Trabecular Bone Structure in the Humeral and Femoral Heads of Anthropoid Primates


  • Timothy M. Ryan,

    Corresponding author
    1. Department of Anthropology, Pennsylvania State University, University Park, Pennsylvania
    2. Center for Quantitative Imaging, Pennsylvania State University, University Park, Pennsylvania
    • 409 Carpenter Building, Department of Anthropology, Penn State University, University Park PA 16802
    Search for more papers by this author
    • Fax: 814-863-1474

  • Alan Walker

    1. Department of Anthropology, Pennsylvania State University, University Park, Pennsylvania
    Search for more papers by this author


The functional significance of three-dimensional trabecular bone architecture in the primate postcranial skeleton has received significant interest over the last decade. Some previous work has produced promising results, finding significant relationships between femoral head trabecular bone structure and hypothesized locomotor loading in leaping and nonleaping strepsirrhines. Conversely, most studies of anthropoid femoral head bone structure have found broad similarity across taxonomic and locomotor groups. The goal of this study is to expand on past analyses of anthropoid trabecular bone structure by assessing the effects of differential limb usage on the trabecular bone architecture of the forelimb and hindlimb across taxa characterized by diverse locomotor behaviors, including brachiation, quadrupedalism, and climbing. High-resolution x-ray computed tomography scans were collected from the proximal humerus and proximal femur of 55 individuals from five anthropoid primate species, including Symphalangus syndactylus, Papio sp., Presbytis rubicunda, Alouatta caraya, and Pan troglodytes. Trabecular bone structural features including bone volume fraction, anisotropy, trabecular thickness, and trabecular number were quantified in large volumes positioned in the center of the humeral or femoral head. Femoral head trabecular bone volume is consistently and significantly higher than trabecular bone volume in the humerus in all taxa independent of locomotor behavior. Humeral trabecular bone is more isotropic than femoral trabecular bone in all species sampled, possibly reflecting the emphasis on a mobile shoulder joint and manipulative forelimb. The results indicate broad similarity in trabecular bone structure in these bones across anthropoids. Anat Rec, 293:719–729, 2010. © 2010 Wiley-Liss, Inc.


Reconstructions of locomotor behavior in the fossil record frequently rely on the hypothesized relationships among bone form, biomechanical function, and the external loads generated during locomotor behaviors. Although analyses have traditionally focused on interspecific comparisons of cortical bone geometry and external morphological characters, advances in computing hardware and imaging modalities such as high-resolution computed tomography have opened up new avenues for studies of the three-dimensional (3D) structure and function of trabecular bone. Studies utilizing these advanced analytical tools have typically focused on the proximal femur and have assumed that trabecular bone architecture reflects the magnitude and orientation of external loads experienced by a bone. Therefore, trabecular structure is believed to hold a strong functional signal that can be used in behavioral reconstructions of extinct forms (Fajardo and Müller, 2001; MacLatchy and Müller, 2002; Ryan and Ketcham, 2002a, b; Viola, 2002; Ryan and Ketcham, 2005; Ryan and Krovitz, 2006; Fajardo et al., 2007; Scherf, 2007; Gosman and Ketcham, 2009).

The mechanical significance of trabecular bone architectural variation has been clearly established, although different studies come to different conclusions regarding the role of various structural features depending on the sample analyzed and the methods employed. The primary determinants of the elastic properties of trabecular bone appear to be the bone volume fraction, apparent density, and fabric anisotropy (Hodgskinson and Currey, 1990a, b; Turner et al., 1990), with up to 94% of the variance in the elastic modulus of trabecular bone explained by these factors. Kabel et al. (1999b), using micromechanical finite element analyses, found that fabric measures and volume fraction explained between 93.9% and 95.6% of the variance in the stiffness values of trabecular bone, and between 89.2% and 89.4% for the compliance values. Odgaard et al. (1997) showed that the elastic principal directions could be accurately predicted by the various anisotropy measures including mean intercept length (MIL) and star volume distribution (SVD), thereby confirming that the mechanical behavior of trabecular bone is governed mainly by its 3D architectural arrangement rather than the material properties of the constituent matrix material (Cowin, 1997). More recently, Mittra et al. (2005) found that the structure model index, a measure of the relative composition of trabecular bone in the form of plates and rods, performs equally well compared to bone volume fraction in predicting the ultimate strength of trabecular bone. What is clear from these analyses is that characterization of the 3D structure of trabecular bone reveals important information about the elasticity of the structure and, by extension, the mechanical and functional adaptation of the bone. It is the task of functional morphologists to delineate the interspecific variation in bone structure and its adaptive significance in primates and other mammals.

