acromion over the humeral head and the less superior orientation of the acromion (Larson, 2007; Voisin et al., 2014). Combined, the reduced PCSA, altered scapular shape and composition of the rotator cuff and deltoids may be evolutionary indications of why humans have a propensity for subacromial impingement syndrome that does not exist in other primates.

Alternatively, the primate shoulder, and the shoulder of some extinct hominin species, has protective musculoskeletal mechanisms in the musculoskeletal morphology of the rotator cuff and scapula. If both humans and chimpanzees have evolved from an arboreal common ancestor, chimpanzees have retained the capability and musculoskeletal system necessary for these behaviors. With more massive rotator cuff muscles, the ratio of the deltoid muscle group to the rotator cuff is closer to 1:1 in chimpanzees. This ensures the rotator cuff can counter the superior pull of the deltoids in arm elevation without the risk of early-onset fatigue that would alter glenohumeral biomechanics during high force upper extremity arboreal behaviors. Chimpanzees also retain a superiorly oriented scapular spine and acromion, and an acromion process that does not project as laterally as that of humans (Voisin et al., 2014). This has widened the subacromial space and reduced the area over which the acromion can impinge the supraspinatus over the humeral head. Chimpanzees resultantly have evolved protection against subacromial impingement syndrome and rotator cuff pathology (Lewis et al., 2001). While there is limited information on extant hominin muscle architecture, australopithecines appear to have a scapular spine that falls between the superior orientation of a chimpanzee and the lateral orientation of a modern human scapular spine (Haile-Selassie et al., 2010). Australopithecus afarensis is believed to be bipedal based upon lower extremity morphology, with an upper extremity that may still have engaged in climbing (Crompton et al., 1998; Haile-Selassie et al., 2010; Green and Alemseged, 2012). The intermediate scapular spine orientation of Australopithecus afarensis would reduce the width and occupational ratio of the subacromial space. Therefore, either the rotator cuff was already becoming smaller to occupy a smaller subacromial space or there was incongruency between the size of the supraspinatus and the subacromial space. Either scenario would indicate the possible mild beginnings of reduced arboreal capacity and increased risk of rotator cuff pathology in Australopithecus afarensis, indicative of the slow adaptation away from arborealism in the human evolutionary tree.

A series of assumptions and limitations accompanied the development of the chimpanzee glenohumeral model and the computational comparison between humans and chimpanzees. Many of these decisions stemmed from the unavailability of desirable relevant datasets, but can be modified in future model iterations. While computational musculoskeletal modeling offers numerous benefits to biomechanical studies of the human body, several limitations constrain the present modeling of the glenohumeral joint. The shoulder is considered a three-joint structure, and the acromioclavicular and sternothoracic joints were not considered, though the torso and clavicle were geometrically positioned to dictate the position of the scapula and influence shoulder rhythm (Voisin, 2006). Soft tissue mechanics were simplified in the study to muscle mass and PCSA. Ligaments were set as inactive in both models, and did not contribute to joint stability. Additionally, optimization routines, as used in the glenohumeral model, may overlook the contribution of small muscles to a task in favor of larger muscles with greater force-producing capabilities (Dickerson et al., 2007), reducing synergistic muscle recruitment. Computation musculoskeletal modeling is limited by how researchers can mathematically represent biological phenomena. To model the entirety of the human musculoskeletal system is computationally expensive and often leads to more assumptions, difficulty in interpretation and erroneous results (Cholewicki et al., 1995). Assumptions are crucial in producing models that adequately address the primary research questions at hand. The present model aimed to compare the glenohumeral musculoskeletal behaviors between two species, particularly differences in muscle patterns and overall usage. The models are considered to have achieved this purpose.

The postural analysis run in the present study did not replicate the full breadth of differences between chimpanzees and humans. The kinematic inputs for the chimpanzee model derived from human experimentation. This would have reduced the realism of the joint center positions, subacromial space width and joint angle decomposition. However, if the kinematic inputs are not representative of a chimpanzee suspension kinematics, then the identification of differences stemming from this comparison between species are likely conservative. The model was run statically, not dynamically. This negated the effect of motion and momentum, which would influence joint forces, and muscular recruitment patterns in a powerful and propulsive behavior such as brachiation (Larson and Stern, 1986). Assumptions were made about hand force in each of the support phase static instances. For the initial static analysis, whole or half body mass operated as a provisional representative of hand force in the three support phase discrete instances of the horizontal bimanual arm suspension cycle and in the direction of gravitational force only. This approach disregarded the multidimensional effect of the hand forces at the handhold. A static analysis was considered the most appropriate initial analysis to make comparisons between species in glenohumeral function, given the complications of assumptions required for dynamic modeling of the suspension cycle in the absence of high-quality chimpanzee kinematic data. The models are both capable of running dynamic assessments of climbing and suspension for future study.

The models are both limited by the choice of individual and musculoskeletal data used to represent each species. Each model was run using postural, anthropometric, task-specific inputs from a single, average individual. As well, the geometry module of both models used bone scan inputs from a single, different individual. Other musculoskeletal features and parameters, such as segment parameters, muscle PCSA, origins and insertions, relied on collected and dissected mean data from a variety of published databases. The choice of these single or mean inputs influenced the results. However, as each of these single inputs or model parameters was selected to represent an average chimpanzee or human, the results as presented provide an initial view of the differential musculoskeletal model outputs. The model can run multiple subject inputs when available, or be run probabilistically through appropriate software, to give better population level predictions. As large variability exists in both kinematics and bone geometry of both species, each model is limited by how much the motion data and musculoskeletal geometry represent an average individual. There is limited data on chimpanzees, however, so the data utilized in the present study represented the best available options. Should improved musculoskeletal or motion data become available, both models would be highly receptive of new modular parameters.

