Comment, Correction, and New Findings for “Foetal Bison Long Bones and Mortality Season Estimates at the Early Holocene Casper and Horner II Sites, North America”, by Ryan P. Breslawski, Tomasin Playford, and Christopher M. Johnston (2020), International Journal of Osteoarchaeology, Vol 30, 425–434

Breslawski et al. (2020) estimated seasons of death for foetal bison (Bison antiquus) at two early Holocene archaeological sites in North America: Casper and Horner II. These estimates were based on growth rates and gestation ages derived from the antero-posterior depth (APD) values of foetal bison long bone diaphyses. Breslawski et al. assessed each site based on APD values originally published by Wilson (1974, 150) for Casper and Todd (1987a, 133) for Horner II, confirming that the Casper foetal bone is inconsistent with previously hypothesized fall mortality and the Horner II foetal bone is consistent with previously hypothesized late-fall or early-winter mortality. Following this finding, present author Wilson identified a discrepancy in the APD measurement procedure used by Breslawski et al. (2020) versus Wilson (1974) and Todd (1987a). Given this discrepancy, both present authors agreed that Breslawski et al. must be updated, and we therefore chose to collaborate on a note that explores and resolves the issue. In the course of this work, we also updated the foetal age estimation method to inform users whether a measured specimen may be neonatal rather than foetal, as is detailed at the end of this comment.

The measurement discrepancy arose due to author Breslawski misreading the original APD measurement procedure outlined by Todd (1987b); to avert future confusion, we reiterate that procedure here. Todd's procedure requires that the minimum medio-lateral diaphysis breadth is first measured. The caliper jaws are then rotated at that anatomical breadth location until they rest on the anterior and posterior aspects of the specimen, providing an APD value at the same place on the shaft. Coincident locations for breadth and APD allow cross-sectional area to be roughly estimated from a modeled ellipse. They were also defined such that measurements are taken near the midshaft, as far as possible from the more complex articular stress environments that characterize each end. Although postnatal stresses will be much greater, foetal limb movements and associated stresses must already influence bone morphology in utero. The initial APD landmark location was defined at the point of minimum breadth to ease replicability, because terminal damage usually prevents a true element-midpoint from being located. Going forward, we refer to this APD dimension as APDminBR.

Todd provided APD dimension codes for the humerus (HM10), radius (RD6), femur (FM17), and tibia (TA9). For the single Horner II humerus, Todd (1987a, 133) indicated that dimension HM13 [sic: HM10 was intended] was for the “minimum antero-posterior diameter of the diaphysis,” though a clearer description would be “antero-posterior diameter at the point of midshaft minimum breadth.” In describing the foetal bison metrics for Casper, Wilson (1974, 151) stated that “… the mid-shaft minimum diameters …” were measured. No dimension codes were presented, as Wilson's work preceded Todd's more explicit descriptions. For this commentary, present author Wilson remeasured the Cactus Flower Site humerus that was included in the Casper Site table (Wilson 1974, 148) to confirm that his stated APD was measured using the same procedure later described by Todd (1987b). As such, all these values are consistent with APDminBR.

Breslawski et al.'s (2020, 428) procedure for measuring APD therefore departed from Todd (1987b), and, by extension, Wilson (1974). In contrast to the minAPD dimensions measured by Todd (1987a) and Wilson (1974), the values described by Breslawski et al. (2020) were taken by placing the caliper jaws on the anterior and posterior aspects of the diaphysis and sliding them along the diaphysis long-axis until a minimum value was reached. Consequently, these APD values are at anatomical locations independent of minimum breadth locations. Henceforth, we refer to this second APD dimension as minAPD.

Breslawski et al.'s (2020) dataset of foetal long bone length and APD values comprises measurements taken by three independent analysts. It was unclear to us whether all three analysts measured minAPD as described in the 2020 paper, or whether one or more followed the original APDminBR measurement procedure described by Todd (1987b). To investigate possible disparities, we plotted length and APD values by analyst (Figure S1). It is apparent that two analysts, Playford and Johnston, obtained comparable APD values for all four elements. Further, their values are similar to those of Wilson for the humerus (paired length and APD data are only available for the humerus for Wilson's measurements). However, Breslawski's APD values underestimate those of Wilson, Johnston, and Playford across all four elements. We therefore conclude that Wilson, Johnston, and Playford measured APDminBR following Todd (1987b), whereas Breslawski measured minAPD as described in the 2020 paper.

To quantify the discrepancy between the two APD measurement procedures, Breslawski measured both APDminBR and minAPD in a foetal and neonatal bison (B. bison) assemblage from the Upper Tucker archaeological site (Woodall 1967). Ten repeat measurements of each dimension for each specimen were taken, allowing for estimated standard errors of measurement (Table S1). Bivariate plots of the Upper Tucker foetal long bone specimen values reveal that the measurement discrepancy varies between element types (Figure S2). For the radius and femur, both APDminBR and minAPD are measured near or at the same anatomical locations. For the femur, the dimensions' values are within a standard error of measurement. In contrast, minAPD locations for the humerus and tibia are typically distal to APDminBR. As such, the difference between values is greater, with a maximum discrepancy of 1.52 mm, or a 10% increase in values from minAPD to APDminBR (Table S1: Tibia 3).