Most analyses of trabecular bone structure in primates have concentrated on the postcranial skeleton (in particular the proximal femur) and have generally produced equivocal results (Fajardo and Müller, 2001; MacLatchy and Müller, 2002; Ryan and Ketcham, 2002a, b; Viola, 2002; Ryan and Ketcham, 2005; Ryan and van Rietbergen, 2005; Ryan and Krovitz, 2006; Fajardo et al., 2007; Scherf, 2007; Cotter et al., 2009). Few of these studies have specifically examined intraindividual or intraspecific variation in trabecular bone structure, and none except one preliminary study with small sample sizes (Fajardo and Müller, 2001) has examined trabecular bone variation in relation to differential loading of the forelimbs and hind limbs during locomotion.

The purpose of this study is to investigate whether there are significant differences in trabecular bone architecture in the humeral and femoral heads reflecting differential limb usage in primates with divergent locomotor behaviors. To assess morphological variation related to limb usage during locomotion, mechanically relevant trabecular bone parameters (e.g., bone volume fraction and anisotropy) and other potentially informative morphological features (e.g., trabecular thickness and number) are analyzed. Brachiating Symphalangus is predicted to have relatively isotropic trabecular bone structure in the humerus, reflecting more diverse loading conditions at the shoulder joint as compared to the hip joint. Additionally, femoral trabecular bone structure in Symphalangus is predicted to have lower bone volume fraction than that in quadrupedal taxa (Rafferty and Ruff, 1994). Terrestrial quadrupeds such as Papio and Presbytis, however, experience relatively equal vertical peak forces in the forelimb and hindlimb during locomotion (Kimura et al., 1979; Kimura, 1985; Demes et al., 1994; Hanna et al., 2006), and thus trabecular bone of the proximal humerus and femur is not expected to differ significantly in volume or fabric anisotropy. Arboreal quadrupeds such as Alouatta and Pan, by contrast, generate higher hindlimb forces than forelimb forces (perhaps as a result of significant climbing), and therefore should have higher bone volume fraction in the femoral head than in the humeral head (indicating higher elastic modulus). Finally, Pan is also expected to have significantly more trabecular bone (higher bone volume fraction) in the femoral head than in the humeral head reflecting higher peak vertical forces in the hindlimb experienced during locomotion (Kimura et al., 1979; Kimura, 1985; Demes et al., 1994; Hanna et al., 2006).


High-Resolution Computed Tomography Scanning and Image Processing

The skeletal sample used in this study consisted of the following anthropoid primates: Papio sp. (baboons) and Presbytis rubicunda (red leaf monkey), general quadrupeds; Alouatta caraya (black howling monkey) and Pan troglodytes (common chimpanzee), quadrupedal climbers; and Symphalangus syndactylus (siamang), a brachiator. Details of the sample are listed in Table 1. All specimens were wild-shot adults with no external signs of pathology on the elements used in the study. No attempt was made to estimate the age at death of the specimens used in this study. The possibility exists that bone pathologies (e.g., osteoporosis or osteopenia) or trauma in other parts of the skeleton could affect the skeletal structure through changes in locomotor behavior or activity levels during the animal's life, but these issues were not specifically addressed in this study. One femur and one humerus from each individual were borrowed from the collections of the American Museum of Natural History (Papio, Presbytis, Alouatta, Pan) or the National Museum of Natural History, Smithsonian Institution (Papio, Symphalangus). Bones from both right and left sides were used in the sample, but only elements from the same side were used within a single individual.

Table 1. Sample used in this study
Taxonn (f/m/u)Body mass (kg)Locomotion
  • a

    Papio group includes 3 P. cynocephalus, 2 P. hamadryas, 1 P.h. anubis, 1 P.h. tesselates, and 4 P. ursinus.

Symphalangus syndactylus7 (4/3/0)10.7–11.9Brachiation
Papio sp.a11 (5/2/4)9.9–25.1Terrestrial quadrupedalism
Presbytis rubicunda4 (1/3/0)6.17–6.29Arboreal quadrupedalism
Alouatta caraya15 (8/7/0)4.33–6.42Arboreal quadrupedalism, climbing
Pan troglodytes18 (5/12/1)33.7–42.7Terrestrial quadrupedalism, climbing

All specimens were scanned on the OMNI-X HD-600 High-Resolution X-ray Computed Tomography (HRCT) scanner (Bio-Imaging Research, Inc, Lincolnshire, IL) at the Center for Quantitative Imaging at Pennsylvania State University. Each bone was individually mounted in florist foam and positioned vertically in the scanner. Serial cross-sectional scans were collected beginning in the shaft and proceeding proximally to cover the entire femoral or humeral head. For the femur, scans were collected beginning at or near the level of the lesser trochanter. In the humerus, scan data were collected beginning just below the surgical neck and progressing proximally. Scans were collected with source energy settings of either 180 kV/0.11 mA or 150 kV/0.2 mA, 1,440 views with three samples per view, and a Feldkamp reconstruction algorithm. The differences in energy settings were the result of refinement of the bone scanning protocols at the Center for Quantitative Imaging (Pennsylvania State University) over the last 5 years. These differences in x-ray source energy should have no effect on the evaluation of trabecular structure in this study for several reasons. First, both energy settings are sufficient for imaging bones of a variety of sizes, providing both sufficient penetration of the bone structure (i.e., no beam hardening like in smaller scanners with lower energies) and a sufficiently small x-ray spot to resolve trabecular structures (between 0.020 and 0.030 mm) given the voxel sizes used in this study (see later). Second, quantitative segmentation (detailed later) was carried out on each image dataset individually rather than using a single threshold value for all scan series in the sample. This protocol accounts for variations in HRCT image data collection and museum specimen preparation.

