The frequencies of various positional behaviors and the precise hand kinematics during locomotion, especially in the wild, among primate taxa are not well characterized and need further research. However, based on major differences in hand kinematics, we make several predictions of the regions of the metacarpo-phalangeal (MP) joint that habitually experience the greatest joint reaction forces (Tsegai et al., 2013 and Zeininger et al., 2011).

Prediction 1. During knuckle-walking, chimpanzees use their digits in a flexed position to support body weight while the metacarpals are extended at the metacarpo-phalangeal joints (Jenkins and Fleagle, 1975 and Tuttle, 1967) and the dorsal surface of the intermediate phalanges make contact with the substrate (Wunderlich and Jungers, 2009). Joint reaction forces are transmitted to the dorsal region of the metacarpal head during knuckle-walking (Fig. 1a), being highest in the 3rd and 4th metacarpals which absorb high compressive forces (Wunderlich and Jungers, 2009). This increased loading dorsally is thought to explain the mediolaterally expanded dorsal articular surface of the metacarpal head as a way to mediate joint reaction forces (Whitehead, 1993). Furthermore, chimpanzee third metacarpals have a dorsal ridge related to the degree of loading at the MP joint (Inouye, 1990). In addition to knuckle-walking, during arboreal locomotion, chimpanzees employ the suspensory transverse hook grip and the suspensory diagonal hook grip (Marzke and Wullstein, 1996), which involves variable flexion of MP joints (Fig. 1b). In this position, the palmar and palmar-distal surfaces of the metacarpal head experience the greatest joint loads. Thus, we predict that in chimpanzees, BV/TV and DA will be relatively similar in the dorsal and palmar regions of the third metacarpal head, both of which are predicted to be greater than in the central region due to greater loading during knuckle-walking and climbing.

Prediction 2. Orangutans engage in a considerable amount of suspensory postures and locomotion, and they rarely engage in behaviors that involve extended MP joints (e.g., Cant, 1985, Thorpe and Crompton, 2006, Tuttle, 1967 and Whitehead, 1993). They employ various hook grips and flexed finger postures during suspension and climbing. As the digits support body weight the palmar and central regions of the metacarpal head are thought to experience high joint forces (Fig. 1c). Therefore, we predict that in orangutans, the palmar and central regions will have relatively high BV/TV and DA compared to the dorsal region.

Prediction 3. Baboons walk on the ground using a digitigrade hand posture, with body weight supported by the metacarpal heads and phalanges (Hildebrand, 1988, Patel, 2010 and Whitehead, 1993). During digitigrady, the distal region of the metacarpal head experiences substrate reaction forces while the dorsal region of the metacarpal head experiences MP joint reaction forces (Fig. 1d). However, at faster speeds, baboons adopt more palmigrade hand postures (Patel, 2010 and Patel and Wunderlich, 2010); as the hands shift from digitigrade to palmigrade, the MP joint reaction force moves from the dorsal to the central region, and the substrate reaction force moves from central to palmar. Joint reaction forces at the MP joint are typically much higher than the substrate reaction forces (Cooney and Chao, 1977, Preuschoft, 1973 and Richmond, 2007). Consequently, in baboons, we predict that there will be greater BV/TV and DA in the central and dorsal regions compared to the palmar regions of the metacarpal head.

Prediction 4. Humans use their hands in a variety of non-stereotypical ways, with actions ranging from very fine finger manipulations (e.g., precision grip) such as knitting to more generalized use such as holding tools (e.g., power grip; Fig. 1b). Some of these hand postures involve MP joint flexion while others do not (Marzke, 1997 and Napier, 1993). Therefore, we predict that compared to all other taxa, humans will show the lowest DA because of generally non-stereotypical hand use, and will have relatively similar and lower BV/TV throughout all regions than seen in other primates as has been reported elsewhere (Chirchir et al., 2015, Tsegai et al., 2013 and Zeininger et al., 2011).

