Early Holocene morphological variation in hunter-gatherer hands and feet

Aims

The focus of this paper is to explore whether or not logistical mobility and domestic economies (subsistence and tool manufacture) are archaeologically visible in the feet (tarsals) and hands (carpals) and, if so, whether or not they reflect a sex-based division of labor. Specifically, we are interested in bone functional adaptation in response to biomechanical stress and use logged length-width ratios to assess relative bone strength (Garn, 1972; Macdonald, Hoy & McKay, 2013; Rauch, 2005). We also examine lateralization using raw measures for left and right sides. Much attention has been paid to sexual dimorphism in carpal and tarsal bones in forensic contexts and with applications to sex identification bioarchaeology; these studies have had varied success with most pointing to the calcaneus and sometimes talus as the most diagnostic bones (Bostanci, 1962; Bunning, 1964; Gualdi-Russo, 2007; Harris & Case, 2012; Hoover, 1997; Introna Jr et al., 1997; Kidd & Oxnard, 2002; Mastrangelo, De Luca & Sanchez-Mejorada, 2011a; Mastrangelo et al., 2011b; Riepert et al., 1996; Steele, 1970; Steele, 1976; Steele & McKern, 1969)—for a review see (Davies, Hackman & Black, 2014). Here, we take a different approach by exploring whether activity is embodied in the often overlooked region of the hands and feet.

Carpals and tarsals have not been examined extensively in bioarchaeological contexts but are potentially very interesting bones. The hands and feet are heavily implicated in the daily activities of hunter-gatherers (e.g., mobility, use of weapons, tool-making, domestic economies) and that activity might be embodied in carpal and tarsal bones. Further, carpal and tarsal bones are less likely to be influenced by the noise created from conflicting signals of genetics and lifestyle that obfuscates differentiation of ultimate and proximate causes of variation that plagues the long bones (Pearson Osbjorn, 2000; Ruff & Larsen, 2014). Further, these dense and small bones tend to be among the better-preserved bones in archaeological contexts (Henderson, 1987; Mann, 1981) and, if daily activities are visible in these bones, we potentially capture data that might otherwise be lost in less well-preserved skeletons. At the very least, if they prove useful in identifying bone functional adaptation, they provide additional data for past activity reconstruction.

We examined carpal and tarsal bones from the Florida Early Archaic Windover Site. The Early Archaic is characterized by major climate change in North America and, along with it, a change in domestic economies. Warmer climate was driving big game north and broad-spectrum foraging was emerging as the primary subsistence economy, de-emphasizing the dietary contribution of males and increasing the contributions of females. The site consists entirely of the mortuary pond where mobile hunter-gatherers buried their dead (usually with grave goods, such as atlatls and stone tools). Due to the semi-tropical environment of Florida, seasonally occupied hunter-gatherer camps are not well-preserved and most of what we know of this period comes from mortuary ponds, rather than occupation sites. As one of the best-preserved and largest collections, Windover remains provide tremendous insights into this period. We generated research questions based on ethnographic and archaeological data from other hunter-gatherer populations and modified them based on specific patterns identified at the Windover Site in other studies on other skeletal elements. As such, the next section takes each area of interest (e.g., mobility) and starts with general patterns in hunter-gatherer populations then narrows down to what we know about Windover. Our specific research questions are placed at the end of Bioarchaeological Context after presenting the material that aided in their generation.

Mobility activity

Domestic economies and activity reconstruction

Windover patterns

Research questions

Mobility

1. Previous research indicates that male hunter-gatherers walk more and at greater speeds than females. Windover data suggest a shared labor load, reduced emphasis on big game hunting, and evidence for heavy load carrying in both sexes. Which model is reflected in bone functional adaptation? We might expect the shared load model at Windover to result in no significant between-sex differences in rear foot tarsal variables that reflect the ground force reaction during locomotion and midfoot tarsal variables that absorb shock during locomotion.

2. The asymmetrical DJD in lower limbs and fracture patterns reviewed previously suggest there may be tarsal asymmetry. Is this evidenced by directional asymmetry in tarsal bones?

Subsistence

1. Previous research finds greater hand trauma and domestic economy production in females but greater DJD in males. Are there between-sex differences in carpal bones?

2. Based on findings of asymmetry in contemporary populations (hunter-gatherer and industrialized) engaged in repetitive manual tasks, we might expect prehistoric hunter-gatherers regularly engaged in repetitive manual tasks (food processing, tool production) to exhibit a similar pattern. Is there directional asymmetry in the carpal bones studied?

