The most common way to examine the postcranium as a complex is to use body proportions. Numerous studies have been done on the partially complete skeletons of fossil hominins to quantify forelimb-to-hindlimb length and joint proportions with the aim of comparing these proportions to those seen in modern taxa in order to make inferences about locomotor patterns. Limb proportions are an important tool in determining possible locomotor repertoires in the fossil record as they are tightly correlated with positional behavior. Species that practice more suspensory, upper limb-dominated modes of locomotion tend to have longer forelimbs (both segment and total length) than hindlimbs. Species that practice hindlimb-dominated locomotor patterns, such as vertical clinging and leaping or bipedalism, have longer hindlimbs in proportion to the forelimb. All of the extant apes have high intermembral indices with Pongo and Hylobates achieving the greatest forelimb-to-hindlimb lengths, whereas modern Homo sapiens has a quite low intermembral index reflecting the lengthening of the hindlimb as an adaptation to obligate bipedalism (Jungers, 1985; Fleagle, 1999). Studies of fossil limb and joint proportions in the Plio-Pleistocene record have yielded complicated and sometimes contradictory results because of the fragmentary nature of the fossil record and the necessity of estimating limb lengths from incomplete specimens such as OH 62, KNM-ER 3735, and BOU-VP-12/1 (Johanson et al., 1987; Korey, 1990; Hartwig-Scherer and Martin, 1991; McHenry and Berger, 1998; Asfaw et al., 1999; Richmond et al., 2002; Degusta, 2004; Dobson, 2005; Reno et al., 2005; Green et al., 2007; Haeusler and McHenry, 2007).
Another potential way to examine patterns of postcranial variation to make inferences about locomotor patterns and morphological affinity to modern taxa is through the lens of morphological covariation between the forelimb and hindlimb. The forelimb and the hindlimb are functionally related and their morphology should be constrained by locomotor requirements; so, specific traits in the forelimb and hindlimb could be selected as a unit to meet the requirements of some locomotor activity (Cheverud, 1996).
If the forelimb and hindlimb shapes covary, then it should be possible to examine patterns of covariation between regions of the appendicular skeleton and compare those patterns to those seen in extant reference taxa to make inferences about potential locomotor capabilities of fossil hominins. In other words, it should be possible to try and “match” the morphological pattern in the forelimb and hindlimb of fossil hominins to what is seen in the extant radiation. Although this sort of analysis relies on comparisons with extant taxa for functional conclusions and lacks the biomechanical information present in limb proportion studies, it does side-step the necessity of having complete bones and allows for the use of the more fragmentary material that is more frequently found in the hominin fossil record.
There have been only a few studies that have examined covariation in forelimb and hindlimb structures in adult primates. Rolian et al. (2010) compared the length and shape of manual and pedal phalanges in hominoids and found that they covaried strongly. These authors have suggested that manual phalanx morphology is a byproduct of strong selection on the pedal morphology for efficient bipedality and not necessarily from selection for manual dexterity. Young et al. (2010) tested covariance in length in serially homologous limb modules in humans, great apes, and some Old World monkeys. These authors found that forelimb and hindlimb integration was more relaxed in apes than in quadrupedal monkeys because of the difference in functional roles of the forelimb and hindlimb.
This study seeks to examine the postcranium as an integrative unit in a way that avoids the necessity of estimating limb proportions. Instead of using indices such as limb or joint proportions, shape covariation is used to answer questions about the way that multiple bones of the postcranium could function as a unit and to address questions about potential morphological shape covariation and mosaic evolution.
This study will address three research questions:
1Is there strong shape covariation between the forelimb and hindlimb bones in extant hominoids and, if so, what are the correlated aspects of shape in the forelimb and hindlimb among the extant taxa?
2Is the morphological pattern of covariance in forelimb and hindlimb elements correlated with intermembral indices in the extant hominoids?
3Does the pattern of forelimb and hindlimb covariance differ among associated skeletons of earlier hominins and to what extent do their morphological patterns match those of the extant taxa? This information could potentially be used to make inferences about their locomotor capabilities.
MATERIALS AND METHODS
Three-dimensional geometric morphometrics (3D-GM) is a morphometric approach that allows for the retention of shape information. This information is preserved in most statistical analyses, which allows for the visualization of shape changes among the original specimens (Rohlf and Slice, 1990). Twenty-five landmarks on the humerus, 16 landmarks on the radius, 23 landmarks on the ulna, 32 landmarks on the femur, and 20 landmarks on the tibia were collected as a series of x,y,z, coordinates (Figs. 1–5, see Supporting Information Table S1 for a description of the points) using a Microscribe 3DX digitizer.