We investigated a sample of 39 third metacarpal heads of four higher primate taxa: Homo sapiens (n = 10), Pan troglodytes (n = 10), Pongo pygmaeus (n = 10), and Papio anubis (n = 9) (Table 1). The non-human samples are wild collected and the modern humans were drawn from the Terry collection (late 19th century and early 20th century working class Americans). All specimens are housed at the National Museum of Natural History (NMNH), Smithsonian Institution, Washington. Following previous studies (Tsegai et al., 2013 and Zeininger et al., 2011), third metacarpals were selected because their central position in the hand buffers them from mediolaterally-directed forces and consistently involves them in bearing weight during locomotion (e.g., Patel and Wunderlich, 2010 and Wunderlich and Jungers, 2009). The sample consists of adult male and female individuals. However, sexually variable factors that may impact bone growth and development are not considered in this study and consequently data from both sexes are pooled. Specimens were scanned at two facilities. Metacarpals of three chimpanzees and three orangutans were scanned at the University of Texas at Austin's High-resolution X-ray CT (HRXCT) facility at 25 μm resolution while the remainder of the samples were scanned at the NMNH using a Stratec XCT Research SA CT (pQCT) scanner, at a 50-μm resolution. Although the two instruments are not the same, both were scanned at relatively high resolution and previous research using different scanning equipment has shown a good relationship between different scanning equipment (e.g., Chirchir et al., 2015). Lazenby et al. (2008) reported bilateral asymmetry in trabecular BV/TV in the hand. Here, we selected bones from the right hand when available (P. troglodytes all right side, H. sapiens; lefts = 1 and rights = 9; P. pygmaeus; lefts = 2, rights = 8; P. anubis; lefts = 1, rights = 8).

The program Quant3D (Ryan and Ketcham, 2002) was used to analyze trabecular bone structure. This program quantifies 3D properties of interest, including BV/TV and DA. The program allows the user to specify the size and location of a spherical volume of interest (VOI) within which trabecular properties are quantified.

In order to compare trabecular properties in the metacarpal head, we partitioned it into three regions: dorsal, central, and palmar VOIs (Fig. 2). We used spherical VOIs because the corners of cubic VOIs may bias the measurement of directionality (Ketcham and Ryan, 2004). VOIs were placed along the midline of the dorso-palmar axis of the joint surface. Utmost care was taken to avoid sampling the surrounding cortical bone at the joint surface. We selected a VOI size that was as large as possible without intersecting any cortical bone in any of the three regions in any of the samples. VOI size was determined to be the largest diameter that could fit within all three regions in a given bone. VOI size remained constant across regions within a specimen, so the smallest region determined the maximum size limit. While there are legitimate questions concerning how to best scale VOIs in studies comparing different species (Fajardo and Müller, 2001, Griffin et al., 2010a, Kivell et al., 2011 and Lazenby et al., 2011), these issues have little impact in this study because the main comparisons are made within, rather than between, species.

Although our regional sampling strategy resulted in some overlap in VOIs, the spherical shape of the VOIs restricted overlap to very small proportions of the total volume. Slight overlap is preferable to making the VOIs so small that they do not adequately represent the structure of the joint regions. Within each VOI, bone was segmented using the FWHM (Full width/half maximum) threshold method (Trussell, 1979).

DA was measured using star volume distribution, which measures trabecular bone orientation by repeatedly sampling bone within the VOI and recording the lengths in all directions from the selected points within the bone to the bone-air interface (Ryan and Ketcham, 2002 and Ryan and Krovitz, 2006). This was measured 2049 times, at 2000 random points as programmed in Quant3D. Eigenvalues were calculated based on the resulting orientation distributions. DA is calculated as the ratio of the primary eigenvalue over the tertiary eigenvalue, as a measure of the strength of orientation in the primary direction of the strength in the weakest perpendicular plane (Ryan and Ketcham, 2002). We performed Kruskal–Wallis tests with Nemenyi tests using R v.3.2.4 (2016) to test for significance across VOIs for each trabecular bone property within a taxon.

Means and standard deviations of BV/TV and DA in each taxon and VOI are reported in Table 2. Within chimpanzee third metacarpal heads, although the central VOI exhibited the greatest mean BV/TV (Table 2), none of the VOIs were significantly different from one another (Table 3, Fig. 3a). Similarly, DA also did not differ significantly across the three VOIs in chimpanzees (Table 3, Fig. 3b). Orangutan third metacarpal heads exhibited significantly greater BV/TV in the palmar VOI than in the dorsal VOI, and significantly greater BV/TV in the central than the dorsal VOI (P < 0.05; Table 3, Fig. 4a). DA in the orangutan third metacarpal heads did not differ significantly among regions (Table 3, Fig. 4b).

In the baboon metacarpal heads, the observed BV/TV was not significantly different across the three VOIs (Fig. 5a, Table 3). The pattern was somewhat similar to the chimpanzee BV/TV (Fig. 5a), with the greatest mean BV/TV in the central VOI but not significantly different from the other VOIs. Unlike BV/TV, the DA pattern in the baboon sample was different from the chimpanzee pattern. Although high DA values are observed in the dorsal region (Fig. 5b), the differences across the VOIs were not statistically significant (Table 3).