For each scan, between 41 and 100 slices were collected during each rotation. Voxel sizes ranged from 0.030 to 0.0687 mm depending on the size of the femoral or humeral head. Images were reconstructed as 16-bit TIFF grayscale images with a 1024 × 1024 pixel matrix. Following scan data collection, the 16-bit images were converted to 8-bit TIFFs using ImageJ 1.40 (, ensuring that the gray values were scaled in the same way for all images in each specimen's dataset.

A single spherical volume of interest (VOI) was defined for each bone for the trabecular bone morphometric analyses following Ketcham and Ryan (2004). The bounding box and center of the joint articular surface for each bone were defined using a multi-step process in Avizo 5.0 (Mercury Systems, Chelmsford, MA). First, a 3D triangulated isosurface of each bone was constructed from the grayscale images with a global threshold value calculated using the iterative segmentation algorithm of Ridler and Calvard (1978; Trussell, 1979). The femoral or humeral head articular surface was then defined for each specimen by selecting only the triangles on the exterior of the reconstructed isosurface that represented the articular surface of the bone (using the surfaceView selection tool). The selection of this surface was accomplished by manually tracing the articular surface and selecting only visible triangle—not the triangles composing the underlying cortical or trabecular bone (i.e., hidden triangles). Although the triangles representing articular surface are generally quite easily distinguished from triangles making up other nonarticular regions of each bone, a precise division between articular and nonarticular regions is not possible to obtain from HRCT data alone, especially at resolutions under 0.07 mm. A conservative approach was taken for all specimens to ensure that nonarticular bone was not included in the articular surface selection. For this reason, the articular surface definitions are not likely to represent the complete articular surface of each bone specimen, which is one reason total articular surface area measurements are not tested in these analyses.

The triangle selection step (described above) resulted in a shell of bone representing a relatively complete articular surface for each specimen (Fig. 1). This articular shell was added to the viewable buffer of the surfaceView module in Avizo, and a new surface was created from these buffer contents. The bounding box for the extracted articular surface was simply the maximum and minimum extents of the articular surface in each of the three orthogonal axes, which because of the mounting scheme used for scanning, were closely aligned with the mediolateral, anteroposterior, and proximodistal axes of each bone. The center of the articular surface shell was calculated by finding the midpoints of the x, y, and z dimensions of the bounding box. For each dimension (x, y, z), the maximum and minimum coordinates of the bounding box were added together and divided by two to find the midpoint of that dimension. While some slight orientation differences may have been present between specimens, it is unlikely they would significantly affect the positioning of the VOIs or the subsequent structural analyses.

Figure 1.

Volume of interest (VOI) selection method. The articular surface of the humeral and femoral heads (shown in red) were extracted from the 3D isosurface reconstruction. The volume of interest was scaled based on the size of a best-fit sphere for the articular surface and was positioned in the center of the humeral or femoral head.

Following definition of the articular surface and determination of the center and bounding box, the diameter of a best-fit sphere for each articular surface was calculated as the average of the x, y, and z dimensions of the bounding box. The radius of the VOI used in subsequent trabecular bone structural analyses was defined as the radius of a sphere with a volume 1/10th the volume of the best-fit articular surface sphere. Each VOI was positioned to superimpose its center with the calculated center of the articular surface constructed in the previous steps. This VOI selection protocol ensured that each VOI was positioned homologously (at the center of the joint), and was scaled to the size of the individual joint being analyzed. The scaled size of the spherical VOIs (i.e., 1/10th the volume of best fit sphere) was selected so that the smallest volumes (i.e., the VOIs from the smallest individuals in the study) had a diameter greater than 5 mm, thereby ensuring that a reasonable amount of trabecular bone was included in these individuals. Volumes ranged in size from ∼7 to 21 mm in diameter for the humerus, and 6 to 17 mm in diameter for the femur.