The specimens were stabilized and oriented using modeling clay and points were taken in the same orientation on each individual. Data were collected on a modern comparative sample of Homo, Gorilla, Pan, and Pongo (Table 1) as well as original fossils (Table 2). No casts were used in these analyses.
Table 1. List of extant specimens used for this study by species, subspecies, and sex
Virunga isolates; RMCA.
Cameroon (and some gorilla males from Republic of Congo); PCM.
Table 2. List of fossil individuals and elements used
Homo sp. indet.
Omo (Kibish) 1
Humerus, ulna, femur
Humerus, ulna, femur
In the extant sample, all individuals were adults and displayed full epiphyseal closure at all joints. Where sex was not indicated on the specimen, it was determined using the pelvis in H. sapiens and using size for Pongo and Gorilla. If the sex was not given for Pan, or if the pelvis was missing for H. sapiens, specimens were listed as unknown. No zoo specimens were sampled. The H. sapiens sample was chosen from maximally different populations to encompass a range of modern human variability. No modern morphological collections were sampled for modern humans as they presumably exhibit a high degree of genetic admixture; instead, archaeological and more isolated populations were sampled.
Previous studies have noted that covariation in elements of two different individuals within a single species does not differ greatly from matched elements (Harcourt-Smith et al., 2008). Thus, KNM-ER 1503 and KNM-ER 1504 were combined as some researchers believe these represent a single individual (McHenry, 1978) and the distal femur of A.L. 129-1 was analyzed with the forelimb elements of A.L. 288-1 as they are both from the same taxon (Johanson et al., 1978) and are of a similar size.
Each postcranial landmark configuration (humerus, radius, ulna, femur, and tibia) was subdivided into proximal and distal subsets and was separately subjected to a generalized Procrustes analysis (GPA). The GPA rotates, translates, and size scales all of the configurations by minimizing the sum of squares distance between each set of landmarks and the mean configuration; this places all of the landmark configurations in the same shape space. After they have been subjected to GPA, the coordinates can be used in standard multivariate analyses (Rohlf and Slice, 1990).
To assess the degree of shape covariation present in this data set, two-block partial least squares analyses (2B-PLS) were conducted using MorphoJ v. 1.02. (Klingenberg, 2008). 2B-PLS is a way of assessing the covariance between two separate blocks of data. A 2B-PLS generates pairs of linear variables (one for each block) that function to explain the maximum covariance between the blocks. These new variables are useful in that they only describe the covariation between the two blocks, not any covariation within a single block. Mathematically, this is achieved by a singular value decomposition (SVD) of a cross-covariance matrix. This generates scores for each individual along each axis (similar to the PC scores in a principal components analysis), loadings of each variable (eigenvectors), and singular values (eigenvalues) (Rohlf and Corti, 2000).
2B-PLS analyses including the fossils and all extant individuals were run to examine patterns of shape covariance across phylogeny in the modern and fossil taxa. Computing a 2B-PLS of all extant hominoids emphasizes species-specific differences in order to make inferences about overall patterns of morphological covariance among hominoids. This analysis can be used to infer patterns of correlated evolutionary change in morphology (e.g., Bastir et al., 2010 for changes in the cranium).
For each analysis, the degree of covariation and correlation on the first axes was recorded, as well as the RV coefficient. The degree of covariation is the amount of variance of the sample that is explained by the first axes of each block, whereas the correlation coefficient (r) is a measure of how well the two blocks covary on the first axis. The first axis will explain the greatest proportion of the covariance in the sample and is not constrained by being orthogonal to the previous axis (Rohlf and Corti, 2000; Bastir and Rosas, 2005). The RV coefficient is analogous to the Pearson's correlation coefficient, but for high-dimensional data sets, and it can be interpreted in the same was as any coefficient of correlation. It is different from the correlation coefficient of each axis in that it takes into account all axes (Escoufier, 1973). Although there is no specific significance value that can be derived from a 2B-PLS, permutation tests can be used to determine whether or not the relationship between blocks is greater than what would be expected in a random sample. In this case, permutation tests of 10,000 replicates were performed to assess the significance of the relationship between the two blocks. These permutation tests compute RV and correlation coefficients in random blocks of data and compare those results to those of the real data (Rohlf and Corti, 2000; Klingenberg, 2008).
The shape changes along the first axes were visualized in Morpheus et al. (Slice, 1999) by generating new landmark configurations representing the change from the maximum and minimum of those axes. This was done by multiplying the eigenvectors by the position along each axis and adding it to the consensus configuration (Polly, 2008).