Raw data collection and variables

Carpal and tarsal bones with standard anatomical reference points intact were included only from adults who had well-defined features used in sex assessment (given the focus on sex-based morphological variation). Sex assessment was carried out by Doran & Dickel (1988a) using nonmetric pelvic and cranial traits and metric analysis of femoral and humeral head dimensions following standard osteological methods (Buikstra & Ubelaker, 1994). The final sample was 44 (27 males, 17 females) but sample size varies by measurement. All measurements were taken on right and left sides, when available for inclusion in the directional asymmetry analysis. Length and width of whole bones were measured in millimeters for four carpals (capitate, hamate, lunate and scaphoid) and four tarsals (calcaneus, intermediate cuneiform, navicular, talus). Length and width of talar (trochlea) and calcaneal (load arm) articular surfaces were also measured. See Table 1 for measurement details using standard anatomical landmarks.

Size difference data and directional asymmetry analysis

Directional asymmetry is assessed by comparing side differences (here, selected right and left hand and foot bones). Confounding factors for directional asymmetry analysis are sexual dimorphism and trait size variation, discussed in-depth in asymmetry methodology papers that were applied to this analysis (Palmer, 1994; Palmer & Strobeck, 1986; Palmer & Strobeck, 2003). Preliminary raw data inspection via scatterplots and outlier statistical tests (e.g., Grubb’s statistic) were performed to eliminate confounding effects of outliers (data recording, trait size, anomalous individuals) (Palmer & Strobeck, 2003). Univariate analysis generated input for a mixed model ANOVA (sides × individuals) which was performed in the Fluctuating Asymmetry Calculations Worksheet (V.11) (Palmer, 1994; Palmer & Strobeck, 1986; Palmer & Strobeck, 2003).

Bone functional adaptation, index data and analysis

As discussed previously, bone functional adaptation in response to biomechanical forces can be assessed by bone width as a proxy for relative bone strength (Garn, 1972; Rauch, 2005). But, sexual dimorphism (Fairbairn, 1997; Jungers, 1984; Smith & Cheverud, 2002) and trait size variation (Huxley & Tessier, 1936; Lewontin, 1966) confound analysis conducted on raw width measurements. For example, a comparison between the width of the intermediate cuneiform and the width of the talus simply demonstrates that the talus is a wider bone because it is a larger bone. Likewise, a comparison between male and female talus widths simply demonstrates that males tend to be larger than females. An approach using raw measurements does not further our understanding of meaningful trait differences beyond absolute size.

A common resolution to the effect from sexual dimorphism is to take a ratio of two variables, such as length and width (Fairbairn, 1997; Huxley & Tessier, 1936; Jungers, 1984; Jungers, Falsetti & Wall, 1995; Lewontin, 1966; Mobb & Wood, 1977; Smith & Cheverud, 2002). We take length and width because width is an indicator of relative bone strength (Garn, 1972; Rauch, 2005)—and length and width are commonly taken, easily replicated measurements. As previously mentioned, resistance to bending force is related to bone diameter because apposition of new bone on periosteal surfaces (a functional adaptation to biomechanical stress) widens the bone (Macdonald, Hoy & McKay, 2013). Thus, the ratio of length to width provides a useful index of relative bone strength and is a marker of biomechanical forces acting on bone functional adaptation from activity. An index of one describes a bone that has equal length and width (1:1). A ratio of greater than one describes a bone that is longer than it is wide and a ratio of less than one describes a bone that is wider than it is long. When comparing bones that have indices that lie exclusively either above or below the set point of one, the relative bone strength can be inferred by smaller numbers. For example, if all the bones are wider than they are long (a ratio value below 1), those with the smallest values are relatively wider than those with larger values (but, see below for interpreting logged values).

The issue of trait size variation is common in asymmetry studies (Palmer, 1994; Palmer & Strobeck, 1986; Palmer & Strobeck, 2003) and often resolved by scaling data to the natural log after taking the absolute value of the length to width index (Mobb & Wood, 1977; Palmer & Strobeck, 2003). The natural log creates a symmetric and homoscedastic dataset that retains the original linear scale of standard deviation to the mean (the spread or variation of the data) (Sokal & Rohlf, 1995). We applied this final transformation to our data and that resulted in a total of 10 index variables (capitate, hamate, lunate, scaphoid, calcaneus, calcaneus load arm, intermediate cuneiform, navicular, talus, talus trochlea tibia) to test bone functional adaptation analysis. We transformed left and right sides separately because each represents an individual measure of biomechanical stress. Equation (1) represents the full transformation of raw length and width variables to the final index that accounts for the confounding issues of sexual dimorphism and trait size. The index is derived from the natural log (log_n) of the absolute value of length-to-width ratio (L:W) for each trait (x) (1)${\displaystyle {log}_{\mathrm{n}}\left|x\left(L:W\right)\right|.}$ An example of how this index mitigates the confounding effects of sex-based and trait-based size differences in bone metrics is seen in (Fig. 1 and Figs. S1–S9). The left panel of Fig. 1 is a scatterplot of raw measurements for the navicular. The right panel of Fig. 1 is a scatterplot of the index values. The distribution of raw data on the left reflects the absolute differences in body size (males larger than females) while the right panel reflects real differences once absolute size differences are eliminated by application of Eq. (1). Our index values describe bones that are absolutely longer than they are wide excepting the calcaneus load arm which is wider (than it is long) and the lunate which is equal in length and width (14.43:14.90 or .99). See Table 2 for an average of ratio values using raw length and width measurements (L:W) and an average of index values using Eq. (1). Table 2 demonstrates that the linear scale of the original data is retained and can be interpreted in a similar manner as before with the exception that values greater than zero are longer than they are wide (i.e., relatively long) and values less than zero are wider than they are long (i.e., relatively wide). To return to the earlier example, if all the bones are wider than they are long (a logged ratio value below 0), those with the smallest values (larger negative values) are relatively wider than those with larger values (closer to zero).