Trabecular Bone Structural Analysis

The trabecular bone morphometric parameters quantified included the bone volume fraction (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), and fabric anisotropy. These parameters were calculated using the morphometric analysis program QUANT3D (Ryan and Ketcham, 2002b; Ketcham and Ryan, 2004). The 8-bit grayscale dataset for each specimen was input into QUANT3D for analysis. These grayscale images were again segmented using a threshold value calculated from the iterative segmentation algorithm of Ridler and Calvard (1978; Trussell, 1979), based on the grayscale values of the VOI only. This localized segmentation approach ensured appropriate definition of the trabecular bone in the VOI. Segmented data were inspected to ensure appropriate thresholding, and the same threshold value was used for all subsequent morphometric analyses for each individual VOI. Studies of trabecular bone which utilize specimens recently collected from experimental animals and all scanned in the same way (on benchtop microCT systems) typically use a single global threshold for all specimens in the sample. This approach is not appropriate for data collected using high-energy industrial HRCT and museum specimens because of minor fluctuations in the output of the x-ray tube as well as differences in the preparation and preservation of the museum specimens. For this reason a unique threshold was calculated for each specimen based on its histogram of grayscale values.

The bone volume fraction (BV/TV) is defined as the number of bone voxels/total voxels in the VOI. Trabecular number (Tb.N) is an estimated parameter based on the number of intersections between a grid of lines and the bone, normalized by total grid line length (Odgaard, 1997). The fabric structure was quantified using the star volume distribution (SVD) method (Cruz-Orive et al., 1992; Ketcham and Ryan, 2004). The SVD describes the distribution of material around a typical point in a structure. Intercept lengths (the longest uninterrupted line lying entirely within bone at a particular orientation in 3D space) are determined for some number of orientations (Cruz-Orive et al., 1992; Ketcham and Ryan, 2004). Details of the SVD method can be found in Ryan and Ketcham (2002b) and Ketcham and Ryan (2004). In this study, linear intercepts were measured using 2,049 uniformly distributed orientations and 8,000 random points in three-dimensional space.

The orientation and intercept data from the SVD analyses were compiled into a 3×3 weighted orientation matrix that describes the distribution of bone in three dimensions. The three eigenvalues of this matrix, equation image, represent the magnitude of the primary, secondary, and tertiary material axes of the bone structure, and the eigenvectors, equation image, represent their orientations in space. The degree of anisotropy (DA) was calculated as equation image/ equation image. Trabecular thickness (Tb.Th) was calculated using the intercept data collected for the SVD analysis by defining Tb.Th as the shortest measured intercept lying in bone at each of the 8000 points.

Statistical Analysis

Within-species differences in trabecular structure parameters between the humerus and the femur were tested using a paired-sample t-test. ANOVA was used to test whether any of the species differed for each of the measured trabecular bone parameters. In cases where ANOVA demonstrated a significant difference, a post hoc Tukey test was used to determine which species differed significantly from one another. In all statistical tests, null hypotheses were rejected for P-values less than 0.05.


Reconstructed volumes extracted from the femoral and humeral heads of representative specimens from each taxon are shown in Figure 2. It is qualitatively clear that femoral trabecular bone is generally thicker and has a higher volume fraction than humeral trabecular bone in all taxa examined. Although these cubes are only 5 mm in size and do not represent the entire VOI used in the analyses, each of these cubes was taken from within the VOI and is representative of the trabecular structure in each specimen.

Figure 2.

Three-dimensional reconstructions of cubic trabecular bone specimens from the humerus (top) and the femur (bottom) of each taxon. The cubes have a 5 mm edge length and are centered within the volumes used for analysis. These cubes are smaller than the volumes measured and analyzed, but illustrate the trabecular bone architecture within the humeral and femoral heads of these taxa.

Summary statistics for the measured trabecular bone parameters and results of the paired t-tests between the humerus and femur within each species are listed in Table 2. BV/TV values are significantly higher in the femoral head than the humeral head in all taxa. The highest mean BV/TV values for the femur are found in Pan followed closely by Presbytis and Papio with mean BV/TV values of 57%, 53%, and 52% bone in the VOI, respectively. The lowest BV/TV in the humerus is in Alouatta with less than 30% of the VOI being made up of bone.

Table 2. Summary statistics and results of paired t-test for trabecular bone structure in the femur and humerus of each species
TaxonBV/TVSVD DATb.Th (mm)Tb.N (mm−1)
  • For each parameter, the mean is listed with standard deviation in parentheses. The t value and significance are listed on the bottom line of each cell.

  • *

    P < 0.05,

  • **

    P < 0.01,

  • ***

    P < 0.001.

