The Appositional Articular Morphology of the Talo-crural Joint: The Influence of Substrate Use on Joint Shape



The appositional articular morphology of the talo-crural joint is the third component of the joint complex. It is a site of internal integration of this highly stable functional evolutionary unit. Prior studies of the other two components, tibia and talus, demonstrated that substrate preference influenced their articular shape. This effect was unrelated to physical attributes (size and mass) and phylogeny (superfamily). The effect of this behavioral factor, substrate preference, on shape and integration of the appositional articular morphology was investigated. Two hundred forty-five matched distal tibial and proximal talar landmarked surfaces from 12 diverse Catarrhine taxa were studied. Shape effects due to the same factors previously studied were examined in the tibial and talar subsets and were highly significant (P < 0.0001). These were assessed using Multivariate Regression and Relative Warps analysis, and Permutation tests, with results consistent with prior unmatched cohorts. Substrate preference influenced shape and was unrelated to the other factors across taxa. Singular Warp analysis of the cross-covariance matrix revealed sorting of taxa by substrate use, unrelated to physical attributes and phylogeny. Finally, the sorting demonstrated a signal of convergent evolution among distantly related taxa and divergent evolution among closely related taxa reflecting substrate use. Results were consistent with a behavioral influence, substrate use, affecting articular shape and integration in this highly stable functional evolutionary unit, and signals with evolutionary implications. Anat Rec, 297:618–629, 2014. © 2014 Wiley Periodicals, Inc.

The concepts of modularity and integration are critically important in evolutionary studies (Muller, 2005). The talo-crural joint provides a model for these processes as a functional unit that fulfills the criterion of an evolutionarily stable configuration (Schwenk, 2001). Prior examination of the bony components of the joint complex, the talus and tibia, identified a signal of substrate preference or use influencing the shape of the talar and tibial appositional articular surfaces (Turley et al., 2011; Turley and Frost, 2013). In this, the third in a series of studies exploring talo-crural joint shape, we evaluate the interaction of the appositional joint surfaces of the talo-crural joint to determine how they covary as a system, the relation of substrate to shape, and if evidence of convergent and divergent evolutionary change are manifest in that substrate-shape association (Turley and Frost, 2013).

The talo-crural joint or upper ankle joint has three osteological components, the tibia, the talus, and the appositional articular morphology. The third component of this joint complex is comprised of the subchondral surfaces of the articular interface encountered in fossil, archeological, and osteological collections (Harcourt-Smith et al., 2008). The presentation, orientation in space relative to the substrate and the organism proper, and the shape of the proximal (tibia) and distal (talus) components of the complex constrain appositional articular shape (Turley et al., 2011; Turley and Frost, 2013).

The ankle joint was addressed in the early literature in comparative anatomy, with interest in the origins of bipedalism in humans. Comparative studies by Morton, in multiple primate taxa Pan and Gorilla, and by Elftman and Manter in Pan demonstrated differences in their respective taxa in talar shape and angulation (Elftman and Manter, 1935; Morton, 1935). Studies were extended to the fossil assemblage and multivariate analysis applied to both extant primate tali and fossil specimens (Lisowski, 1967; Lisowski et al., 1974, 1976). Concurrently, Lewis expanded the understanding of the function and phylogenetic history of the ankle joint, describing the evolution of the talo-crural joint from arboreal marsupial structure and the implications of arboreal and terrestrial use for functional anatomy and joint shape (Lewis, 1980, 1989). Semiplantigrade, plantigrade, and bipedal ankle articular surfaces among extant and fossil taxa have been explored by Langdon in Miocene apes (1986), and Latimer et al. (1987) in African apes, humans and AL-288-1 examining ankle function in relation to use, and talar trochlear shape, distal tibial obliquity, and the location and implications for function of the talo-crural axis. Gebo (1992) has described the osteological structure in African apes shared with early hominids. Gebo and Sargis (1994) observed the relation of behavior to ankle shape among terrestrial, semiterrestial, and arboreal guenons, and Gebo and Schwartz (2006) demonstrated comparative talar shape among fossil hominids. All employed standard measurements and angles. Finally, Harcourt-Smith (2002) introduced the use of geometric morphometrics for examination of talus employing landmark data to compare the shape of great ape tali. Using these same methods the authors examined both the tibia and talus (Turley et al., 2011; Turley and Frost, 2013). Geometric morphometric methodology is also used in the current study to avoid the problems, associated with traditional measurements and angles (Harcourt-Smith, 2002; Turley et al., 2011). Whereas these previous studies examined the individual bones of the upper ankle joint, generally in their entirety, this study focuses just on the appositional articular morphology of both elements as a unified functional unit.

