Questionable Affiliation of Five Curated Crania: Using MicroScribe 3D Digitizer and FORDISC 3.1 Computer Program to Estimate Ancestry and Sex

Keywords

Cranium • Forensic anthropology • Repatriation • FORDISC 3.1 • Microscribe 3D digitizer • 3D Osteometric coordinate landmarks • Posterior probabilities • Typicality probabilities

Introduction

In the latter part of the 20th century, indigenous and marginalized peoples spoke out against the continued ownership of their biological and cultural materials by U.S. state and federal agencies and institutions. This push culminated with enacting into law the Native American Graves Protection and Repatriation legislation to address long-standing claims by federally recognized tribes that human remains and cultural artifacts-unlawfully removed from precontact, post-contact, former, or current Native American homelands-should be returned to lineal descendants for reburial [1]. With the addition of a stiff penalty of 12 months’ imprisonment and a $100,000 fine for violation of this law, state and federal institutions are complying. While a hefty fine is a great motivating factor in complying with this law, scientists in these institutions are largely motivated by shame because they know that, historically, documented human skeletal collections were built with the bodies of impoverished and marginalized peoples, and these scientists want to do the right thing by repatriating the remains despite their importance in forensic anthropological research, education, and training in the United States [2].

After the 2020 murder of George Floyd in police custody in the United States, the history of racial injustice perpetrated by 19th- and early 20th century scientists re-emerged with a vengeance. These scientists focused on scientifically “proving” the superiority of the White race over other races by measuring skull size, and in the wake of Floyd’s death, academics and activists turned their attention to the Morton human skeletal collection (formerly on display at the University of Pennsylvania Museum of Archaeology and Anthropology) as part of this history-and current perpetuation-of racism [3]. The Morton Collection consists of more than 1,300 skulls, approximately 900 of which were acquired by Philadelphia-based physician Samuel George Morton during the 1830s and 1840s; some of these belonged to enslaved individuals [4]. Now, as in 1991, when more than 400 individuals were reburied after New York City construction uncovered the largest-known African American burial ground in the United States, academics and activists are advocating for an African American Graves Protection and Repatriation Act. The repatriation of Native and African American cultural and biological remains is influencing indigenous and marginalized peoples in other parts of the world to fight for protection of their ancestral remains.

As a result of the renewed reckoning with racism caused by the curation and public display of human skulls in the Morton Collection, the administration at Bloomsburg University-a public university in Pennsylvania-directed this researcher to report on the ancestry of five human crania curated by the Anthropology Department. There are no substantive records on the origins of these five crania. Based on the limited information obtained, these crania were purchased in late 1990 (this researcher arrived at Bloomsburg University in 2005) from an unknown seller, and-while very little clear information exists about the identity of these individuals-they were supposedly East Indian and/or Chinese in origin. This researcher is no fan of the bone trade, despite the fact that it is legal and one way to obtain skeletal materials for teaching. It simply continues the tradition of 19th century archaeology and anthropology, when the acquisition of human skulls was the primary goal with little regard to the fact that these skulls once belonged to people. Over the years, this researcher saw no reason to question the India-China grouping, however ambiguous, because the focus in teaching was human anatomy and skeletal variation, but now that the ethical concerns about curation of and research on skeletal remains of marginalized peoples has reached the halls of university administrators, this researcher has decided to approach the problem as a forensic case to assess the origins of these crania.

The aim of this study is to estimate the ancestry and sex of five unknown human crania using the MicroScribe 3D digitizer to collect coordinate data from osteometric landmarks that were simultaneously recorded by analytical software called ThreeSkull (3 Skull) and subsequently imported into the FORDISC 3.1 discriminant functions computer program for processing. Depending on the results, steps will be taken to repatriate any Native American crania to their lineal descendants. All crania were handled in a respectful manner.

