Previous research on human carpal bones and other hand bones has revealed significant differences in size between males and females (Scheuer and Elkington, 1993; Falsetti, 1995; Lazenby, 1994, 2002; Barrio et al., 2006; Sulzmann et al., 2008; Mastrangelo et al., 2011). These studies usually use this variation in size to determine sex from the bone in a forensic context. For example, variation in the size of some carpal bones can be used to successfully (up to 88.6%) discriminate males from females (Sulzmann et al., 2008). However, the question as to whether this size variation in male and female human carpal bones manifests as shape variation between the sexes as well has been much less explored.
Ateshian et al. (1992) showed that, compared to males, females have a significantly different shape of the trapezium's first metacarpal facet, resulting in less congruency of the first carpometacarpal joint. Wang et al. (2010) found significant sex-related differences in overall carpus length. Others have described variation in facet morphology of the lunate (Viegas et al., 1990; Nakamura et al., 2001), capitate (Yazaki et al., 2008), and carpometacarpal joints (Viegas et al., 1991; Marzke et al., 1994; Nakamura et al., 2001), but sexes were pooled for these studies. Few studies have investigated shape dimorphism in human carpals (Ateshian et al., 1992; Wang et al., 2010; Marzke et al., 2012) yet higher incidences of osteoarthritis in the female first carpometacarpal joint (Bagge et al., 1991; Haara et al., 2004; Wilder et al., 2006) have been explained by variation in shape between sexes (Ateshian et al., 1992). Thus, investigating potential shape variation in the human carpus may have clinical implications. Dimorphism in other aspects of carpal morphology may partially account for significantly higher prevalence of osteoarthritis in particular wrist joints in either males or females (Butler et al., 1988; Bagge et al., 1991; Haara et al., 2004; Wilder et al., 2006). Sex variation in the relative degree of wrist mobility (Brumfield et al., 1966; Moritomo et al., 2006) and hand strength (Chao et al., 1989; Morse et al., 2006; Phu, 2010) may further suggest some degree of shape variation between the male and female carpal bones that has functional implications.
Most movement in the wrist takes place at the radiocarpal and midcarpal joints. The radiocarpal joint is responsible for more than half of the range of motion of the wrist (Crisco et al., 2005), and the midcarpal joint is the functional crux of the wrist, allowing for range of motion in all planes (Ruby et al., 1988; Viegas et al., 1993). Thus shape differences, if any, between the sexes may appear in the carpal bones that comprise these joints. For example, females have a larger range of range of motion in flexion-extension at the radiocarpal joint compared to males (e.g., 65 vs. 60 degrees, respectively; Brumfield et al., 1966; Marshall et al., 1999). Similarly, it has been shown that males and females have a slightly different axis of rotation on the capitate proximal facet during flexion-extension and radioulnar deviation (Neu et al., 2001; Rainbow et al., 2008). Therefore, although variation in soft tissue anatomy likely also plays a critical role, we would predict that if differences in carpal shape exist between human males and females, that these differences may occur in the relative shape of the radial facets on the lunate and scaphoid and the proximal facet of the capitate.
Furthermore, variation in androgen expression during fetal development has been correlated with sex differences in the relative length of the second and fourth digits in human males and females (Manning et al., 1998; Brown et al., 2002; but see Buck et al., 2003). Thus, it is possible that androgen expression affects other bones of the hand as well. Although the underlying shape of the carpal and its facets are present in the cartilage anlage (Cihák, 1972; Lovejoy et al., 2003; Kivell, 2007), the prolonged period of carpal development in humans [i.e., roughly 12–15 years of ossification (Scheuer and Black, 2000)] may allow carpal shape to be influenced by other ontogenetic factors, such as variation in endocrine levels or epigenetic factors. Abnormalities in the development of carpal bones are common with several congenital diseases associated with pathological expression of endocrine, such as hypothyroid epiphyseal dysgenesis and ovarian agenesis (Wilkins, 1941; O'Rahilly, 1953; Poznanski and Holt, 1971), demonstrating that carpal development is susceptible to changes endocrine expression.