Prior studies of the tibia and talus have demonstrated the effect of multiple factors on the shape of their subchondral articular surfaces (Turley et al., 2011; Turley and Frost, 2013). However, substrate preference (a proxy for behavior), unrelated to several other factors such as size and mass (as a proxies for physical attributes, scale, and allometry) and superfamily (a proxy for phylogeny), affected the shape of the appositional articular surfaces of the talo-crural joint, the distal tibial and proximal talar facets (Turley et al., 2011; Turley and Frost, 2013). While these previous studies have examined distal tibial and proximal talar morphology independently, here the interaction of the appositional joint surfaces of these two bones is explored.

The subchondral surfaces were chosen since these are available in museum collections of the taxa examined, have been used in the prior studies suggesting the effect of substrate preference on shape, and avoid the issues of cartilage, tendon, synovial fluid and capsule, and soft tissue pathology, as well as, the difficulties of scanning living subjects. Although the importance of the soft tissue elements in joint function is recognized, the study of the appositional subchondral articular morphology can provide a proxy for the joint interface shape and provide insights into the response of underlying bone below the thin layer of cartilage to genetic and epigenetic influences (Hamrick, 1999a; Harcourt-Smith et al., 2008: Hammond et al., 2010).

The current study examined the subchondral surfaces of this articular interface using Singular Warp analysis of the cross-covariance matrix of matched talo-crural landmarked surfaces (Bookstein et al., 2003). When shape coordinates are employed singular warps have been used to assess “morphological integration” among anatomical structures (Bookstein et al., 2003; Bastir and Rosas, 2005; Mitteroecker and Gunz, 2009). Using this methodology, matching of such reciprocal surfaces has been possible (Harcourt-Smith et al., 2008; McNulty, 2009). Using matched specimens in integrated structures, such as, the talo-crural joint, a combined joint interface, a virtual plane of the opposing surfaces, may be studied as proposed by D'Arcy Thompson (Thompson, 1917). Beyond the separate analysis of its proximal and distal articular shapes, this approach allows an examination of how these two joint surfaces covary and how their shapes work together in the joint complex.

The objective of this study is first to evaluate the integration of this evolutionarily stable configuration, the talo-crural joint, by examining the covariation (correlation, r) of the tibial and talar articular interface using Singular Warp analysis (Schwenk, 2001; Bookstein et al., 2003; Bastir and Rosas, 2005). To this end, additional specimens and taxa have been included to improve the scope and expand the range of the study. Second, examine the relation of the tibial and talar subchondral articular shapes to the factors examined, including substrate. Finally, identify if signals of convergent evolution and divergent evolution are evidenced in the appositional articular morphology.



The study group consisted of matched distal tibial and proximal talar articular facets from single individuals. Two hundred forty-five specimens from 12 catarrhine taxa were included. All specimens were adults with M3 erupted, and all were without evidence of pathology. Nonhuman specimens had provenience documented, and all were wild-shot or in the case of Macaca mulatta from a free-ranging captive population. The non-human primate cohort included 195 specimens, 128 Hominoids, and 67 Cercopithecoids. The 50 human specimens were from six populations to provide diversity, California Paleoamericans (13), Southwestern Paleoamericans (6), Inuit (9), 4th Century Egyptian (7), 19th century European American (7), and 19th Century African American (8) (see Acknowledgements for institutional sources) (Table1).

Table 1. The number of specimens of the twelve taxa used in this analysis with their sex estimated mass and estimated substrate preference documented
TaxonNMFMass (kg) M/FSubstrate (0–10) M/F
  1. Mass values obtained from Smith and Jungers (1997) and Delson et al. (2000). Substrate preference obtained from Fleagle (1999) and Wells and Turnquist (2001).

Homo sapiens50262259.3/53.29.99/9.99
Pan troglodytes54262756.6/44.04/3
Gorilla spp.422511169.8/73.57/4
Pongo spp.105578.1/35.72/0.2
Hylobatidae12658.6/8.00.1/ 0.1
Macaca fascicularis13655.3/3.61/1
Macaca thibetana75115.2/9.58/8
Papio hamadryas125525.1/13.39/9
Nasalis larvatus116520.4/9.80.1/ 0.1
Macaca mulatta126614.0/12.06.4/6.4
Colobus guereza12749.0/7.50.1/0.1
Pan paniscus105545.0/33.20.5/0.5

The taxa were chosen from each of the superfamilies to provide a spectrum of size and substrate preference, with comparisons possible of both phylogenetically distantly and closely related taxa (Xing et al., 2007). The substrate preference estimates were coded from 0 (most arboreal) to 10 (most terrestrial), and were obtained from the literature along with taxon mean mass estimates of males and females for the taxa examined (Turley et al., 2011).

Data Collection

Each specimen was laser surface scanned and digitally reconstructed using either a Konica Minolta Vivid 910 Noncontact 3D Digitizer and Geomagic Studio 8 software, or a NextEngine Desktop 3D Scanner and ScanStudio HD software. Twenty-seven landmarks were placed by a single observer (KT): 12 on the distal tibial facets, and 15 on the proximal talar facets, using Landmark Editor software. These landmarks were illustrated in prior publications examining the shape of the tibia and talus (Turley et al., 2011, Fig. 1, Table 2; Turley and Frost, 2013, Fig. 1, Table 2).