Tests for between-group differences using index variables characterized differences in relative width as an indication of biomechanical stress acting on the area of interest. The General Linear Model (GLM) was used to test for between-sex differences in tarsal and carpal variables. All results were evaluated relative to confidence intervals, power, and estimated effect size.

Tarsal variation

Based on previous analysis, we speculated whether the biomechanical stress of load carrying might influence bone functional adaptation in the tarsals. Typical hunter-gatherer males have faster walking speeds and tend to walk greater distances than females (Pontzer et al., 2014). If males were more specialized in hunting activities that required them to walk longer distances at faster speeds than females, we might expect relatively wider (i.e., stronger) weight-bearing bones in males (lower index values for talus and calcaneus variables)—this expectation is predicated on body weight being born by the rear foot (talus and calcaneus) (Nordin & Frankel, 2012; Trinkaus & Shang, 2008). We also might expect males to have greater impact shock from locomotion, which is absorbed by the midfoot (navicular, cuboid, and cuneiform bones) (Nordin & Frankel, 2012). The archaeological interpretation of the division of labor at Windover, while limited to grave goods, suggests a shared model of labor. If males are less specialized and share domestic economy tasks with females, we would expect no differences in the sexes and this is what we found. That said, both tests are underpowered and there may well be significant differences in the shock-absorbing bones that were not identified due to small sample sizes per variable. Still, the effect is very low (in this study), which suggests that any differences that might exist are not likely to have biological significance.

The calcaneus index values are the only ones that exhibit a clear sex-based trend (even if not significant after correction for multiple hypothesis testing). Recall that the bipedal heel strike during locomotion transmits body mass from the tibia to the rear foot (talus and calcaneus) to the ground (Nordin & Frankel, 2012). Extant Hadza hunter-gatherers tend to favor a midfoot strike (Pontzer et al., 2014) while the Daasanach pastoralists favor a rearfoot strike (Hatala et al., 2013), the latter of which significantly impacts the calcaneus more so than other bones and which increases dorsal spurs on the calcaneus (Weiss, 2012). Calcaneal dorsal spurs are correlated with running economy (long calcaneal tuber = greater energy cost) (Raichlen, Armstrong & Lieberman, 2011)—this has been noted in endurance runners in Kenya who appear to favor a forefoot strike (Lieberman et al., 2015). Because the Windover population was experiencing a comparatively wetter climate than in previous archaeological periods (Halligan et al., 2016), big game was harder to find. We know that the population had an abundance of local resources and emphasized broad spectrum foraging over big-game hunting (Doran & Dickel, 1988b; Milanich, 1994). Male burials, however, sometimes are associated with atlatls which suggests some big game hunting and might explain why males have relatively longer calcaneus bones than females (higher index values, closer to zero, are interpreted as relatively longer and lower index values, larger negative numbers, are interpreted as relatively wider). While some research argues that barefootedness causes wider feet (Lieberman, 2013), the situation is more complex. Clinical evidence of the effect on the foot due to barefoot walking and running during growth and development suggests that while the forefoot does widen (Franklin et al., 2015), the overall length is increased at the expense of width even when controlling for potentially confounding effects of demographic and developmental variation in activity and body weight (Hollander et al., 2017). A large cohort (n = 520) of children aged 6–18 in South Africa (habitually barefoot, even in school) and Germany showed that the wider foot cedes to a longer foot in the South African barefoot cohort. Variation in published findings may be related to factors such as the confounding factors influencing adult foot morphology and, in the one other study conducted on habitually barefoot children (D’AoÛt et al., 2009), factors such as BMI and ligament tensility (Hollander et al., 2017). The story of foot morphology and barefootedness is not yet complete.