Here we examine patterns of shape variation between the sexes in five bones of the carpus that are involved in the radiocarpal and midcarpal joints; the scaphoid, lunate, triquetrum, capitate, and hamate. We test the null hypothesis that the shapes of the carpal bones in male and female humans are similar. For the shape variables that show significant differences between males and females, we further examine whether this variation is due to a shared, intraspecific pattern of size-correlated shape change or whether females and males show divergent scaling patterns. There are several potential developmental, functional, and clinical implications of the patterns of shape variation in males and females that should be further investigated.
MATERIALS AND METHODS
We investigated sexual dimorphism in carpal shape in a sample of male (N = 70) and female (N = 74) modern humans. The human sample was composed mainly of white and black individuals derived from the Grant collection (University of Toronto) and the Terry Collection (National Museum of Natural History, Smithsonian Institute) from the late 1800s to mid-1900s (Table 1). In addition, a small sample of small-bodied Khoisan individuals was included to incorporate a larger range of variation in size (Table 1). All specimens were considered adult based on complete eruption of the third molar and/or complete epiphyseal union throughout the skeleton and lacked osseous pathology. Sex identification was based on museum records and/or diagnostic skeletal features of the associated pelvis or skull. All specimens for which sex could not be clearly determined were excluded from the analysis.
Table 1. The human sample
University of Toronto Grant Collection
White (1928 to early 1950s)
NMNH Terry Collection
Black (late 1800s to early 1900s)
NMNH Terry Collection
White (late 1800s to early 1900s)
University of Vienna
Khoisan (early 1900s)
Forty-one linear measurements (e.g., proximodistal length, dorsopalmar height, and/or mediolateral breadth) of each carpal bone and its articular facets were taken on the scaphoid, lunate, triquetrum, capitate, and hamate using Mitutoyo (Mitutoyo Corporation, Kanagawa, Japan) digital calipers (Table 1). All measurements were taken by one observer (T.L.K.) and on the right side, unless unavailable (then measurements were taken on the left). Intraobserver measurement error was tested on N = 25 specimens on three separate occasions. Measurement errors were calculated using the methods outlined by White (2000), and the average error was less than 1% for most variables, although closer to 1.5% for measurement of the length of the scaphoid body (LSB), length and breadth of lunate's radial facet. This measurement error is consistent with other morphometric studies of the hand (Weinberg et al., 2005).
Since body mass is rarely available for museum specimens, a geometric mean derived from all carpal bone dimensions was used as the size variable (Jungers et al., 1995). For each bone, each linear measurement was divided by the geometric mean of all measurements to create a dimensionless shape ratio (Mosimann, 1970; Darroch and Mosimann, 1985). Differences in the shape ratios between males and females were assessed using a Mann-Whitney U test, and results were considered statistically significant at the P = 0.05 level (Quinn and Keough, 2002).
The shape ratio will incorporate size-correlated (or allometric) shape variation (Falsetti et al., 1993; Jungers et al., 1995). The size-correlated shape variation can be due to a common allometric slope or due to different allometric slopes in males and females. For each shape ratio that differed significantly between males and females, we assessed the pattern of intraspecific scaling by calculating slopes of the ordinary least squares (OLS) regression and standard major axis (SMA) regression (Warton et al., 2006; Smith, 2009). The appropriate log10 variable was regressed on the log10 geometric mean. For variables that displayed statistically similar SMA slopes between the sexes, tests of equal elevation (i.e., equal intercept values) were evaluated with the Wald statistic (Warton et al., 2006) to address the potential for sex-specific variation in the intercept to affect the values of the shape ratios. However, we interpret differences in male and female intercept values with caution for several reasons. In all of the SMA regressions, the intercepts are negative and very close to zero indicating that the intercept will have negligible influence on the value of the shape ratios in males compared with females. Moreover, even though several of SMA slopes were determined to be statistically similar, the male and female regression lines often crossed one another indicating a significant interaction (height of the lunate's radial facet [HLRF], height of the proximal facet [HCPF], and breadth of the capitate body [BCB]) or converged near the intercept (length of the lunate triquetrum facet [LLTF], length of the triquetrum body [LTB], and longer hamate body [LHB]). All regression analyses, including tests for equal slopes and elevation, were performed in SMATR (Falster et al., 2006; Warton et al., 2006).