Finally, a multiple regression was performed in PAST (Hammer et al., 2001) to test whether forelimb and hindlimb shape covariance can predict intermembral index. In this analysis, average intermembral indices for males and females of each species or subspecies were taken from the literature (Table 3) and were used as the dependent variable, and average PLS scores for males and females of each species or subspecies on the first two pairs of axes were used as the independent variables. Individuals of unknown sex were not included in the species or subspecies means for the PLS scores. In this analysis, an r2 value of 1.0 would indicate that the PLS scores perfectly predict intermembral indices in extant hominoids, whereas a 0 would indicate that there is no correlation whatsoever (Cohen, 1968). Both r2 and an adjusted r2 that takes into account the number of variables and observations in the model are presented, as well as which independent variables were most significant in the regression model. Overall significance of the model was tested using an ANOVA (Hammer et al., 2001).
Table 3. Average intermembral index for males and females by species or subspecies
2B-PLS analyses for all possible pairs of forelimb and hindlimb segments—including fossil individuals that preserved the relevant portions—were completed, and the results of which are presented in Table 4. All pairings were significantly different from the null hypothesis of independence in permutation tests. There were four pairings with relatively high RV scores and strong covariance (P < 0.0001) between the first and second axes: proximal ulna/proximal femur, proximal ulna/distal femur, distal humerus/proximal femur, and distal humerus/distal femur. 2B-PLS graphs were generated for these four pairings (Figs. 2–5; see also Supporting Information Fig. 1 for three-dimensional plots), and the scores were tested to see how well they predict intermembral index (Table 4). The average PLS scores from the first axes of all four pairings have a significant relationship with average intermembral index. In most cases, the multiple regression including all axes performed better than each axis on its own. Of the four pairs, the distal humerus and distal femur had the poorest adjusted correlation coefficient ( ) at 0.43 and P value at 0.04. In all other cases, was above 0.75 indicating that PLS scores are a relatively good indicator of intermembral index and thus likely an equally good proxy for locomotor repertoire.
Table 4. Percent covariance (% cov) and correlation (corr) along the first and second common axes and RV coefficients for pairs of forelimb/hindlimb segments
% cov 1
% cov 2
Bolded entries indicate comparatively high RV values.
Prox ulna/Prox femur
Prox ulna/Dist femur
Prox ulna/Prox tibia
Prox ulna/Dist tibia
Dist radius/Prox femur
Dist radius/Dist femur
Dist radius/Prox tibia
Dist radius/Dist tibia
Prox radius/Prox femur
Prox radius/Dist femur
Prox radius/Prox tibia
Prox radius/Dist tibia
Dist humerus/Prox femur
Dist humerus/Dist femur
Dist humerus/Prox tibia
Dist humerus/Dist tibia
Prox humerus/Prox femur
Prox humerus/Dist femur
Prox humerus/Prox tibia
Prox humerus/Dist tibia
Considering the pattern of shared covariance within this sample, for the distal humerus/proximal femur pairing A.L. 288-1 and KNM-ER 1503/1504 are both Homo-like (Fig. 2), whereas KNM-WT 15000 seems to have a more unique pattern of covariance with a human-like femur and a slightly more ape-like humerus, although this specimen is a juvenile which could affect its placement (Brown et al., 1985) and it is not far outside the range of human variation. The traits that covary most strongly on the first axis are, for the humerus, the height of the medial and lateral epicondyles with Homo having them more distally flexed and the apes having them more proximally flexed. It is also driven by the shape of the trochlea, particularly the proximal-anterior margin. In Homo, that margin is straight, whereas in apes it has a “V” shape. Additionally, Homo has a less symmetrical trochlea than do the apes. The first femoral axis is largely driven by the lateral extension of the greater trochanter and the height of the femoral head with Homo having laterally projecting trochanters and more proximally oriented femoral heads in comparison to the apes. On the second axis, the depth of the trochanteric fossa in the proximal femur covaries with the degree of asymmetry in the trochlea and the position and orientation of the capitulum in the distal humerus. Pan has a deep trochanteric fossa with minimal keeling on the trochlea and a small, less distally projecting capitulum, whereas Gorilla has a shallow trochanteric fossa and a large, more distally projecting capitulum (Table 6).