Unique assumptions related to chimpanzee musculoskeletal behavior were incorporated into the creation of the chimpanzee model, owing once more to limited data on chimpanzees. The shoulder rhythm algorithms were adapted from a previously developed human shoulder rhythm and used two-dimensional x-rays to define modified chimpanzee scapular and clavicular orientations. The limited postures of the chimpanzees in these x-rays likely affected the approximated boney orientations and resulting shoulder rhythm, particularly the anterior/posterior tilt. This may have also affected the realism of the subacromial space width predictions. Scapulothoracic contact force application sites, joint centers and glenohumeral contact force constraints were estimated from regression equations designed for the human upper extremity. Specific tension was assumed to be equivalent to human estimates. The paucity of complete information required considerable flexibility, experimentation and adaptability during the construction of this model. However, the model was created to enable efficient algorithm adjustment based on alternative hypotheses, and to allow modifications as novel future musculoskeletal and kinematic data on chimpanzees arise. The very existence of the initial exploratory model is imperative for progress and the success of future studies on primate evolutionary shoulder function.

The novel chimpanzee model is designed for a specific use and contains a series of biomechanical and physiological assumptions, both of which must be considered when interpreting results. This model is intended for comparative analyses with other species models only. It is not meant to be a standalone model to predict chimpanzee musculoskeletal shoulder behavior. Rather, the model is designed to run concurrently with a parallel model, to provide insight into differences between species in musculoskeletal function in analogous computational environments and conditions. These musculoskeletal outputs can include muscle moment arms, muscle patterns and coordination, and joint kinetics and stability for any upper extremity task. The model utilizes kinematic, anthropometric and task-specific inputs and any interpretation of comparative model results are limited by the quality of these inputs, particularly kinematic inputs. The model is designed with specific internal settings, and the outputs are also dependent on the quality of these parameters. These factors have been outlined in Table 3 to elucidate the limitations of the model in its current form and how improvements can be implemented should new chimpanzee data become available.

The model can be used to test additional specific, comparative hypotheses through further analyses. The chimpanzee glenohumeral model does not represent a substitution for in vivo and in vitro evolutionary and comparative studies. Rather, the model serves as an alternative method of analysis for testing a variety of musculoskeletal computational ‘what if’ scenarios that may be difficult to test morphometrically or experimentally (Hutchinson, 2012). Hypotheses that can be tested with comparative models include between-species differences in muscle coordination and contribution, subacromial space width, joint stability, and joint forces and moments. Many specific hypotheses can be tested using the model by modifying inputs and internal model parameters. Future analyses can test different inputs, including motion data, anthropometrics and task-specific inputs. Additional upper extremity behaviors that may be valuable to analyze computationally could include different throwing techniques, reaching and arm swing during gait. Internal model parameters can also be modified to test other comparative hypotheses. Different bone scans, PCSA values, muscle origins and insertions, shoulder rhythm equations or glenohumeral stability ratios can be applied to test the effect of these geometric musculoskeletal settings on musculoskeletal outputs. Finally, performing multi-subject analyses using multiple matching kinematic, anthropometric and task-specific inputs will provide data samples for greater population-level comparative analyses. Alternatively, probabilistic analyses can be applied to both model inputs to assess functional variability between species, and model parameters such as PCSA, bone shape and lines of action to assess functional sensitivity (Langenderfer et al., 2008; Chopp-Hurley et al., 2014).

While prevalent in biomechanics and engineering, computational modeling is still recent and largely unexplored in evolutionary science. Classical measuring techniques in physical anthropology have limitations that may discount the manner in which features of the musculoskeletal form operate synergistically within a complex mechanical system (Hutchinson, 2012). Computational models such as the presently developed chimpanzee model allow assessment of specific features of the musculoskeletal system, and how they interact to produce strategies for movement at the shoulder. Primate computational models provide benefits to exploring and answering evolutionary, fundamental physiological and biomechanical, and modern human functional questions. Understanding evolutionary adaptations of the modern human shoulder can aid our understanding of specific strengths and weaknesses of the modern shoulder, the root causes of injury risk, and how to avoid them.

The present results confirm that while chimpanzees and humans have very similar gross musculoskeletal anatomy, changes to the musculoskeletal system have influenced muscle force production in overhead postures. The laterally orientated glenoid and laterally oriented and projected acromion have narrowed the human subacromial space. The reduced PCSA of many muscles crossing the human glenohumeral joint have modified joint stabilizing force couples. As a result, muscles like the infraspinatus may be overloaded and highly susceptible to fatigue in overhead postures. These adaptations have enabled essential modern human behaviors, but may also explain the modern human propensity for subacromial impingement and rotator cuff tears.

The authors would like to acknowledge the contribution of Dr Susan Larson (Stony Brook University), Dr Nathan Thompson (New York Institute of Technology), Dr Brian Umberger (University of Michigan) and Dr Matthew O'Neill (Midwestern University), for providing chimpanzee EMG, x-ray and skeletal data, respectively, which greatly helped in the development and evaluation of the chimpanzee model. Some results, figure legends and discussion in this paper are reproduced from the PhD thesis of Kathleen MacLean (MacLean, 2018).