Finally, to assess the overall utility of carpal shape to distinguish males from females, we performed a discriminant function (DF) analysis for each carpal bone. DF analyses included all metric variables for each carpal bone and were performed on both on shape ratios (reported here) and raw data for comparison (Table 2).
Table 2. A description of the linear measurements
Each linear measurement (“msmt”) was divided by each geometric mean to create a dimensionless shape variable.
max. proximodistal length of scaphoid body
max. dorsopalmar height of scaphoid body
max. mediolateral breadth of scaphoid body, excluding tubercle
max. dorsopalmar length of scaphoid radial facet
max. proximodistal height of scaphoid radial facet
max. dorsopalmar length of scaphoid lunate facet
max. proximodistal height of scaphoid lunate facet
max. proximodistal length of lunate body
max. dorsopalmar height of lunate body
max. mediolateral breadth of lunate body
max. dorsopalmar height of lunate scaphoid facet (at distal edge)
max. proximodistal length of lunate scaphoid facet
max. dorsopalmar height of midcarpal joint facet
max. mediolateral breadth of midcarpal joint facet
max. dorsopalmar height lunate radial facet
max. mediolateral breadth of lunate radial facet
max. dorsopalmar height of lunate triquetral facet
max. proximodistal length of lunate triquetral facet
max. mediolateral breadth of triquetrum body
max. dorsopalmar height of triquetrum body
max. proximodistal length of triqetrum body
max. dorsopalmar height of triquetrum lunate facet
max. proximodistal length of triquetrum lunate facet
max. mediolateral breadth of triquetrum hamate facet
max. dorsopalmar height of triquetrum hamate facet
max. proximodistal length of capitate body
max. dorsopalmar height of capitate body
max. mediolateral breadth of capitate body
max. dorsopalmar height of capitate hamate facet
max. proximodistal length of capitate hamate facet
max. mediolateral breadth of capitate proximal facet
max. dorsopalmar height of capitate proximal facet
max. proximodistal length of hamate body
max. proximodistal length of hamate body, excluding hamulus
max. dorsopalmar height of hamate body
max. dorsopalmar height of hamate body, excluding hamulus
max. mediolateral breadth of hamate body
max. dorsopalmar height of hamate capitate facet
max. proximodistal length of the hamate capitate facet
max. dorsopalmar height of hamate triquetrum facet
max. proximodistal length of hamate triquetrum facet
Mann-Whitney U tests were conducted on the complete sample and also excluding the Khoisan sample to test if the small-bodied humans were providing an “extreme” morphology that biased the analysis. Results with and without the Khoisan sample were the same (i.e., significant differences in same shape ratios and slopes between males and females did not change), and thus only the results of the complete sample are presented.
Table 3 presents the mean and standard deviation of each shape ratio and the results of the Mann-Whitney U tests of significant differences between human males and females. The majority (76%) of shape ratios did not differ significantly between the sexes, supporting the null hypothesis that males and females have similar carpal bone shapes. However, in each carpal bone there was some aspect of shape that was significantly sexually dimorphic. In total, 10 shape ratios (24%) differed significantly between males and females. Each of these is discussed in detail below.
Table 3. Sex dimorphism in shape ratios between males and females
Mean and standard deviation (SD) of size-adjusted shape ratios are provided for males and females. See Table 2 for measurement abbreviations.
Significantly different shape ratios are shown in bold with:
DF analyses revealed that carpal shape did not discriminate well between males and females for any carpal bone (Fig. 1). Carpal bones displaying the most sexual dimorphism (i.e., the lunate and capitate) had the highest number of specimens classified correctly (Fig. 1), but all bones ranged between only 62% and 67% correct classification (as compared to 80%–87% using the raw data in which variation in size is included).