Table 6. Summary of the characters in the forelimb and hindlimb segments that covaried most strongly within this sample
Distal humerus—axis 1
Proximal femur—axis 1
• Height of the medial and lateral epicondyle • Proximo-anterior margin of the trochlea • Degree of symmetry of the trochlea
• Degree of lateral extension of greater trochanter • Length of femoral neck
Distal femur—axis 1
• Degree of symmetry of the proximal border of the patellar articular surface • Keeling of the patellar groove • Relative condyle size
Proximal ulna—axis 1
Proximal femur—axis 1
• Depth of the trochlear notch • Keeling of the trochlear notch • Size of the olecranon process • Projection of olecranon process
• Degree of lateral extension of greater trochanter • Length of femoral neck
Distal femur—axis 1
• Degree of symmetry of the proximal border of the patellar articular surface • Angle between the patellar surface and the distal articular surface • Relative condyle size
Distal humerus—axis 2
Proximal femur—axis 2
• Size and position of capitulum • Size and lateral keeling of trochlea
• Depth of the trochanteric fossa
Distal femur—axis 2
• Outline of the distal articular surface (square vs. trapezoid)
Proximal ulna—axis 2
Proximal femur—axis 2
• Height of olecranon • Width of olecranon • Width of trochlear notch
• Orientation of the lesser trochanter • Depth of femoral neck Distal femur—axis 2
• Outline of the distal articular surface (square vs. trapezoid)
For the distal humerus/distal femur pairing in this sample, Omo (Kibish) 1 has a shared pattern of forelimb/hindlimb covariance most similar to modern H. sapiens, whereas the other two fossils have a unique pattern of covariance. Australopithecus afarensis (represented by a composite A.L. 288-1 and A.L. 129-1) has a human-like humerus combined with a more ape-like distal femur, whereas KNM-WT 15000 has a human-like distal femur combined with a more ape-like distal humerus (Fig. 3). Again, the fact that KNM-WT 15000 is a juvenile (Brown et al., 1985) cannot be discounted. The traits that covary between the distal humerus and distal femur are, for the humerus, the position of both the medial and lateral epicondyles, which are relatively low in humans as opposed to the apes, as well as the proximal-anterior border and shape of the trochlea. In humans, the proximal-anterior border is straight and the medial corner projects somewhat anteriorly; the trochlea as a whole is less symmetrical than those of the apes. The relative symmetry of the patellar articular surface, degree of patellar keeling, and relative size differential between the medial and lateral condyles covary with the humeral traits; in H. sapiens, the patellar surface is more keeled and asymmetrical than in apes, but the medial and lateral condyles are of approximately equal size. On the second axis, the size and position of the capitulum and the size and degree of lateral keeling of the trochlea in the distal humerus covary with the outline of the distal articular surface. Individuals toward the positive values have more trapezoidal distal outlines and those toward the negative values have more square-shaped distal outlines (Table 6). Overall, there is more overlap between the extant taxa for this pairing.
The pattern of covariance between the proximal ulna and proximal femur was examined. KNM-WT 15000 and A.L. 288-1 both have a modern human pattern of covariance for the proximal ulna and proximal femur (Fig. 4). The shape of the olecranon process and the depth and keeling of the trochlear notch in the ulna covary with the length of the femoral neck and lateral projection of the greater trochanter. Modern humans have a relatively flat and shallow notch in comparison to the apes and an olecranon process that is smaller and does not have a strong posterior projection. The first femoral axis is driven by the low, laterally projecting greater trochanter in modern humans as opposed to the higher, less projecting trochanter of the apes. In addition, the length of the femoral neck also seems to be contributing to this axis with modern humans having a longer neck than the apes. On the second axis, the orientation of the lesser trochanter and the depth of the femoral neck covary with the height and width of the olecranon process and the width of the trochlear notch. Chimpanzees have a narrow trochlear notch with a tall and narrow olecranon process in combination with a more medially projecting lesser trochanter and deep femoral neck when compared with the other taxa (Table 6).