Figure 1.

Scatter plot showing Singular Warp scores for individual talo-crural joints based on the Singular Warp of Procrustes aligned coordinates. Tibia 1 (TIB 1) is on the X-axis and talus 1 (TAL 1) on the Y-axis. Convex polygons are used to show the range of scatter with individual points hidden for clarity, and the mean values for each taxon recorded.

Table 2. Aspects of talar shape examined in each dataset
Distal tibial facets dataset
• Trochlear facet shape-oval/trapezoid
• Anterior trochlear facet margin (plane, lateral to medial)
• Posterior trochlear facet margin (plane, lateral to medial)
• Medial trochlear facet margin (trochlear medial malleolar groove) (axis posterior to anterior)
• Lateral trochlear facet margin (axis posterior to anterior)
• Trochlear-medial malleolar facets angle
• Central trochlear facet concavity (depth)
• Relative medial malleolar height (midline)
• Relative medial malleolar length (base)
• Medial malleolar base shape (concave to convex)
• Medial malleolar central shape (concave to convex)
• Medial malleolar apex shape (position, posterior–central–anterior)
Proximal talar facets dataset
• Trochlea shape
• Central groove relative depth
• Anterior trochlear margin (plane and lateral to medial shape)
• Posterior trochlear margin (plane and lateral to medial shape)
• Medial trochlear margin shape
• Lateral trochlear margin shape
• Medial crest angle, shape, and position of the apex
• Lateral crest angle, shape, and position of the apex

Observational error was previously assessed in both the tibial and talar subsets using single specimens of Homo sapiens landmarked 10 times (Turley et al., 2011; Turley and Frost, 2013). Principal component analysis demonstrated tight clustering of the repeated measures compared to variation within and among the taxa studied. This study, like the prior examinations of tibial and talar shape involves variation above the species level; thus, precision was deemed satisfactory.

Generalized Procrustes Analysis

Generalized Procrustes Analysis (GPA) was performed using Morpheus (Slice, 1998). GPA superimposes landmark configurations and removes variance due to position and rotation, and scales each to unit centroid size (Rohlf and Slice, 1990). Centroid size is the square root of the sum of the squared distances of each landmark to the centroid (Rohlf and Slice, 1990), and is stored as a separate variable during GPA. Centroid size was transformed to its natural logarithm for subsequent analysis. GPA was done with reflection allowed, since our dataset included both the right and left specimens. Separate GPAs were performed for the tibial and talar landmark subsets. All subsequent statistical analyses were performed using SAS 9.1 (SAS Institute, Cary, NC). As GPA aligned coordinates have a very high correspondence with their Euclidean tangent space projections, unprojected aligned coordinates were used. Shape differences among landmark configurations were measured by Procrustes chord distance; the Pythagorean distance between the two Procrustes superimposed landmark configurations (Bookstein, 1991; Slice, 2001; Turley et al., 2011).

Relative Warps Analysis

Relative warps analysis, principal component analysis (PCA) of shape coordinates, was performed on the covariance matrix of the GPA superimposed landmark coordinates, and used as a data reduction and exploration technique (Neff and Marcus, 1980; Bookstein, 1991). Here it is used with the components unweighted, although classically it weights the components by bending energy (Bookstein, 1991). Initial visualization of the effect on shape described by each Principal Component (PC) both within and among superfamilies was performed using Morphologika software (O'Higgins and Jones, 1998; O'Higgins, 2006).

Permutation Test

Differences in shape between individual taxa were assessed using pair-wise permutation tests with 100,000 replicates performed for each landmark subset. Individuals, with equal random samples drawn, were randomly permuted across the two taxa and Procrustes distance was calculated between the permuted groups' means. The α was the fraction of permuted values that were greater than the actual Procrustes distance between group means (Good, 2000). Results were reported with a Bonferoni-adjusted significance level of P < 0.0008 (Turley et al., 2011; Turley and Frost, 2013).

Descriptive examination of the studied taxa is compared to the prior studies of tibia and talus, and expanded to include Colobus guereza, not included in the prior talar study, and two new taxa Macaca mulatta, and Pan paniscus (Turley et al., 2011; Turley and Frost, 2013).

Regression Analysis

The relationship between shape coordinates and variables related to biological role was assessed using multivariate regression analysis (sensu Bock and Wahlert, 1965; Bookstein, 1996; Frost et al., 2003). Substrate preference, body mass, log centroid size, and superfamily (Hominoidea and Cercopithecoidea) were used as independent variables with GPA aligned coordinates the dependent variables.