The SMA regressions are depicted in Figs. 2–8. The results for both the SMA and the OLS are reported in Table 4. The r2 values for the SMA regressions range from low (0.34 for HCPF in males; 0.35 for HLRF in females) to high (0.94 for LHB excluding the hamulus [LHB-H] in males; 0.84 for LTB in females). The SMA regressions indicate that most linear variables scale with positive allometry in both males and females. For eight of the shape ratios, sex-related shape variation is associated with statistically similar sex-specific scaling relationships, although females consistently show higher slopes than males. Only the LSB and the height of the lunate triquetrum facet (HLTF) showed a significantly higher SMA slope in females (Table 4). These findings indicate that the significant dimorphism in the majority of the shape ratios was a function of subtle (i.e., not statistically significant) scaling differences between males and females. Four of the shape ratios with common SMA slopes showed significantly different intercept values which range from −0.01 to −0.65 in log10 space. The largest difference in the intercept value between males and females was log10−0.04 for LLTF in males as compared with females.
Table 4. Results from regressions and comparisons of slopes and intercepts in sexually dimorphic carpal shape ratios
Male versus female
OLS slope (P)
SMA slope (P)
SMA intercept (M/F)
The coefficient of determination (r2), OLS slope, and SMA regression values for males and females for all carpal variables that displayed significant differences in shape between sexes.
Male and female regression values are reported separately in left columns:
* P ≤ 0.05 and
**P ≤ 0.01 represent the probability that the slope is significantly different from 1 (i.e., isometry).
Comparisons between male and female regression values are reported in the right columns; results of tests for significant differences in OLS and SMA slopes and significant differences in SMA intercepts using the Wald statistic:
* P ≤ 0.05 and
**P ≤ 0.01
ns, not significant; –, test does not apply because slopes are not common.
The LSB was significantly dimorphic, with females having a relatively longer length for their size compared to males (Table 3, Fig. 2). In the SMA regressions, LSB was the only variable to show negative allometry in males and isometry in females (Table 4, Fig. 2). Thus, as carpal size increased, females had a relatively longer LSB compared with males.
The lunate displayed significant shape dimorphism in the radial and triquetrum facets. Males had a relatively taller radial facet (HLRF) compared with females (Table 3, Fig. 3). OLS regressions revealed that males and females shared similar negative scaling with carpal size (i.e., slopes were equal; Table 4), but that males were slightly less negatively allometric (Table 4). Interestingly, the SMA regressions reversed this relationship indicating a slightly higher slope in females (Fig. 3) and suggesting that for HLRF the OLS regression slopes provide a more accurate reflection of the sex differences in the shape ratios. The SMA reversal is due to the low correlation coefficient (r = 0.59) for HLRF relative to the geometric mean in the females.
Humans also displayed significant shape dimorphism in the lunate's triquetrum facet, with females having a taller and longer triquetrum facet (HLTF and LLTF, respectively) compared with males (Table 3). For the HLTF, both the OLS and SMA regressions indicated that females had a higher slope compared with males, and this slope difference was significant for the SMA regression (although the correlation coefficient [r = 0.64] for females was low; Fig. 4). For the LLTF, both OLS and SMA regressions indicated that females were positively allometric.
The relative LTB shape was significantly larger in males compared with females (Table 3; Fig. 5). Both males and females displayed a similar positive scaling with carpal size for SMA and OLS regressions (Table 4), and the higher LTB shape ratio in males was attributed to a shared pattern of positive allometry in both males and females (Fig. 5).
In the capitate, the shape of the proximal facet was sexually dimorphic, with females being significantly broader breadth of the proximal facet (BCPF) and males significantly taller (HCPF) (Table 3; Fig. 6). Males also had a significantly broader BCB (Table 3; Fig. 7). Male and female slopes were not statistically distinct for any of these three variables (Table 4; Figs. 6 and 7). The elevation of the BCPF female SMA slope was significantly above the male slope (Table 4; Fig. 6). The significantly higher shape ratio for BCPF (i.e., broader facet) in the females thus was likely due to the combined influences of a narrower range of geometric mean, several individuals with high values for BCPF, and shape variation captured by the intercept comparisons (Fig. 6). The higher HCPF and BCB shape ratios in males were attributed to a shared pattern of positive allometry in both males and females (Figs. 6 and 7).