Symphalangus syndactylus0.386 (0.041)0.479 (0.060)1.75 (0.32)2.61 (0.45)0.186 (0.020)0.231 (0.041)1.88 (0.29)1.90 (0.24)
Papio sp.0.382 (0.060)0.516 (0.068)1.78 (0.54)2.52 (1.04)0.157 (0.023)0.256 (0.085)2.28 (0.30)1.98 (0.56)
Presbytis rubicunda0.354 (0.076)0.532 (0.051)2.13 (0.60)3.02 (0.87)0.136 (0.025)0.191 (0.027)2.34 (0.18)2.47 (0.19)
Alouatta caraya0.298 (0.022)0.491 (0.043)1.46 (0.26)2.40 (0.69)0.149 (0.013)0.209 (0.039)1.86 (0.21)2.16 (0.32)
Pan troglodytes0.407 (0.042)0.569 (0.074)1.59 (0.15)2.43 (0.76)0.173 (0.018)0.251 (0.057)2.17 (0.22)2.09 (0.31)

Trabecular bone in the humeral head is significantly more isotropic than trabecular bone in the femoral head in all taxa except Presbytis. Presbytis has the highest DA in both the humerus and femur, with Alouatta and Pan having the lowest values in each bone. Just as with the BV/TV, mean Tb.Th is significantly lower in the humerus than in the femur in all taxa. Humeral trabeculae are on average less than 0.200 mm in size, with Presbytis and Alouatta having the thinnest trabeculae in both bones. Mean trabecular thickness in the femur ranges from 0.191 to 0.251 mm. One Papio individual has extremely thick trabeculae with a mean value of 0.470 mm. This individual is a Papio hamadryas of unknown sex. Based on the dimensions of its femoral head, it is likely this individual is a male, but it is not significantly larger than any of the other Papio males in the sample.

The estimated number of trabeculae per line length in the femur and humerus are generally similar across all taxa. Values for Tb.N range from 1.88/mm to 2.47/mm. The only significant differences between the femur and humerus occur in Papio and Alouatta. In Papio, the estimated number of trabeculae in the femoral head is significantly smaller than that in the humeral head. The opposite pattern was found in Alouatta in which there appear to be significantly fewer trabeculae in the humerus than in the femur. So, although the trabeculae are significantly thicker in the femur than in the humerus in all these taxa, the number of trabeculae is only reduced in Papio. These results indicate that in these taxa higher BV/TV results from thicker trabeculae, not from increased number of trabeculae.

The results for individual specimens are plotted in Figure 3 as a comparison of femoral to humeral bone structure. Comparing the femoral and humeral results graphically, few interspecific trends emerge. The overall pattern for BV/TV, DA, and Tb.Th is for higher values in the femoral head than the humeral head. Several of the Symphalangus individuals fall at the lower end of the femoral BV/TV distribution. These five individuals appear to have more similar humeral and femoral bone volume fractions than most of the other taxa in the sample, but the difference between the humeral and femoral bone structure remains significant (Table 2). The Tb.N results depart from the other three measured parameters in that they are more variable with most of the Pan and Papio specimens having more trabeculae in the humerus than in the femur.

Figure 3.

Comparison of femoral and humeral bone structure in each individual in the sample. The lines in each graph represent unity between the humeral head and femoral head measurements for each variable.

The results of the interspecific comparisons for the humerus are shown in Table 3. Alouatta has significantly lower humeral head BV/TV than any of the other taxa except Presbytis. Presbytis has significantly more anisotropic humeral head trabecular bone than Alouatta or Pan. Symphalangus has significantly thicker humeral trabeculae than any other taxon except Pan. Symphalangus has significantly fewer trabeculae (Tb.N) in the humerus than Papio or Presbytis, while Alouatta has fewer trabeculae in the humerus than any other species in the sample except Symphalangus. The only significant differences between taxa in the measured femoral parameters is that Pan has significantly higher BV/TV than both Symphalangus and Alouatta (Pan vs Symphalangus: P < 0.05; Pan vs Alouatta: P < 0.01). All other interspecific comparisons in the femur are not significant.

Table 3. Results of ANOVA statistical comparisons for the humerus between species
  S. syndactylusPapioP. rubicundaA. caraya
BV/TVS. syndactylus   
 P. rubicundaNSNS 
 A. caraya<0.001<0.0001NS
 P. troglodytesNSNSNS<0.0001
DA S. syndactylusPapioP. rubicundaA. caraya
 S. syndactylus   
 P. rubicundaNSNS 
 A. carayaNSNS<0.001
 P. troglodytesNSNS<0.05NS
Tb.Th S. syndactylusPapioP. rubicundaA. caraya
 S. syndactylus   
 P. rubicunda<0.001NS 
 A. caraya<0.001NSNS
 P. troglodytesNSNS<0.01<0.01
Tb.N S. syndactylusPapioP. rubicundaA. caraya
 S. syndactylus   
 P. rubicunda<0.05NS 
 A. carayaNS<0.0001<0.01
 P. troglodytesNSNSNS<0.01