The proportion of the total variance accounted for by each independent variable (mass, log centroid size, superfamily, and substrate preference) was calculated for each of the subsets by subtracting the residual variance after regression from the total variance and dividing the difference by the total variance (multivariate multiple regressions) (Frost et al. 2003). The sample size did not allow for a robust test of the interaction of all four variables together, or combinations of three. However, pair-wise analysis of the interactions among all four variables was possible. The angles among the shape vectors for each factor were calculated as the arccosine of their vector correlation (dot product) to evaluate the closeness of the relationship among the independent variables (Cobb and O'Higgins, 2004). Relatively parallel angulation (0°–30°) demonstrated a strong correlation, the closer to 0°, the tighter the relation, while relatively orthogonal angulation (60°–90°) demonstrated a weak relation, the closer to 90°, the less related (Cobb and O'Higgins, 2004; Turley et al., 2011).

Singular Warp Analysis

Singular warp analysis of the cross-covariance matrix of the appositional articular shape was used in this study to explore the relation of the matched surfaces of the distal tibial facets and proximal talar facets. This is the application of Partial Least Squares within morphometrics, which computes two unit vectors such that the covariance is a maximum. It describes how two sets of variables covary with each other. In this case, X variables (shape coordinates) from the tibia form one submatrix and Y variables (shape coordinates) from the talus form the other, that have the greatest mutual linear predictive power (Rohlf and Corti, 2000; Bookstein et al., 2003: Frost et al., 2003; Harcourt-Smith et al., 2008). The correlation r, a measure of integration, was computed for each (Bookstein et al., 2003; Bastir and Rosas, 2005). A scatter plot of the first Singular Warp of the cross-covariance submatrix of the appositional articular shapes was created with convex polygons used to show the range of scatter, but the individual points hidden for clarity, and the mean values for each taxon recorded. A scatter plot showing the male/female mean values of each taxon was created using the same technique. Finally, a scatter plot of the mean scores of the second Singular Warp was created (Bookstein et al., 2003).


Visualization of shape differences was accomplished by comparing landmark configurations directly in Morpheus and by warping an exemplar surface to fit those landmark configurations using Landmark Editor. Mean landmark configurations were computed for each taxon. Shape differences associated with substrate preference were visualized by adding the vector of regression coefficients from multivariate regression to the consensus landmark configuration for the landmark subsets. Features used in the resulting descriptions are explained in Table 2. The exemplar appositional articular surfaces from the singular warp analysis were likewise visualized using Landmark Editor. The singular vectors were shown as surface-morphs (Gunz and Harvati, 2007).


Relative Warps Analysis

The results of the relative warps analysis of the distal tibial and proximal talar matched subsets are provided, with eigenvalues of PC 1–5 and cumulative percentage of variance, in Table 3. These values are consistent with analysis of unmatched distal tibial and proximal talus data previously reported, with 57% of the distal tibial and 58% of the proximal talar variance accounted for by PC 1–5 (Turley et al., 2011; Turley and Frost, 2013). Shape differences generated from plots of these PCs likewise separated the studied taxa as in prior analysis (Turley et al., 2011; Turley and Frost, 2013).

Table 3. The first five eigenvalues of the covariance matrix on Procrustes aligned coordinates for principal component analyses on the two different datasets with the proportion and the cumulative values of the total variance provided
Distal tibia dataset
Proximal talus dataset

Regression Analysis

Multivariate regression analysis of the shape differences observed between four independent variables studied was again significant (P < 0.0001) in both the distal tibial and proximal talar subsets, and are presented in Table 4. Substrate preference accounted for a higher percentage (6.7% distal tibia and 9.7% proximal talus), with mass the least in each subset, as was the case in previous analyses (Turley et al., 2011; Turley and Frost, 2013). Table 5 presents the arccosine of the vector correlation (dot product) of the four variables. Substrate preference was unrelated to mass and superfamily, with size correlating with superfamily, the latter consistent with the larger size of most of the extant hominoids included in this study.

Table 4. Percent of total variance explained by the different factors within the proximal talar and distal tibial landmark subsets
AAM 3Substrate (%)Mass (%)Lcentroidsize (%)Superfamily (%)
Proximal talus9.
Distal tibia6.
Table 5. Angular differences (dot product) between the vectors for the different factors examined presented in degrees

The results generated from the examination of the regression coefficients of the taxon shapes associated with the matched appositional articular subcondral surfaces of the terrestrial and arboreal forms differed in the distal tibial trochlear facet and talar trochlea, with the former being a trapezoid in outline rather than an oval, with a flatter medial to lateral plane. The matched central convexity and concavity were greater and the crests were more acute with a more anteriorly displaced apex rather than obtusely angled and symmetrical. Compared to the terrestrial form, the medial malleolus of the arboreal form was longer and had a convex rather than concave base, curved rather than flat facet, and central rather than anterior apex. The talar facets were larger and concave, versus the flatter, smaller and wedge shaped terrestrial form. Superfamily forms differed with wider and shorter trochlea and trochlear facet dimensions in the hominoids. There was also a flatter plane with flatter crests with anterior displacement of the apex. Compared to hominoids, cercopithecoids had deeper central concavity and convexity, increase medial malleolar surface area relative to trochlear surface area with smaller medial and lateral facets with less concavity. Greater body mass was associated with a trapezoid trochlea and distal tibial trochlear facet with increased length and decreased width, while lesser mass was associated with more equal dimensions. The latter also had a flatter plane and central groove and convexity, square rather than acute lateral and obtuse medial crest angles, flat wide facets without distal wedge configurations. The medial malleolus in taxa with greater mass had greater height and width with less convexity and the apex anteriorly displaced than less massive taxa. Finally, larger forms (log centroid size) had equal trochlear and distal tibial trochlear facet dimensions, shallower central groove and lower convexity, greater plane, more acute crest angles, higher, distally displaced crests and grooves compared to smaller forms, and curved lateral and medial facets with central rather than distally displaced apices. The medial malleolus of the larger form had less convexity with apex anterior rather than central.