Linear regressions show that, on average, APDminBR is 1%–6% greater than minAPD (Figure S2), with the tibia displaying the largest difference. Breslawski et al. used these APD values to estimate diaphysis lengths and gestation ages, so smaller values correspond to shorter diaphysis lengths and, consequently, younger foetal ages.

To evaluate how these smaller APD values affect the conclusions in Breslawski et al. (2020), we rescaled the minAPD values measured by Breslawski using the Figure S2 regressions, making them comparable to the APDminBR values measured by Wilson, Todd, Johnston, and Playford. This shifted the quantile regressions that Breslawski et al. (2020) used to model relationships between APD and diaphysis length (Figure 1), producing shorter modeled lengths for the Casper and Horner II specimens (Table S2).

The shorter diaphyses correspond to younger modeled gestation ages for three specimens: a femur and tibia from Casper and the Horner II humerus (Table S3). For the femur, this is a difference of a single day at the upper bound. For the tibia, the age with the adjusted comparative data becomes 1 day younger at the lower bound and 13 days younger at the upper bound. For the humerus, the gestation age is 1 day younger at the lower bound and 5 days younger at the upper bound. The three specimens with gestation ages unaffected by the adjusted APD values comprise two radii and a femur. The lack of changes in these cases is due to the model representing ages as integer days, preventing small continuous metric differences from affecting age estimates in some cases.

When reassessing the hypotheses for deposition dates, the probabilities for each hypothesized date range show limited shifts (Table S4; Figure S3). For Casper, changes in the probabilities for fall mortality are too small to be captured at the 0.01 resolution, as published in Breslawski et al. (2020). Though these probabilities do differ from those in Breslawski et al. 2020, the small magnitude of this difference does not affect the original conclusions. The Casper foetal humerus is associated with a detectable shift, where the probability of late-fall/early-winter mortality changes from 0.16 to 0.18. This increase remains consistent with the initial conclusions presented by Breslawski et al.

Henceforth, the season of death estimation method detailed in Breslawski et al. (2020) should be employed using only the APDminBR measurement procedures detailed in Todd (1987b). Diaphysis lengths and regression-adjusted minAPD values (i.e., APDminBR) are included here as the default dataset for the method's R script (Appendix S1; Data S1), along with the dataset for adult bison long bone measurements required by the script, unchanged from its publication in 2020 (Data S2). The default dataset contains measurements (mm) for archaeological and modern foetal bison long bone diaphyses, as reported in four previous studies (Breslawski and Playford 2018; Frison et al. 1978; Johnston 2016; Playford 2015). Measurement landmark codes follow Todd (1987b). In the interest of clear reporting in future work, we recommend that analysts measure and report both APD dimensions for foetal bison diaphyses.

The method's R script now also displays information on whether a measured specimen is potentially neonatal rather than foetal. The method's growth curves extend to 320 days, longer than observed modern gestation durations, to capture potential neonatal specimens mistaken for foetal remains, as it can be challenging to distinguish late-term foetal from young neonatal bone in archaeological contexts. If the method is applied to young neonatal remains, failing to extend the curves into the early neonatal period would produce misleadingly young ages as well as narrower and earlier season of death estimates. The skeletal growth curves are based on mean gestation durations of 264–272 days as estimated for modern Yellowstone bison (B. bison) (Gogan et al. 2005), and herds elsewhere have been reported with mean durations as long as 292.5 days (Berger 1992, 324). Individual bison have recorded gestation durations spanning ~246 to ~316 days (visually estimated from Berger 1992: Figure 1), shown in Figure S4. Gestation duration varies with maternal condition and copulation date (Berger 1992; Berger and Cunningham 1994, 120–123), and wildlife management practices likely influence these durations accordingly. As such, modern animals are an imperfect reflection of pre-contact bison populations, though it is unclear exactly how gestation durations should be adjusted to account for this.

The R script now displays the percentages of a growth curve's area contained within a specimen's estimated diaphysis length range that are foetal (1–245 days), indeterminate foetal vs. neonatal (246–316 days), or neonatal (317–320 days). Table S5 displays these percentages for the six Casper and Horner II specimens. If there is strong contextual evidence that a specimen is foetal rather than neonatal, such as in situ recovery from the abdomen of an adult female skeleton, users can now adjust the growth curves in the R script based on this evidence. If a specimen is known to be foetal rather than neonatal, the curves can be truncated at the longest gestation duration, 316 days, which excludes the early neonatal period. Alternatively, the curves can be truncated at the midpoint of reported gestation durations, 281 days. However, the midpoint is earlier than some reported modern gestation durations detailed in the previous paragraph and may therefore produce misleadingly young ages.

In closing, this comment has clarified the procedure for APDminBR measurements and detailed that the season of death estimate R script now displays information about developmental stages informed by those measurements. This provides more accurate insights into the seasonal timing of depositional events and their implications for human hunting in North America's past.

The authors declare no conflicts of interest.