Females had a significantly longer length of the hamate body (with and without the hamulus; LHB and LHB-H, respectively) compared with males (Table 3; Fig. 8). The OLS regressions showed that the LHB and LHB-H scaled isometrically relative to hamate size in both males and females. However, the SMA regressions indicated that both males and females were positively allometric in the LHB, with females displaying a slightly higher slope (Table 4; Fig. 8). SMA regressions also revealed isometry in the LHB-H, but males had a slightly higher slope. For both LHB and LHB-H, the female SMA slope was significantly above that of the males (Table 4; Fig. 8). This pattern suggests that the significantly longer hamate body, particularly when the hamulus is excluded, may be explained at least partially by the shape variation captured by the intercept comparisons.
Although past studies have shown that human carpal bones vary in their absolute size between males and females, it was unknown if this size difference was associated with shape variation and, if so, whether shape variation was size-correlated (i.e., showed size allometry) or was independent of overall variation in carpal size. We hypothesized that males and females would show similar carpal shapes, despite variation in size. If sex-related shape dimorphism was present in the human wrist, we predicted that this variation would be found in the radiocarpal joint (i.e., radial facets of the scaphoid and lunate) and the proximal facet of the capitate.
This study revealed that most aspects of carpal shape were similar between human males and females, supporting the null hypothesis. The majority of shape ratios showed no within-sex correlation with size (Table 4), and many shape ratios showed very similar mean values between the sexes (Table 3). The poor classification of males and females in the DF analyses is consistent with this pattern (Fig. 1). These results indicate that most aspects of carpal shape are similar despite variation in overall size within and between sexes. However, not all aspects of carpal morphology showed this pattern. Each carpal bone displayed some aspect of sex-related shape dimorphism and this dimorphism could be most often explained by subtle (i.e., not statistically significant) scaling differences with carpal size or slight differences in shape represented by the intercept comparisons.
Patterns of Shape Dimorphism in Humans
As predicted, the lunate's radial facet (but not the scaphoid's radial facet) varied in shape between the sexes; males had a relatively taller height of the radial facet compared with females. Previous research has shown, however, that females, not males, have a larger range of motion in flexion-extension at the radiocarpal joint (Brumfield et al., 1966; Marshall et al., 1999), which is not consistent with the hypothesis that a larger radial facet facilitates a larger range of motion. Males exert a significantly stronger grip force and wrist torque compared to females (Chao et al., 1989; Hallbeck, 1994, Morse et al., 2006), which may be associated with a larger articular surface at the radiocarpal joint to distribute load. However, the functional consequences, if any, of this shape dimorphism in the lunatoradial joint remain unclear.
This study revealed that, as predicted, human males and females differed in the shape of the capitate's proximal facet. Males had a relatively taller proximal facet compared with females, which is consistent with previous research showing that males have more distal axes of rotation on the proximal head of the capitate during flexion-extension and radioulnar deviation (Neu et al., 2001; Rainbow et al., 2008). The difference in midcarpal rotation between the sexes has been attributed to the larger size of the capitate in males compared to females and not differences in function (Neu et al., 2001; Crisco et al., 2005; Rainbow et al., 2008). Our study suggests the shape of the proximal facet, with females also being significantly broader than males, may also facilitate sex-related differences in midcarpal joint motion.
This study also revealed significant carpal shape dimorphism between males and females that we did not predict. Females displayed a relatively longer LSB compared with males, which was explained by significantly different sex-specific scaling patterns with scaphoid size.
Compared with males, females had a larger triquetrum facet on the lunate but a relatively shorter length of the triquetrum itself. This dimorphism could be explained in part by significant variation in scaling with lunate size and may also suggest females have a larger range of movement at the lunatotriquetrum articulation. Females have been documented to have a larger range of wrist flexion than males (82 and 73 degrees, respectively; Brumfield et al., 1966), in which flexion and radial deviation of the triquetrum contribute substantially to this overall movement of the wrist (Fiepel et al., 1994; Werner et al., 1997). However, few studies have discussed differences in intercarpal joint motion between the sexes (Craigen and Stanley, 1995 [scaphoid]; Neu et al., 2001; Crisco et al., 2005; Rainbow et al., 2008 [capitate]) while most studies are based on pooled samples (e.g., Gellman et al., 1988; Ruby et al., 1988; Feipel et al., 1994; Moojen et al., 2002; Moritomo et al., 2006; Foumani et al., 2009). Data presented in Moritomo et al. (2006) suggest that females may have slightly greater range of motion at the lunatotriquetrum joint compared to males, but sample sizes were small and no formal comparison was made.