Permutation Test

The results of the permutation tests for differences in shape of the distal tibial and proximal talar subsets of the individual landmarked taxa are presented in Table 6. The great apes and humans were significantly different from each other. They were also significantly different from all other studied taxa excepting Pongo spp. from Hylobatidae, in the distal tibia. P. hamadryas was significantly different from all other taxa excepting the proximal talus of M. thibetana and M. mulatta in this study. Hylobatidae were not significantly different from M. fascicularis in the distal tibia and proximal talus, or from N. larvatus, C. guereza, or M. thibetana in the distal tibia, consistent with prior studies (Turley et al., 2011; Turley and Frost, 2013). Likewise, C. guereza and N. larvatus, N. larvatus and M. fascicularis, and M. fascicularis and both M. thibetana and M. mulatta, were not significantly different in either subset.

Table 6. Shape differences observed among taxa
  1. Procrustes distance between taxon means are shown above the diagonal and

  2. P values below the diagonal (100,000 permutations). P values <0.00001 correspond to 0.

  3. Legend: Hs, Homo sapiens; Pt, Pan troglodytes; Pp, Pan paniscus; Go, Gorilla spp.; Po, Pongo spp.; Hy, Hylobatidea; Ph, Papio hamadryas; Mt, Macaca thibetana; Mm, Macaca mulatta; Mf, Macaca fascicularis; Nl, Nasalis larvatus; Cg, Colobus guereza.


The mean shapes of the studied taxa, in this study matched as right or left articular complexes of distal tibial facets and proximal talar facets, were consistent with the findings in prior studies of the random distal tibial and proximal talar facets in Gorilla spp., H. sapiens, P. troglodytes, Pongo spp., Hylobytidae, P. hamadryas, M. fascicularis, M. thibetana, N. larvatus, and the distal tibia of C. guereza (Turley et al., 2011; Turley and Frost, 2013). The proximal talar facets of Colobus guereza, as well as, the mean shape of the tali and tibiae of the two additional taxa examined in the current study, Macaca mulatta and Pan paniscus are described.

Colobus guereza

The proximal talar facets showed a lesser plane to Nasalis larvatus. The trochlea, likewise, was more oval and wider with shorter less curved proximal margin than Nasalis larvatus. There was also a lower lateral crest but more acute angle. The central groove was shallower. The medial and lateral facets were comparable in size and shape to Nasalis larvatus.

Macaca mulatta

The proximal talar plane was decreased from that observed in Macaca fascicularis, as was the central groove depth and crest height. The crests were symmetrical but angles more acute. There was a smaller medial facet.

The plane of the distal tibial facets was decreased from M. fascicularis. The trochlea was wider centrally and narrower distally yielding a more oval lateral margin, but with straight proximal and distal margins and a convex medial groove. The central convexity was decreased, as was the trochlear fact medial malleolar angle. The medial malleolus was comparable but with a flatter base and facet, and more distally displaced apex.

Pan paniscus

The proximal talar facet plane was comparable to Pan troglodytes. The trochlea was more oval, and longer due to a longer central but shorter medial and lateral, and more curved, proximal margin. There was increased distal-medial extension toward the head. There was a shallower groove centrally but it was deeper distally and proximally. The crests were symmetrical with the lateral angle more acute. There was a deeper larger lateral facet with extension proximally.

The distal tibial facets plane was comparable. The trochlear facet was oval but with decreased proximal margin and distal medial and lateral widening. There was a more obtuse trochlear facet medial malleolar angle with a shorter but more convex medial malleolar base. The medial malleolus was smaller but with a central apex.

Singular Warp Analysis

The scatter plot of the scores from singular warp analysis of the cross-covariance matrix of the appositional articular shape of the talo-crural joints in the current sample was highly significant (P < 0.0001) (r = 0.68) and revealed sorting by taxa (Fig. 1). This sorting was made more evident by plotting the mean values of each taxon (r = 0.94) (Fig. 1). The sorting of taxa was consistent with substrate preference or more precisely substrate use, rather than, superfamily, mass or size, the other variables studied. Likewise, sexual dimorphism had little effect on the sorting, although behavioral differences (terrestrial patrolling among the males) among the sexes may have influenced results in the case of Pongo spp. (r = 0.88) (Fig. 2). The arboreal taxa clustered toward the origin including large bodied taxa such as N. larvatus, and P. paniscus, while the remaining taxa sorted by increasing terrestrial substrate preference. Exceptions to be addressed in the discussion included Pongo spp., which clustered with more terrestrial taxa, and M. mulatta,which in this study, clustered with the more arboreal taxa.