Primates employ a variety of locomotor behaviors ranging from generalized terrestrial and arboreal quadrupedalism to more highly specialized behaviors such as brachiation, vertical clinging and leaping, and bipedalism. One of the distinct differences between these behavioral patterns is the differential use of the forelimb and hindlimb during locomotion, especially comparing brachiators with bipeds or leapers. The goal of this analysis was to assess whether this divergent limb use during locomotion is reflected in humeral and femoral head trabecular bone architecture. Overall, the results indicate that patterns of trabecular bone structural variation in the femur and humerus are broadly similar across locomotor groups and do not appear to reflect differential limb usage in brachiating and quadrupedal taxa. The only taxa that appear to follow the predicted patterns of forelimb to hindlimb structural organization are the arboreal quadrupeds Presbytis and Alouatta and quadrupedal Pan in which the femoral head has significantly higher bone volume fraction. The fact that a similar pattern was found in Papio and brachiating Symphalangus, however, strongly suggests that this result is not necessarily driven by particular locomotor behavioral patterns, but is indicative of broad functional and/or developmental similarities across these postcranial joints in primates. The strongest potential functional signal for higher forces in the hindlimb among the quadrupedal primates is the structural divergence between the trabecular bone of the forelimb and hindlimb of Alouatta and Pan. The significantly lower humeral trabecular bone volume in Alouatta when compared with the other taxa suggests very low magnitudes of loading on the forelimb during quadrupedal walking and climbing.

Contrary to expectations, Symphalangus displays the same pattern of trabecular bone structural variation between forelimb and hindlimb as all other taxa in the sample in spite of its unique brachiating form of locomotion. Trabecular bone structure in the proximal femur of brachiators is indistinguishable from that of quadrupedal and climbing taxa in all mechanically important features, which is a surprising result considering the presumably more limited use of the hindlimb in locomotion by hylobatids (i.e., Symphalangus) compared to quadrupedal primates. As predicted, Symphalangus has relatively isotropic humeral head trabecular bone structure compared to the femur, but this pattern is also present in other quadrupedal and climbing taxa analyzed here. Additionally, humeral head trabecular bone structure is not significantly more isotropic in Symphalangus than in any other taxon in the sample. Humeral head trabecular bone may be relatively isotropic across a range of primates due to their generally more mobile shoulder joint and manipulative forelimbs. Symphalangus does have significantly thicker trabeculae in the humerus relative to the other taxa, but the significance of this result is unclear given the relatively low importance of trabecular thickness compared to other structural features like bone volume fraction and structure model index (a measure of the amount of trabecular plates and rods) in determining the elastic properties of trabecular bone (Ulrich et al., 1999; Mittra et al., 2005).

The lack of predicted results for Symphalangus suggests either that loading of the femur is higher than predicted for this taxon, or that trabecular structure in the femur is being driven strongly by other developmental mechanisms (Lovejoy et al., 2002; Lovejoy, 2005). In their analysis of femoral and humeral articular structure and function in brachiators and quadrupeds, Rafferty and Ruff (1994) suggested a decoupling of articular surface area and trabecular bone structure in the primate postcranial skeleton. Using optical luminance measurements from radiographs of the proximal humerus and femur, they found that humeral head mass relative to joint surface area was not significantly higher in Symphalangus than in quadrupeds. On the basis of these results, they suggested that muscular contractions at the shoulder joint act to maintain trabecular bone mass, but that locomotor loading of the shoulder joint is not particularly high in this taxon because it does not support significant proportions of body mass during locomotion. Rafferty and Ruff (1994), therefore, suggested that humeral head trabecular structure in brachiators should not be different than in quadrupedal primates. However, in contrast to the humeral results, Rafferty and Ruff (1994) found that femoral head mass in Symphalangus is lower relative to joint surface area than in quadrupedal monkeys because the hindlimbs are used only rarely in locomotion.

The results of this study only partially support the conclusions of Rafferty and Ruff (1994). Humeral head trabecular bone volume is not significantly higher in Symphalangus, further supporting the idea that the shoulder joint is not loaded excessively during locomotion in spite of the use of the forelimb in brachiation. However, although mean bone volume in the femur is lower in Symphalangus than in any other species in this sample, it is not statistically significant, which is in contrast to the patterns found by Rafferty and Ruff (1994). The differences between the current results and those of Rafferty and Ruff (1994), as well as the lack of significant differences in trabecular structure found by Fajardo et al. (2007), strongly suggest that methodological differences can have a significant impact on studies of trabecular bone structure.