Foot lateralization

Despite evidence suggesting footedness would not be visible in the archaeological record (Zverev, 2006), bioarchaeological evidence suggested there might be—namely, lateralization in the lower limbs might have resulted in lateralization of tarsal bones with the dominant foot having larger mid foot widths (due to biomechanical pressure to strengthen the bone during impact) or rear foot widths (due to biomechanical pressure to strengthen the bone during weight-bearing). Specifically, males had greater DJD on both knees and the left talus and calcaneus (Smith, 2008:47). These lateral patterns did not translate to differences in the tarsals bones in this population. As previously discussed, footedness in humans develops in late childhood (11–12 years old) with a right skew (Gabbard, 1996; Gentry & Gabbard, 1995) but its influence on walking gait is not significant and not likely to affect the musculoskeletal system in the absence of other evidence of lateralization (Zverev, 2006). A future study might confirm this by examining bone functional adaptation in response to foot preference in populations with known physical activity patterns that evidence clear lateralization. Further, the hallux has been identified as a potentially significant bone that might be examined in bioarchaeological contexts due to its involvement in the extreme plantarflexion associated with barefoot locomotion (Franklin et al., 2015).

Carpal variation

We postulated we would find between-sex differences in carpal bones due to increased hand trauma (Wentz, 2010) and cervical vertebra DJD in females (possibly linked to food and textile processing activities) (Adovasio et al., 2001; Adovasio, Soffer & Page, 2009; Wentz, 2010) compared to males who had a higher frequency of severe DJD in both hands (18% of the sample) (Smith, 2008). The multivariate test conducted on the index variables for the four carpal bones did not result in a significant finding of difference between males and females but, as noted in Results, the test was underpowered and the p-value was only slightly higher than the significance threshold. Thus, a Type II error may have occurred. The female carpal sample is smaller than the male carpal sample and there is no way to predict what the differences might actually be. The between-subjects tests did not indicate any individual variable was significantly different between the sexes but these were underpowered due to small sample size, particularly for females and the hamate index. Given that the female sample is much smaller than the male sample, we cannot comfortably make any conclusions relative to the research question. We can, however, suggest that there is very small (if any) effect size from activity marking on the bones due to biomechanical stress (in this study). And we note that a lack of difference circumstantially supports prior analysis suggesting heavy overlap between the sexes in labor as evidenced by male graves containing tools for domestic labor (Hagaman, 2009; Hamlin, 2001; Smith, 2008; Smith, 2003; Wentz, 2006).

There were some interesting trends worth noting even if they are not part of statistical hypothesis testing. First, female carpals (excepting the scaphoid) exhibit a greater range of index values than males (Table 4, Table S2). Females engage in broader non-diet activity (e.g., weaving, cordage) when male contributions to diet are larger (Waguespack, 2005). Given that the Early Archaic is characterized by rapid change from big game hunting to broad spectrum foraging (Doran & Dickel, 1988a; Halligan et al., 2016; Milanich, 1994) and given that Windover females are engaging in both subsistence and non-subsistence domestic activities (Hagaman, 2009; Hamlin, 2001; Wentz, 2010), perhaps our data reflect this transition and greater variety in female workload. First, female specific activities include health care as evidenced by grave goods for medicine preparation (Hamlin, 2001; Wentz & Gifford, 2007) and the Windover population experienced a variety of ailments (e.g., post-fracture bone alignment, surviving childhood stressors) (Wentz, 2006; Wentz, 2010; Wentz et al., 2006). Second, the sex-based preference for tool materials suggests a gendered ideology surrounding tasks (Hamlin, 2001). Finally, a between-sex health comparison found females had poorer health, which might be attributable to greater life stress from a heavier workload (Wentz et al., 2006). Thus, while we cannot conclude there are differences in biomechanical stress on carpal bones, we can suggest there may be evidence of females engaging in a wider variety of tasks. Ultimately and partly as a result of the very small effect size, the carpals do not appear to be useful (at least in this study) in identifying between-sex differences in biomechanical stressors from activity. But, perhaps the results reflect the ‘shared-load’ model put forward for Windover division of labor.

Hand lateralization

Our final area of inquiry was whether there was any evidence for lateralization in the carpal bones; this was based on previous research that suggested that hunter-gatherers exhibit strong handedness when engaged in complex tasks (Cavanagh et al., 2016; Robira et al., 2018; Stock et al., 2013). The complex tool kit at Windover would have provided an avenue for handedness to be archaeologically visible, but neither male nor female carpal bones exhibited significant directional asymmetry that would suggest lateralization. We might have expected some lateralization in males if the carpals were implicated in spear-throwing and males were engaged in heavy hunting activities but the Windover archaeological record does not suggest this was the case. Further, the bioarchaeological record shows that DJD and MSM patterns were shared between the sexes which suggests that males were not likely regularly engaged in hunting or the markers of habitual spear-throwing were offset by changes in other aspects of wrist anatomy (Maki, 2013: 238). A previous study on a population with clear handedness identified only two bones exhibiting directional asymmetry, the lunate and trapezium (Reina et al., 2017) which suggests handedness is not likely to be archaeologically visible.