Figure 2.

Scatter plot using the same method as Figure 1 with the mean male and female values for each taxon recorded.

Figure 3 is a scatter plot (P < 0.0001) (r = 0.91) of the means of the second Singular Warp (tal2-tib2) with sorting of the African hominoids from cercopithecoids, except H. sapiens, which sorted with the cercopithecoids. Pongo spp. and Hylobatidae were on the fringe of the latter cluster. Of note, PC-2 of both the talus and tibia in prior studies was associated with the shape of the apposing trochlea facets, in particular, the anterior margin, and the shape of the medial talar crest and the curvilinear shape of the medial malleolar–trochlear groove. These shape changes are consistent with the articular morphology associated vertical climbing (Turley et al., 2011; Turley and Frost, 2013). The sorting of the second Singular Warp was again consistent with a behavioral effect, substrate use.

Figure 3.

Scatter plot of the second Singular Warp scores of the M3 mean values for each taxon recorded.


The talo-crural subchondral surfaces were visualized using Landmark Editor and an exemplar surface (Pan troglodytes) was warped to fit the estimated configurations. The shapes related to the arboreal and terrestrial forms are illustrated (Fig. 4). Likewise, the surface (Pan troglodytes) was warped to fit the estimated configurations of the Singular Warp vectors and a surface-morph constructed (Fig. 5).

Figure 4.

Visualization of the appositional articular morphology, the distal tibial and proximal talar surfaces, of the substrate preference variable with the terrestrial and arboreal shapes presented. See Materials and Methods section for details.

Figure 5.

Visualizations of the shape change of the Singular Warp of tibia 1 and talus 1 from negative to positive on the X and Y axes, respectively. See Materials and Methods section for details.

A visualization of the shape differences represented by the second Singular Warp is shown in Figure 6, as well as, a visualization of a composite African ape morphed surface from the relative warps data, and a picture of a representative Gorilla spp. surfaces. The similarity of the latter with the visualization of the negative pole of the distribution of the second Singular Warp scores is observed. These include a curvilinear medial extension of the tibial groove about the base of the medial malleolus with superio-medial displacement of this groove, as well as, the anterior tibial trochlear facet margin, and anterior displacement of the medial talar trochlear crest, anterior and medial extension of the crest and medial displacement of the anterior medial facet. This wedge shaped medio-superior appositional articular shape stabilizes and locks the articular surface in dorsiflection and medial rotation optimizing safe vertical climbing in large bodied hominoids (DeSilva, 2009; Turley et al., 2011; Turley and Frost, 2013).

Figure 6.

Visualizations of the shape change of the second Singular Warp of the M3 sample from negative to positive on the X and Y axes, respectively. Relative Warps of the African apes (Gorilla spp., and Pan spp.) appositional articular morphology morphed to an exemplar surface. Visualizations of the of a representative M3 Gorilla spp. Laser scan digital reconstruction.


The talo-crural joint was chosen to examine evidence of homology, homoplasy (convergent evolution), and morphological differentiation (divergent evolution) among catarrhine taxa due to qualities of both its structure and development. It is a well-defined component of the primate Bauplan, and a conserved and canalized character trait. It is consistent with the designation of a functional unit within the concept of biological role (sensu Bock and von Wahlert, 1965). Further, it has the characteristics of “evolutionary stable units” as proposed by Wagner and Schwenk (2000), character groups which are structural and mechanical, are a composite of morphological components (tibial, talus, and appositional articular shape), and demonstrate internal integration modulating external pressure to change, but providing a mechanism for reorganization of the phenotype in response to such pressures (Schwenk, 2001). During development it forms the interface of two of the modules of the hindlimb, the zeugopodium proximally, and the autopodium distally (Cachel, 2006). It is the region of greatest developmental modularity evident by the sequencing of ossification centers in the postcranium (Blomquist, 2009). Finally, the appositional articular surface shapes reflect integration, the complex pattern of structural correlation and covariance at the joint interface (Klingenberg, 2008).

This study examined integration of the talo-crural joint by utilizing both singular warp analysis of the matched appositional articular subchondral bone, as well as the more customary relative warps analysis and multivariate regression analysis of each of the matched elements, the distal tibia and proximal talus (Bookstein et al., 2003; Mitteroecker and Gunz, 2009). Using singular warp analysis, we were able to explore information unavailable to our prior data analysis (Turley et al., 2011; Turley and Frost, 2013). It provided data about the relation of the matched elements, their covariance, allowing inferences as to the relationship of the shapes, and how they work as a coordinated system (sensu Bock and von Wahlert, 1965). The way they are used is influenced by their interrelation, which expresses how the system is integrated; the “intrinsic functional integration” observed by Schwenk (2001) in “evolutionarily stable configurations.”