Studies of positional behavior in Symphalangus and other hylobatids may provide some insight into these results. Postural behaviors in Symphalangus are apparently limited to sitting and arm-hanging, with arm-hanging being observed 53% of the time by Chivers (1972) and 61.7% of the time by Fleagle (1976). Bipedal postures were not recorded, so do not likely account for significant loading on the hindlimb. Locomotor behaviors, however, provide potentially interesting insights into hindlimb loading in Symphalangus. In a small sample of observations, Chivers (1972) found that Symphalangus brachiated 80% of the time, but climbed and walked bipedally for the remaining 20% of their locomotion (10% each). Fleagle's (1976) later study, with a larger number of observations, found a much lower amount of brachiation in Symphalangus, reporting brachiation during only 37.9% of locomotor bouts. The observation that Symphalangus climbed during 54.3% of their locomotor bouts, with either leaping/hopping or bipedally walking accounting for the remaining 7.8% of their locomotor bouts (3.2% leap/hop and 4.6% bipedal walk) (Fleagle, 1976), has important implications for this study because a full 62.1% of locomotor bouts were behaviors in which the hindlimb was active. Studies of vertical climbing in gibbons (Hylobates gabriellae and H. leucogenys) indicate that they usually support the body with either one forelimb and one hindlimb or with a combination of three limbs (Isler, 2002). Bipedal support is less frequent but not unusual (between ∼2% and 14% of strides), while bimanual support is rare (Isler, 2002). If the behaviors observed for Hylobates also characterize the behaviors of Symphalangus, they would suggest that hylobatids load their hindlimbs more frequently than predicted based on the generalization of their locomotor behavior as “brachiation.”

The similarities in the trabecular bone structure of the femoral head among primate taxa examined here, and the differences observed between femoral head and humeral head trabecular bone architecture in the same taxa may reflect the fact that the hindlimb is loaded more significantly in compression in a diverse array of postures. Most primates experience higher vertical peak forces on the hindlimb than on the forelimb during quadrupedal locomotion as a result of the shift of the center of mass toward the hindlimbs (Kimura et al., 1979; Kimura, 1985; Demes et al., 1994; Hanna et al., 2006). The results of this study strongly suggest that trabecular bone of the femoral head reflects higher compressive forces in the hip joint, resulting from normal locomotor and postural behaviors in which primates support a significant amount and, at least during some positional behaviors, all of their body weight with the hindlimbs. In this way, the results can be interpreted as a more or less accurate reflection of the differential loading of the forelimb and hindlimb during locomotion, even if specific locomotor patterns can not be distinguished.

Over the last decade, analyses of 3D trabecular bone architecture have become more commonplace and represent an area of burgeoning interest among paleontologists and primate morphologists. Although experimental and modeling work consistently confirm the mechanical significance of architectural variation in trabecular bone structure (Hodgskinson and Currey, 1990a, b; Turner et al., 1990; Cowin, 1997; Odgaard, 1997; Kabel et al., 1999a, b; Mittra et al., 2005), many studies examining trabecular bone structure and locomotor behavior in primates have produced conflicting results (Fajardo and Müller, 2001; MacLatchy and Müller, 2002; Ryan and Ketcham, 2002a, b; Viola, 2002; Ryan and Ketcham, 2005; Ryan and van Rietbergen, 2005; Ryan and Krovitz, 2006; Fajardo et al., 2007; Scherf, 2007). Some of the variation in results derives from differences in the methodologies employed, the taxonomic and locomotor groups analyzed, and the skeletal elements studied. Also contributing to the conflicting results is the somewhat unfortunate historical focus on the proximal femur, which is of obvious interest for the evolution of hominin bipedalism, but which also represents a complex joint not easily understood through simple comparisons of trabecular structure (Skedros and Baucom, 2007).

In analyses of 3D bone structure in the femoral heads of small strepsirrhines, both Ryan and Ketcham (2002b, 2005) and MacLatchy and Müller (2002) found a significant and reliable locomotor signal. In both of these studies, the focus was on the specialized leaping locomotion employed by small-bodied galagids and, in the case of Ryan and Ketcham (2002b), the larger-bodied indrids as well. By contrast, separate studies by Fajardo et al. (2007) and Viola (2002) on larger-bodied anthropoids failed to detect a strong relationship between specialized locomotor behaviors, such as brachiation in gibbons, and trabecular structure in the femoral head and neck (but see Saparin et al., 2009). These results on trabecular microarchitecture in anthropoids also contrast with those obtained from planar x-rays of the femoral neck (Rafferty and Ruff, 1994; Rafferty, 1996).

The results of this study provide further evidence of broad similarity in trabecular bone structure across anthropoid primates. Fajardo et al.'s (2007) analysis of femoral neck trabecular bone structure in a diverse sample of anthropoids found general similarity across locomotor behaviors, phylogenetic groups, and body sizes. The current analysis corroborates Fajardo et al.'s (2007) results and extends them to the proximal humerus. Most significant results in analyses of trabecular architecture have been obtained in comparisons of leaping and nonleaping strepsirrhines (MacLatchy and Müller, 2002; Ryan and Ketcham, 2002b, 2005), which suggests that femoral trabecular architecture can only be used effectively in analyses of extreme locomotor differences (Fajardo et al., 2007). Carlson et al. (2008) recently found no significant differences in trabecular bone structure between groups of mice confined to locomote either linearly or through a series of turns, calling into question the responsiveness of trabecular bone to distinct locomotor behaviors.