Finally, the proximal ulna and distal femur also covaried strongly along the first axis. In this analysis, the morphological pattern in the forelimb and hindlimb of KNM-WT 15000 most closely resembles the way that modern human proximal ulnae and distal femora covary. The ulna of Omo (Kibish) 1 has a unique shape relative to that of its distal femur in that it has a smaller olecranon process in comparison to a much wider trochlear notch than seen in the other H. sapiens in this sample. KNM-ER 1500 and A.L. 288-1/129-1 both have a unique pattern of covariation among the extant taxa with a human-like proximal ulna paired with slightly more ape-like distal femur (Fig. 5), although their pattern of covariation is not far outside the range of modern humans. The first ulnar axis was driven by the shape of the olecranon process and the depth and keeling of the trochlear notch. Modern humans have a relatively flat and shallow trochlear notch and an olecranon process that is smaller and does not have a strong posterior projection paired with an asymmetrical patellar articular surface, a tight angle between that surface and the distal articular surface and similarly size medial and lateral condyles in the femur, whereas apes have an ulna with a strong posteriolateral projection of the olecranon and a deep, keeled trochlear notch paired with a distal femur that has a symmetrical patellar surface and a wider angle between the patellar and distal articular surfaces. On the second axis, a narrow trochlea with a narrow olecranon covaries with a more trapezoidal distal femur in Pan, whereas Gorilla and most fossil hominins have a wider trochlear notch and a wider, more posteriorly protruding olecranon process paired with a more square-shaped distal femur (Table 6).
In summary, for all of the segment pairings, there is a clear difference in the pattern of covariation seen in Homo versus that of the other large bodied hominoids. KNM-WT 15000 (H. erectus) was included in all analyses and has a morphological pattern that is consistent with that seen in modern humans in the proximal ulna and both segments of the femur, and a unique morphological pattern between the distal humerus and both segments of the femur. A.L. 288-1/129-1a (A. afarensis) was also included in all analyses and had a human-like morphological pattern for the proximal femur with both the proximal ulna and the distal humerus, but a unique morphological pattern given the shape of the distal femur with those same forelimb segments. KNM-ER 1500, attributed to Paranthropus boisei (Grausz et al, 1988), was included only in the analyses of the proximal ulna and distal femur and it was found to have a unique morphological pattern of with a human-like proximal ulna combined with an ape-like distal femur. Omo (Kibish) 1 (H. sapiens; McDougall and Brown, 2008) was excluded from the analyses using the proximal femur as that region is not preserved in the specimen. In the other analyses, it largely fell within the distribution of modern humans, the exception being for the proximal ulna where it is unlike any extant taxon. Finally, for the single analysis where KNM-ER 1503/1504 (Homo sp.) was used, it also had a human-like morphological pattern. This information is summarized in Table 7.
Table 7. Morphological pattern present for each fossil included in these analyses
Distal humerus/ Proximal femur
Distal humerus/ Distal femur
Proximal ulna/ Proximal femur
Proximal ulna/ Distal femur
“Human” indicates a morphological most similar to modern humans, “Ape” indicates that the morphological pattern is most similar to extant great apes, and “Unique” indicates that the morphological pattern is outside the range of the extant taxon, with those that have an asterisk being close to the human distribution.
A. afarensis (A.L. 288-1/A.L. 129-1)
P. boisei (KNM-ER 1500)
H. erectus (KNM-WT 15000)
Early H. sapiens (Omo [Kibish] 1)
Homo sp. indet. (KNM-ER 1503/1504)
Is the Morphological Pattern of Covariance in Forelimb and Hindlimb Elements Correlated With Intermembral Indices in the Extant Hominoids?
For the elements examined here, there is a strong correlation between shape covariance in the forelimb and hindlimb and intermembral indices (Table 5). Intermembral indices are often used to model locomotor capabilities in hominoids and fossil hominins (Jungers and Stern, 1983; Crompton et al., 1998; Kramer, 1999; Kramer and Eck, 2000; Wang et al., 2004) but these studies are generally limited to the few specimens that have a complete forelimb and hindlimb. Although shape covariance data do not allow for true biomechanical analyses, the correlation between these data and intermembral index indicates that they are a valuable proxy for inferring locomotor pattern in the less complete specimens commonly found in the fossil record. In all cases, multiple axes representing elements in the forelimb and hindlimb had higher correlation coefficients with intermembral index than a single axis, indicating that for the anatomical regions tested here, interlimb covariance is a better proxy for locomotion than any single element.
Table 5. Results from multiple regression analyses
For each analysis, the r2 and adjusted r2 ( ) are given to assess how well the PLS scores can predict intermembral index. The F-statistic and associated P value demonstrate the significance of the model. Below each analysis, the r2 for each individual axis and its significance to the model as a whole is recorded.
Distal humerus/proximal femur
Distal humerus axis 1
Distal humerus axis 2
Proximal femur axis 1
Proximal femur axis 2
Distal humerus/distal femur
Distal humerus axis 1
Distal humerus axis 2
Distal femur axis 1
Distal femur axis 2
Proximal ulna/proximal femur
Proximal ulna axis 1
Proximal ulna axis 2
Proximal femur axis 1
Proximal femur axis 2
Proximal ulna/distal femur
Proximal ulna axis 1
Proximal ulna axis 2
Distal femur axis 1
Distal femur axis 2
Is There Strong Shape Covariance Between the Forelimb and Hindlimb Bones in Extant Hominoids and, if so, What Are the Correlated Aspects of Shape in the Forelimb and Hindlimb Among the Extant Taxa?