The factors studied functioned as proxy measures to identify signals of the relative importance and interrelationship of the main categories examined: superfamily for phylogeny, mass, and size for absolute scale of the studied taxa, and substrate preference for the behavioral interface of the taxa with the environment (Turley et al., 2011; Turley and Frost, 2013).

The articular surfaces of the distal tibia and proximal talus were first analyzed separately, since the matched cohort of the current study had additional taxa and a change in sample size from the prior studies (Turley et al., 2011; Turley and Frost, 2013). All of the above factors had a significant influence on shape in both the distal tibial and proximal talar subsets (P < 0.0001). Substrate preference and superfamily, as in the prior unmatched studies were major factors, but substrate preference was unrelated to both proxies for phylogeny and scale (Turley et al., 2011; Turley and Frost, 2013).

Each of the ape taxa was significantly different from each other, with the exception of Pongo spp. and Hylobatidae, in the permutation test of the distal tibial data set (P < 0.0027). The great apes were significantly different from all other taxa. Of note, this despite problems of the hominoid samples currently available for study in museum collections.

P. hamadryas, likewise, was significantly different from most other taxa excepting in the proximal talus of M. fascicularis and M. thibetana. Hylobatidae were not significantly different from M. fascicularis in the proximal talus, and were significantly different from all cercopithecoids, excepting M. mulatta, in the distal tibia. Among the other cercopithecoids, significant differences were observed in shape of M. thibetana and both M. mulatta and N. larvatus, M. mulatta and both colobines, and M. fascicularis and C. guereza in the proximal talus. These findings were consistent with the highly conserved nature of the two components of the joint complex, but demonstrated the differences evident between superfamilies.

The articular surfaces were then analyzed together using singular wrap analysis. The singular warps can provide important insights into the appositional articular morphology of the studied taxa not evident in the individual analysis. Since the cross-covariance matrix has only the covariances of the matched distal tibial and proximal talar landmark sets without the within set covariances, morphological integration of the appositional surfaces can be assessed. Figure 1 demonstrates the first Singular Warp summarizing the cross-covariation between the two subsets sets and integration (mean, r = 0.94) of the two components of the talo-crural module. The mean values of the studied taxa sorted by the behavioral proxy, substrate preference, rather than phylogeny or scale (Fig. 1) or sex (Fig. 2). Sorting was not evident in the relative warps of the tibial and talar sets. However, it should be emphasized that this does not appear to be a size effect with the terrestrial cercopithecoids, M. thibetana and P. hamadryas sorting with H. sapiens, Gorilla spp., P. troglodytes and Pongo spp., while P. paniscus, comparably sized to P. troglodytes, and M. mulatta, N. larvatus and C. guereza comparatively sized to M. thibetana and P. hamadryas sorting with the smaller bodied Hylobatidae, and M. fascicularis.

Several taxa had inconsistencies between substrate preference documented from the literature and the sorting observed (Table 1). The substrate preference of Pongo spp. is highly arboreal (with females even more so than males), however, its positional behavior (males more so than females) has been stated by Thorpe and Crompton to be most similar to lowland gorillas, and the stresses induced by squatting and arboreal walking suggest the influence of use on the sorting, as well as, a behavioral effect influencing the male cohort greater than the female in Figure 2 (Thorpe and Crompton, 2006; Thorpe et al., 2007; Young, 2008). All specimens of M. mulatta were from the free ranging colony at Cayo Santiago, Puerto Rico. Wells and Turnquist (2001) have documented the terrestrial substrate preference (here scored as 6.4) of the adult population, as well as, a lesser (4) juvenile preference and the nature of the island habitat. Greater terrestrial use has been suggested in native Asian populations, and the nature of the substrate, both terrestrial and arboreal on Cayo Santiago versus Asian habitats may influence the findings of the current study (Lindberg, 1980; Goldstein and Richard, 1989; Fleagle, 1999). Of note, the permutation tests of tibial and talar elements demonstrated significant differences between the M. mulatta and M. thibetana specimens in the current study. Finally, the difference observed between P. paniscus and P. troglodytes are consistent with terrestrial use of only 0.5 in P. paniscus observed by White (1992). This may reflect a greater difference in substrate preference than previously documented due to terrestrial provisioning in some prior studies in the P. paniscus, the fact that the specimens sampled, were from the Royal Museum of Central Africa wild shot during the Belgian Colonial period prior provisioning at primate study sites, and prior to the greater appreciation of terrestrial use (patrolling, etc.) in P. troglodytes (Doran, 1992, 1993, 1997).