Lastly, in humans, the dorsal VOI had significantly lower BV/TV than the central VOI but not the palmar (P < 0.05; Table 3, Fig. 6a). There was no significant difference in DA across the three VOIs (Fig. 6b, Table 3). In comparison to other taxa, BV/TV was significantly lower in all VOIs compared with other taxa (Table 4), whereas DA was not significantly lower than other taxa except central VOI of P. anubis (P < 0.05).

The results show that the distribution of trabecular bone volume (BV/TV) matches some of the predictions based on species-specific differences in habitual hand use, whereas the distribution of degree of anisotropy (DA) does not. The findings here show that habitual biomechanical loading has an influence on trabecular morphology consistent with reports by other researchers (e.g., Griffin et al., 2010a, Pontzer et al., 2006, Rafferty, 1998, Ryan and Ketcham, 2002 and Tsegai et al., 2013), albeit not in both trabecular properties investigated.

The first prediction that chimpanzees have relatively similar BV/TV and DA in both palmar and dorsal regions of the third metacarpal head is supported by our findings. Specifically, there is no significant difference in either BV/TV or DA between the two regions in chimpanzees (Table 3). However, contrary to predictions, the central region is not significantly lower in BV/TV than the dorsal and palmar regions (Fig. 3a). This might be explained by the diversity in hand postures and positional behaviors among chimpanzees, which result in loading at the MP joint from a variety of postures. Specifically, knuckle-walking and flexed-joint hand postures during climbing could explain the high DA values in the dorsal and palmar VOIs, respectively. Perhaps the BV/TV in the central region remains high to keep the entire metacarpal head strong through the range of regularly used hand postures, including partly-flexed MP joints used in a variety of climbing grips (e.g., Fig. 1b and c). Chimpanzees engage in a variety of postures and locomotor behaviors in the trees (e.g., Carlson et al., 2006, Doran, 1993a and Doran, 1993b). Predicting joint reaction forces distribution in the chimpanzee hand is complicated by factors, such as age changes in locomotion, body mass, and type and size of substrate (Doran, 1993a, Doran, 1993b and Doran, 1997). However, chimpanzees spend ∼ 84% of the time on the ground; during knuckle-walking, the MP joints are stereotypically loaded in extended postures and the second and third digits experience the greatest stresses (Wunderlich and Jungers, 2009). If trabecular bone is sensitive to differences in loading then, despite the highly flexible nature of hand use during locomotion, habitual behaviors as different as knuckle-walking and suspension should be detectable. If not, then it raises serious concerns about using trabecular structure to make behavioral inferences in the fossil record.

Zeininger et al. (2011) show that patterns of bone mineral density better match the hypothesis that knuckle-walking and climbing concentrate mechanically demanding loads on the dorsal and palmar regions. However, in that study (Zeininger et al., 2011), the lower bone mineral density (associated with elevated remodeling levels) were most pronounced in subchondral bone towards the margins of the joint. Similarly, Tsegai et al. (2013) report that the BV/TV values are especially high in the dorsal-most region and dorsal ridge. The VOIs selected here may not sample quite the same region (e.g., spherical VOIs cannot sample much trabecular bone very close to the joint surface).

Tsegai et al. (2013) also report that, compared to the dorsal region, BV/TV values appear to be lower in the palmar regions of their African ape samples, based on visual inspection of scaled color maps. Our results contrast slightly, in that the palmar VOIs in the chimpanzee sample are equivalent to the dorsal VOIs in BV/TV (Table 3, Fig. 3). Similarly, our results show equivalent DA values in those regions (i.e. the dorsal and palmar VOIs). The visualization maps in Tsegai et al. (2013) suggest a pattern of somewhat lower values in these regions. It would be valuable to follow up our and Tsegai et al.’s (2013) study with other strategies to quantify and statistically examine the distributions of trabecular characteristics. Our understanding of the link between trabecular morphology and behavior could be greatly improved with the development of more thorough quantitative strategies coupled with much more thorough documentation of behavior, such as hand postures used in natural settings that should be a high priority for future research (Richmond et al., 2016).