Covariation between the proximal ulna and distal humerus in the forelimb with the proximal and distal femur in the hindlimb is higher than for other limb segments within extant hominoids. One possibility for this difference could be morphological bias. There is higher degree of overlap in the overall shape in the less complex proximal humerus, proximal radius, and proximal tibia. In contrast, the distal humerus, proximal ulna, and the proximal and distal femur are all relatively complex joints. In the more complicated joints of the elbow, hip, and femoral portion of the knee, there are many functionally important traits in that relate directly to their specific shapes. It is possible that selection is stronger on these regions in order to maintain functional integrity for the specific movements that are performed across them.
For instance, the size and projection of the olecranon process of the ulna are associated with maximizing the mechanical advantage of the triceps brachii muscle. A posteriorly oriented, retroflexed olecranon process will maximize the strength of the triceps brachii muscle during extended forearm postures in quadrupedal locomotion, whereas a more proximally located olecranon will increase the leverage of the triceps brachii in flexed forearm positions (Drapeau, 2004). Additionally, a deeper trochlear notch has been associated with stabilization of the elbow joint during hyperextension, and a greater degree of keeling has been linked with resistance to transverse forces across the elbow joint during suspension (Knussman, 1967; Drapeau, 2008).
In the distal humerus, the degree of symmetry present in the trochlea and the shape of the proximo-anterior border are related to stability or mobility at the elbow. Humans have a more asymmetric trochlea, which has the effect of reducing the stabilizing aspects of the elbow joint and maximizing muscular control of the forearm muscles while the forearm is in a flexed position, whereas the more symmetric trochlea seen in extant apes is related to weight support during extension and functionally separating the radial articulation from the ulnar articulation. The shape of the proximoanterior border of the trochlea is likely related the presence of a strong lateral trochlear ridge in the great apes, which functions to stabilize the ulna through flexion and extension (Aiello and Dean, 1990; Rose, 1993). The heights of the medial and lateral epicondyles are related to the muscle lever arms of the flexors and extensors of the digits. The higher position of the epicondyles in the apes when compared with humans is related to powerful hand musculature for use during locomotion (Senut and Tardieu, 1985).
In the analyses presented here, these features of the ulna and humerus covaried with the proximal femur. Of particular importance is the degree of lateral extension of the greater trochanter, which is indicative of the positioning of the gluteal muscles for bipedal versus quadrupedal locomotion, and the length of the femoral neck, which is associated with mobility in the hip (Aiello and Dean, 1990). In the distal femur, these same humeral and ulnar traits covaried with the degree of symmetry of the proximal border of the patellar articular surface, relative size of the medial and lateral condyles, and, for just the ulna, the angle between the patellar surface and the distal articular surface; these traits are related to stability at the knee joint in a biped versus more mobility at the knee joint in a quadruped. The increase in angulation between the two surfaces increases the lever arm for the quadriceps femoris, which is the main flexor of the knee, and orients the patella in a plane that is more anterior to a quadrupedal ape (Aiello and Dean, 1990). The deep, asymmetrical patellar groove could function to prevent dislocation of the patella as the quadriceps femoris pulls on it along the bicondylar angle (Heiple and Lovejoy, 1971; Stern and Susman, 1983).
The traits in the forelimb and hindlimb outlined above are part of the larger functional complex that is adapted for either quadrupedalism in the extant apes (i.e., posteriorly oriented and flexed olecranon, deep and keeled trochlear notch, symmetrical trochlea with a strong lateral ridge, more proximally located epicondyles, less laterally extended greater trochanter, lower femoral neck angle, symmetrical patellar articular surface, and greater angulation between the articular surfaces in the distal femur) or bipedalism and freeing of the hands in H. sapiens (i.e., a more proximally oriented olecranon, flatter and less keeled trochlear notch, asymmetrical trochlea, more distally located epicondyles, higher femoral neck angle, laterally flared greater trochanter, asymmetrical patellar surface, and a tighter angle between the two articular surfaces in the distal femur; see Table 6 for summary).