The apparent lack of a strong functional signal differentiating locomotor behaviors such as brachiation and quadrupedalism in either the proximal femur or the proximal humerus is likely related to the complexity of loading at the shoulder and hip joints in anthropoids. Linking trabecular structural variation with other bone structural properties, such as patterns of subchondral bone density, may provide better resolution of locomotor differences in joint loading. Recent work on subchondral bone density suggests that this property may carry a stronger functional signal capable of differentiating fine-scale differences in joint loading across various locomotor behaviors (DeRousseau, 1988; Carlson and Patel, 2006; Patel and Carlson, 2007; Polk et al., 2008). The work by Carlson and Patel (2006; Patel and Carlson, 2007) on the distal radius, and Polk et al. (2008) on the knee joint, suggests that subchondral bone density is highly sensitive to peak compressive forces generated during different locomotor behaviors. The underlying trabecular bone, therefore, is likely also to reflect the compressive forces experienced during locomotor and postural behavior. It may be the complexity of the joint forces at the shoulder and hip joints that masks fine-scale locomotor signals in the trabecular architecture of anthropoids, at least when trabecular bone structure is measured in the center of the femoral and humeral heads.

The utility of trabecular bone for locomotor inferences in the fossil record, therefore, remains an open and unresolved issue. One aspect of trabecular bone that is apparent from all these analyses is that it is a complex structure that varies at all levels—interspecific, intraspecific, within a single individual, and within a joint. Further work is needed to more finely delineate the response of trabecular bone to locomotor loading, to determine the most appropriate skeletal regions with the highest potential for resolving functional patterns, and to ascertain the relative mechanical contributions of cortical bone, trabecular bone, and subchondral bone in various joints.

The most significant potential limitation to this study is the size and positioning of the volumes of interest. Previous analyses have clearly shown significant variation in bone structure within a single bone (Ryan and Ketcham, 2002b; Ryan and Krovitz, 2006; Fajardo et al., 2007), so using only one centrally located VOI to characterize bone structure in the femoral and humeral heads may be an oversimplification. An alternative approach would be to position multiple volumes throughout the heads with the idea of characterizing structure across the joints. The use of multiple volumes has been successful in previous analyses (Ryan and Ketcham, 2002b; Fajardo et al., 2007), but presents a challenge in comparing different bones with distinctly different shapes and sizes.

The new method introduced here, in which the VOI is positioned at the calculated center of the joint and scaled based on the size of the articular surface, represents an improvement over previous attempts at homologous VOI positioning and sizing. Ideally, volumes would be positioned to optimize effective analysis of clearly-defined biomechanical hypotheses concerning joint posture and loading. A potentially promising approach is to sub-divide joints into a series of smaller volumes of interest to allow morphometric comparisons between regions within the joint, as well as between individuals and species (Ryan and Test, 2007; Saparin et al., 2009). Such an approach could capture variation across the femoral head and could track interesting structural features such as the relative bone volume fraction, anisotropy, and orientation of the principal compressive arcade. Analyses employing other VOI positioning protocols and a broader range of trabecular structural features such as structure model index, connectivity, and trabecular separation are ongoing. Expanded structural analyses may allow for a better understanding of the interrelationships among the various structural components as well as the elastic properties of trabecular bone volumes via micromechanical finite element models. The possibility exists that other trabecular bone features not quantified here may provide better discrimination of locomotor differences in the primate postcranial skeleton (Scherf et al., 2009).

In conclusion, trabecular bone structure in the proximal humerus and femur is broadly similar across anthropoid primates and appears to reflect higher loading on the hindlimb than the forelimb in all of these primates. Unique loading patterns generated during brachiation and quadrupedalism are not reflected in the trabecular architecture of the proximal femur or humerus. Instead, femoral trabecular bone volume is always higher than humeral head trabecular bone volume, and trabecular bone in the humerus is consistently more isotropic than in the femur. Based on these results, trabecular bone structure in the shoulder and hip joints of anthropoid primates does not provide significant discrimination of locomotor behaviors and therefore may not be a useful predictor of locomotor behavior in the fossil record.


The authors would like to thank Jason Organ, Qian Wang, Timothy Smith, and Valerie DeLeon for inviting them to contribute to this special issue. They also thank Darrin Lunde and Eileen Westwig at the American Museum of Natural History, and Richard Thorington and Linda Gordon at the National Museum of Natural History, Smithsonian Institution, for their assistance with specimens and their willingness to loan specimens for scanning. Thanks to Richard Ketcham who wrote the QUANT3D program used in this analysis, and Ingmar Carlson, Matthew Test, Pat Recinto, and Tyler Wagner who helped with various aspects of image processing and data entry. They thank the two anonymous reviewers who provided helpful comments and insights that improved this manuscript.