The second prediction that orangutans will have greater BV/TV and a high DA in the palmar and central regions compared to the dorsal region is partially supported by the results – the BV/TV values match expectations whereas the DA results do not. We found that there is significantly greater BV/TV in the palmar and central VOIs compared to the dorsal VOI (Fig. 4a, Table 2 and Table 3). These results are consistent with the BV//TV color maps and stiffness maps reported in a different sample of orangutan third metacarpals (Tsegai et al., 2013), and with data showing lower bone mineral density in the palmar-distal regions (Zeininger et al., 2011). These results support the hypothesis that the dorsal portion of the orangutan metacarpal is not typically subjected to joint forces, as extended MP joint postures are not usually used during suspension and climbing. Suspensory taxa instead use highly flexed fingers when grasping relatively small supports, and neutral MP joint postures during hook grips (Cant, 1985 and Sarmiento, 1988; Fig. 1c). Here, trabecular BV/TV shows a clear pattern related to habitual use.

Orangutan third metacarpal DA on the other hand, did not follow the predicted pattern. This demonstrates the complexity of interpreting DA. Experimental research finds no differences in DA between running and control mice in the hindlimb (e.g., Carlson et al., 2008). Additionally, Matarazzo (2015) found that when individual trabecular bone properties are examined in phalanges and metacarpal heads of primates, they do not fully reflect locomotor behavior. Thus, we suggest that the bone in the orangutan dorsal metacarpal head remains anisotropic with struts oriented towards the joint surface in order to resist the forces that occur, however infrequently.

Thirdly, the prediction that baboons will have the greatest BV/TV and DA on the dorsal and central regions is not well supported by our results because there were no significant differences across VOIs in either BV/TV or DA (Tables 3, Fig. 5a and b). The lack of difference suggests that the MP joint posture is not sufficiently stereotypical to leave a distinct trabecular pattern. For example, in human feet, the metatarsal heads have a striking pattern of increasing BV/TV and DA from plantar to dorsal that contrasts with the patterns in great ape metatarsal heads (Griffin et al., 2010a). Humans, of course, use their feet in far more stereotypical ways than do great apes, and humans walk and run with metatarsophalangeal joints that are more highly extended than those of bonobos, for example (Griffin et al., 2010b). Baboons use extended MP joints on the ground but shift to palmigrade postures with neutral MP joints at higher speeds (Patel, 2010). This, in combination with varying amounts of tree-climbing activities, such as sleeping and feeding (Altmann, 1974), which may explain the fairly consistent distribution of bone throughout the third metacarpal head.

Lastly, we predicted that modern humans would have low and relatively uniform BV/TV and low DA in all regions due to the non-stereotypical hand use and relatively low (non-locomotor) levels of force. Our results partially support this prediction. First, BV/TV in the dorsal VOI was significantly lower than the central but not palmar VOI (Table 3, Fig. 6a). Like orangutans, human hand functions typically involve little hyperextension at the MP joints. Human MP joints are used in neutral and flexed positions during carrying and manual manipulation (Goldfarb and Dovan, 2006, Marzke, 1997 and Napier, 1993). In all regions, BV/TV was very low compared with the other taxa (Table 4, Fig. 6a), adding further support to research showing that low bone volume throughout the skeleton appeared very recently in human evolution (Chirchir et al., 2015), likely after abandoning a hunter-gatherer lifestyle (Ryan and Shaw, 2015). The low BV/TV throughout the third metacarpal head observed in this study are consistent with Tsegai et al.’s (2013) findings as well. Human DA was not significantly different from most taxa, suggesting that even with low BV/TV, trabecular bone is structured to resist loads.

In summary, the distribution of BV/TV within the third metacarpal head reflects species-specific differences in habitual joint loading. While some species show fewer differences among regions than predicted, all of the observed differences are consistent with hand function. In contrast, the spatial pattern of DA did not match predictions, and remained at similar levels in regions with high and low bone volume. This suggests that the spatial pattern of DA may be more informative in situations involving highly stereotypical loading, such as the distinct pattern of dorsally-increasing DA in the metatarsal heads of humans (Griffin et al., 2010a), who habitually hyperextend the toes during walking and running and do little else demanding with those joints. We also note that the principal orientations of anisotropic structure and variations in anisotropic structure (e.g., rods, plates) can be informative (e.g., Ryan and Ketcham, 2002).

This study used a simple approach of examining three distinct regions within the epiphysis of a mobile joint. We suggest that it would be useful to develop more sophisticated approaches to statistically compare spatial patterns of trabecular structure. Also needed are data documenting function with finer resolution, such as hand postures used by primates in natural settings. Spatial analytical approaches together with finer-scale data on behavior will improve our abilities to understand the form-function relationship. This in turn will lead to more robust interpretations of the functional behavior of our long-extinct ancestors and relatives. Indeed with the development of improved quantification strategies, trabecular bone morphology holds promise of improving behavioral inferences in the fossil record.