The second axes of each analysis comprised a much smaller—yet still significant—proportion of the shape covariance in this sample (Table 6). In the distal humerus, the size and position of the capitulum and the size and degree of lateral keeling in the trochlea were the most important aspects of shape on the second axis in all analyses and covaried with the depth of the trochanteric fossa in the proximal femur. This combination of traits best separates Gorilla and Pan from the others in this sample (Fig. 2). A distally facing capitulum and laterally keeled trochlea are correlated with stability at the joint between the humerus and radius (Rose, 1993). The trochanteric fossa is the insertion point for the lateral rotators of the hip (Aiello and Dean, 1990). The shallow trochanteric fossa in gorillas would shorten the lever arms of the rotators of the thigh allowing for more powerful rotating capability. The trochanteric fossa is particularly distinctive in chimpanzees as it is deep and very vertically oriented, and sometimes continuous with the medullary cavity; however, this is generally considered to be a trait unrelated to function (Gregory, 1950). The size and position of the capitulum and the size and degree of lateral keeling in the trochlea also covary with the distal articular shape in the distal femur, but the combination of these traits does not separate the extant taxa in any meaningful way (Fig. 3).
In the proximal ulna, the shape of the olecranon and the width of the trochlear notch covaried with the depth of the femoral neck and orientation of the lesser trochanter in the proximal femur. This combination of traits best differentiates Pan and, to some degree, Homo from the rest of the sample (Fig. 4). The widening of the trochlear notch in large bodied species has been suggested to be related to withstanding large loads on the center of the joint. Smaller taxa with a more varied locomotor repertoire (like Pan) have more waisted notches (Drapeau, 2008). The depth of the femoral neck is associated with rotational ability at the hip with taxa with deep necks having the greatest rotational ability and those with shallow necks the greatest stability (Aiello and Dean, 1990; Harmon, 2007). The position of the lesser trochanter is related to the insertion of the iliopsoas muscles, and changes in its orientation will change the mechanical advantage of these muscles (Jungers, 1977; Aiello and Dean, 1990; Anemone, 1993).
The shape of the olecranon and the width of the trochlear notch in the ulna also covaried with the shape of the distal articular surface of the femur. This suite of traits best differentiates fossil hominins—but particularly Omo (Kibish) 1—and Gorilla from the rest of the sample at one extreme and, to a lesser degree, Pan at the other extreme (Fig. 5). An anteriorly broad distal articular surface provides greater stability at larger body sizes with repetitive movements, as in Gorilla and Homo (Heiple and Lovejoy, 1971; Aiello and Dean, 1990; Lovejoy, 2007).
Overall, this information is useful in considering the sequence of morphological changes that led to the acquisition of bipedal locomotion. There is considerable evidence that the total morphological pattern in the forelimb and hindlimb of modern humans is a result of mosaic evolution and that there was considerable locomotor diversity in the hominin fossil record (e.g., Harcourt-Smith and Aiello, 2004; Green et al., 2007; Haeusler and McHenry, 2007; Young et al., 2010; Zipfel et al., 2011). Understanding which traits in the forelimb and hindlimb covary in the extant hominoids has implications for understanding which traits may have been selected as an adaptive suite of characteristics for a particular functional use, versus smaller scale changes that could have been acquired at other points in time as adaptations to satisfy some requirement or as the result of genetic drift.
Does the Pattern of Forelimb and Hindlimb Covariance Differ Among Associated Skeletons of Earlier Hominins and to What Extent Do Their Morphological Patterns Match Those of the Extant Taxa?
KNM-ER 1503/1504, Omo (Kibish) 1, and KNM-WT 15000 all had a basically human pattern for forelimb and hindlimb covariation. A.L. 288-1 was also human-like in all analyses except for those that used the distal femur. KNM-ER 1500 had a unique pattern of covariance with a proximal ulna with human-like characteristics with an ape-like distal femur (Table 7). Omo (Kibish) 1 is a modern H. sapiens (Brown and Fuller, 2008) and KNM-WT 15000 is reconstructed to have modern human limb proportions (Ruff and Walker, 1993); thus, a modern human pattern of covariation in the forelimb and hindlimb is expected for both of these specimens. It is interesting that these two specimens deviated, in some ways, more from the human pattern of covariance than earlier specimens, but both of these are somewhat exceptional. KNM-WT 15000 is a juvenile (Brown et al., 1985) and this could affect forelimb to hindlimb covariance. Omo (Kibish) 1 is a large, robust early anatomically modern human (AMH) (Brown and Fuller, 2008). Statistical differences between early AMH and current H. sapiens have been noted in previous cranial studies (e.g., Tubón et al., 1997; Pearson, 2008; Gunz et al., 2009) and are qualitatively more robust than current modern humans (Keith, 1925). This cranial robusticity is mirrored in the postcrania (Ruff et al., 1993) and often postulated to be the result of differences in activity patterns over time (Brose and Wolpoff, 1971; Stock, 2006). Thus, the unique pattern of covariance in the Omo (Kibish) 1 ulna could be related to statistical differences correlated with a more robust and muscular forearm.