Human males had a broader capitate body relative to carpal size compared with females. The broader capitate, which is measured at the distal end of the body (Table 2), may reflect a broader base of the articulating metacarpals (Manolis et al., 2009). However, relative differences in metacarpal size between the sexes have not been documented.
Finally, compared with males, females had a relatively longer hamate body, both including and excluding the hamulus (although the hamulus often does not extend distally beyond the length of the hamate body in humans as it does in other hominoids). Increased length of the hamate may elongate the overall length of the carpus or the length of the carpal tunnel in females. However, Wang et al. (2010) demonstrated that males have significantly longer carpus length (termed “carpal height” in their analysis but measured as a ratio of proximodistal carpus length to third metacarpal length) than females, which is not consistent with this hypothesis.
Potential Causes and Implications of Human Sexual Dimorphism
Human evolutionary history may permit shape dimorphism in the wrist. Compared with other primates, the human wrist and hand are no longer load-bearing during locomotion. Human carpal morphology may be more “free” to vary in shape because it is not constrained by the potentially high loads of quadrupedal locomotion. Although the human wrist is still subject to potentially high loads during manipulation (Marzke et al., 1999; Rolian et al., 2011; Williams et al., 2010), this loading may be less predictable, less habitual, and potentially more variable between the sexes than those produced during locomotion in other primates.
A study of carpal shape throughout ontogeny is necessary to determine when shape variation between males and females develops. The underlying shape of the carpal and its facets are present in the cartilage anlage (Cihák, 1972; Lovejoy et al., 2003; Kivell, 2007) and thus these differences may appear prenatally. Sex-related variations in the intercepts in log10 space (i.e., regression line elevation) were statistically significant, but small. Where intercepts differed, females always showed an elevated regression line. It would be interesting to determine whether these small differences in carpal shape between males and females appear early in ontogeny, like variation in metacarpal length (Manning et al., 1998; Brown et al., 2002).
Functional factors, such as slight variation in wrist biomechanics and loading, may also influence subtle shape differences between males and females throughout the process of ossification and remodelling. Sex-related variation in range of motion (Brumfield et al., 1966; Marshall et al., 1999) and axes of rotation within the wrist (Neu et al., 2001; Crisco et al., 2005), or differences in hand strength and muscle force between human males and females (Chao et al., 1989; Marzke et al., 1998; Morse et al., 2006; Puh, 2010) are examples of external factors that could either (1) produce variation in carpal morphology during development or (2) be the result of sex variation in carpal morphology. Understanding the potential cause(s) and effects of sexual dimorphism in human carpal morphology requires further investigation, especially regarding the clinical implications.
Clinical, biomechanical, and anthropological studies often pool male and female samples, with the assumption that carpal shape is similar between the sexes (e.g., Ryu et al., 1991; Feipel et al., 1994; Foumani et al., 2009). This study reveals that this assumption may not be valid for some aspects of shape. Differences in carpal shape have been linked to the higher frequency of osteoarthritis in females in some wrist joints (Ateshian, 1992). Thus, recognizing dimorphism in carpal shapes could help to explain the sex-related variation in the prevalence of osteoarthritis in other aspects of the wrist (Butler et al., 1988; Bagge et al., 1991; Wilder et al., 2006).
Limitations and Future Considerations
The majority of the human sample used in this study (from the late 1800s to mid-1900s) may be subject to secular changes similar to those identified in the long bones over the last two centuries (Meadows and Jantz, 1995). Therefore, shape differences between human males and females may be not as pronounced in a more recent human sample (and, alternatively, may be accentuated in an older human sample).
The authors thank all of the people and institutions that provided access to skeletal collections: K. Hunt and L. Gordon (Smithsonian Institution), S. Pfeiffer (University of Toronto) and S. Kirchengast (University of Vienna). They also extend their gratitude to D. Begun, M. Schillaci, D. Schmitt, M. Skinner, A. Sylvester, and two anonymous reviewers for helpful discussions, comments and statistical advice that greatly improved this manuscript, and to Jean-Jacques Hublin for his support.