The singular warp scores also provided potential signals of convergent and divergent evolution. Phylogenetically distant, more terrestrial taxa, such as, H. sapiens, M. thibetana, Gorilla spp. and P. hamadryas, sorted together. Likewise, both large bodied arboreal taxa, such as, P. paniscus, and N. larvatus, and small bodied arboreal taxa, such as, M. fascicularis and Hylobatidae sorted together consistent with homoplasy. Conversely, closely related taxa, such as, P. troglodytes and P. paniscus, as well as, M. thibetana and M. fascicularis, sorted away from each other suggesting morphological differentiation within these genera.

When the second Singular Warp was examined clustering of Gorilla spp., Pan troglodytes, and Pan paniscus scores was observed. The differences in shape identified in the submatrix were consistent with the superior displacement of the antero-medial margin of the tibial trochlear facet, and the shape of the medial talar crest, as well as, the curvilinear shape of the medial malleolar base and medial malleolar-trochlear groove observed in these taxa in prior studies (Turley et al., 2011; Turley and Frost, 2013). All are critical in locking the joint in dorsiflexion and medial rotation, providing safety and stability when these large bodied hominoids employ their technique of vertical climbing.

The etiology of the shape differences observed may be related to both intrinsic and extrinsic influences, namely, the nature of the joint complex itself or its environmental interface. The talo-crural joint is a highly modular and integrated structure consistent with an evolutionarily stable configuration, that is, a functional unit with a biological role sensu Bock and von Wahlert (Schwenk, 2001). It is a complex unit consisting of multiple elements including the synovial fluid and sac, tendons and ligaments, and the cartilaginous surfaces with underlying subchondral bone, and appositional articular surfaces (Sarrafian and Kelikian, 2011). The interrelation of these elements during development and in the adult organism is reflected in the end point morphology. The development of the joint involves progenitor cells responsible for all its elements (Koyoma et al., 2008). The formation of articular cartilage and subchondral bone involves cartilage resorption to a single postnatal articular layer in the former and modeling of the underlying bone by the deep chondrocyte layer (Hunziker et al., 2007; Stempel et al., 2011). Environmental variances have been observed to increase distally in the mammalian limb (Hallgrimsson et al., 2002). Further, the constraints due to covariation have been proposed to be “weak enough” that differences in development of distal components may occur (Young and Hallgrimsson, 2005). Differences in phenotypes reflect genetic variation and changes in genetic networks are a mechanism proposed for change (Carroll, 2008). Evolutionary developmental theory suggests that these may be due to natural selection for “deep homology”, commonality of genetic developmental processes among distantly related organisms (Shubin et al., 2009; Hall, 2012). They also may result from adaptive developmental plasticity manifest by either modification in gene expression or evolving reactivity to systems responsive to external signals (West-Eberhard, 2003, 2005; Beldade et al., 2011). Conversely, they can result from plasticity, an epigenetic response of bone to strain, producing permanent changes in shape from a behavioral difference altering the mechanical loads experienced (Cobb and O'Higgins, 2004; Singleton, 2012).

This begs the question of whether the phenotypic differences observed among the catarrhine taxa in this study were the result of natural selection or plasticity. Bone morphogenic proteins have been shown to be involved in both joint development and bone remodeling in response to external stimuli (Francis-West et al., 1999; Young and Badyaev, 2007). They are, therefore, likely to be important in both the way natural selection can shape a joints predisposed developmental pathway and in the way that bone remodels. In terms of remodeling, a chondral response to different loads has also been proposed to shape joints (Frost, 1999). Hydrostatic pressure differences with development and changes in locomotor and postural behavioral activity are also a purported mechanism producing a mechanically induced change in joint shape (Hamrick, 1999a,b). Finally, remodeling of subchondral bone to load and strain has been demonstrated (Hammond et al., 2010).

Thus, the mechanism by which the shape differences observed in the sorting of the singular warps cross-covariance matrices occur remains to be elucidated. However, the developmental timing of the observed differences among the taxa used in this study can provide a lens to separate genetic from epigenetic (Frelat and Mittereocker, 2011). The study of the ontogeny of shape can be used to compare taxa as subadults and adults to determine if sorting by substrate use is consistent throughout (genetic) or changes with behavior during development (epigenetic).


The authors are grateful to Will Harcourt-Smith and Eric Delson of the Department of Vertebrate Paleontology, AMNH for their assistance with many aspects of this project, and for providing some of the humans and great apes scans from the AMNH. They thank Eileen Westwig and Gisselle Garcia, AMNH, Yohannes Haile-Selassie and Lyman Jellema, CMNH, Terry Kensler, CPRC, Judith Chupasko, HMCZ, Tracy Damitz, FMNH, Linda K. Gordon NMNH, Angela Gill, Powell-Cotton Museum, Emmanuel Gilissen and Wim Wendelen, RMCA, and Natasha Johnson of the P.A. Hearst Museum Department of Anthropology, UC Berkeley, for access to their collections, as well as, Tim White and Mike Black for help with the Hearst Museum. They also thank the reviewers of our manuscript, Kieran P. McNulty and an anonymous reviewer, for their helpful comments improving its quality.