The other three earlier specimens are somewhat more interesting to consider. A.L. 288-1 has consistently been found to have more ape-like limb proportions than seen in later hominins (Jungers, 1982). KNM-ER 1500 and KNM-ER 1503/1504 have been reconstructed to have intermediate body/joint proportions similar to A.L. 288-1 (McHenry, 1978; Grausz et al., 1988), but they have no complete long bones for direct comparison. Typically, the more ape-like body proportions of these individuals have been interpreted to mean that they practiced a less efficient form of bipedal locomotion (Berge, 1994) and (when interpreted in the light of other characteristics) maintained other adaptations for continuing arboreality (Stern and Susman, 1983, but see Lovejoy, 1988). The data presented here generally support those conclusions. A.L. 288-1 and KNM-ER 1500 had a unique pattern of covariance pairing a largely human-like elbow with a somewhat more ape-like knee.
The presence of an ape-like knee with a human-like forelimb seems contradictory on the surface; however, this is largely due to the assumption that humans evolved from an ancestor that was postcranially chimpanzee-like (Senut, 2003), an assumption that is based in part on the genetic similarity between extant chimpanzees and humans (Patterson et al., 2006). However, evidence is being accumulated that extant great apes are quite morphologically derived from the ancestral hominine condition just as humans are (Lovejoy et al., 2009; Harrison, 2010), and the directionality of the shape change in the postcranium from the last common ancestor to modern humans is not completely clear. Therefore, it is possible that the unique pattern of covariance present in these earliest hominins represents a largely symplesiomorphic condition; alternatively, it is possible that selection toward a bipedal morphology was stronger in the elbow and proximal femur earlier in the fossil record, and the retention of a more mobile knee was a compromise that allowed for more agility in continuing arboreal behaviors. This finding of unique covariation is not dissimilar from the findings in the foot of Australopithecus sediba, where many human-like traits were found in combination with an ape-like calcaneum (Zipfel et al., 2011). After all, there is more and more evidence that indicates that bipedal characteristics in early hominins could be homoplastic (Harcourt-Smith and Aiello, 2004; Zipfel et al., 2011).
Regardless, the similarity between the conclusions of limb proportion studies on A.L. 288-1 and conclusions here lend support to the idea that comparison of forelimb and hindlimb covariance could be used as a tool to bypass the necessity of estimating limb proportions when specimens are fragmentary (and therefore prone to error) in considering possible locomotor function.
Overall, all of the fossils tended toward having a human-like morphological pattern with no fossils falling solely within the distribution of extant great apes. It is possible that earlier specimens, particularly the recently published skeleton of Ardipithecus ramidus, would yield more ape-like results especially if it truly was a more generalized quadruped (Lovejoy et al., 2009).
When taken as a whole these results tentatively indicate that, compared to the extant sample, the fossils all had morphological patterns most similar to Homo, indicating that by the appearance of A. afarensis, a more modern human-like potential for division of use could already have been in place for the forelimb and hindlimb but with the retention of a slightly more flexible knee as a possible compromise for continuing arboreal ability. This would mean selection away from the forelimb as a supporting limb and a more flexible, arboreal hindlimb toward a more stable hindlimb and a forelimb adapted for a variety of uses. However, this does not indicate that these individuals were all perfect, human-like bipeds; just that the overall pattern of covariance for these elements was generally either more human-like or completely unique (as opposed to ape-like) when considered in its entirety.
The author thanks the following curators of the museums around the world for their help in accessing this material: Emma Mbua, National Museum of Kenya; Mamitu Yilma, National Museum of Ethiopia; Alan Morris, University of Cape Town; Malcolm Harmon, Powell-Cotton Museum; Louise Humphrey, Rob Kryzynsky, and Paula Jenkins, Natural History Museum London; W. Wendelen, Royal Museum of Central Africa; and Ian Tattersall, Eileen Westwig, and Gisselle Garcia, AMNH. The author also thanks Siobhán Cooke, Eric Delson, William Harcourt-Smith, William Jungers, F. James Rohlf, and the anonymous reviewers for their helpful comments in the preparation of this manuscript. This is NYCEP morphometrics contribution number 55.