Material and Methods

MicroScribe G-2X digitizer

To better capture cranial size and shape, three-dimensional (3D) osteometric landmark coordinate data were collected using a MicroScribe G-2X digitizer. Practitioners have reported that 3D landmark coordinates show greater discrimination among modern cranial sample groups than traditional one-dimensional (linear) measurements and are, therefore, more valuable in a modern forensic setting [5]. Other advantages of 3D landmark data have been reported; for example, more nontraditional measurements (i.e., arcs and angles) can be calculated, a better representation of cranial morphology can be captured, a much lower error rate can be obtained, and conversion to linear measurements is easier, which results in greater data efficiency [6].

A tri-column stand 12.7 centimeters high made of non-hardening modeling clay was placed next to the digitizer and used to hold each of the five crania stationary while digitizing (Figure 1). The proximity of the cranium to the MicroScribe is important because the digitizing arm of the MicroScribe with a stylus attached must be able to reach all landmarks around the cranium. The stylus is used to capture one landmark at a time and must be in the homing position (Figure 1) prior to use in order for coordinates (X-Y-Z) to be accurately recorded [7].

Figure 1.Tri-column clay stands with skull secured on top in close proximity to MicroScribe positioned in the homing position and mandible secured on a piece of plexiglass.

A mirror was placed between the clay columns to assist with collecting osteometric landmarks on the base of each cranium. 3Skull was used to record 3D landmarks coordinates collected by the MicroScribe and to facilitate import of data to FORDISC 3.1. Before placing the cranium on the tri-column stand to begin digitizing, osteometric landmarks were located and marked with a pencil. In this research, the landmarks digitized and collected by 3Skull are listed by landmark, measurement, and brief description in Supplement 1 (S1). Not all of these 111 landmarks are used in craniometric analysis [8]. In fact, the Forensic Data Bank (FDB) in FORDISC 3.1 uses 56 landmarks (out of 111) and William W. Howells’ global Craniometric data set [9] (also in FORDISC 3.1) adds 39 new landmarks in addition to the 56 (S1). The osteometric landmarks digitized are indicated in (Figure 2a-g). Cranial interlandmark distances, angles, chords, elevation, radii, and subtenses were automatically calculated by 3Skull (Table 1). If a mandible accompanied any cranium, plexiglass with a bevelled edge and clay were used to hold the mandible stationary for digitizing (Figure 2).

Figure 2(a-g). Showing osteometric landmarks automatically calculated (red), marked before digitizing (blue), measured on work surface (brown), and arrows depicting digitizing arcs (anterior-posterior, medial-lateral). (Adapted from Fleischman and Crowder 2019).

Statistical analysis

FORDISC 3.1 generates the unknown’s posterior and typicality probability of membership in each reference group in the database. Posterior probabilities sum to 1 (100 percent) and is based on the unknown’s relative similarities (all Mahalanobis distances [D2 ]) to all groups) [10]. A high posterior probability, which in turn creates a small distance, indicates a greater similarity than to other groups. Typicality probabilities, in contrast, are the unknown’s probability of membership in each group, based on the unknown’s absolute similarity. The percentage of correct group allocations-or groups with the typical profile of the unknown case-is an indication of how well groups can be separated using the available variables. The word “typical” used above is important because distance probabilities or “typicality probabilities” can be calculated to ascertain whether an individual is typical for a specific group (and not assumed to belong to a respective group, as in posterior probabilities). When the typicality probabilities are uniformly low (i.e., less than 0.01 for each group), the posterior probabilities and classification should be disregarded because classification accuracy is critical in biological evidence for affiliation [11]. An important result is that the D2 values will follow a chi-square distribution with p degree of freedom.

Additionally, FORDISC 3.1 uses canonical variates to display data in graphic form. Canonical variate analysis is most effective in problems where many variables are used to compare differences among and within many reference groups. It is a technique that uses raw data to produce coefficients (or eigenvectors), and these coefficients are used to obtain new variables called canonical variates which maximize the among-groups variation (eigenvalues) relative to the standardized within-groups variation [12]. The variables (or measurements) are combined into a reduced number of functions to maximize the separation between groups. Such plots provide visual information as to which sample means (or centroids) are close or distant to one another in multivariate space. Moreover, multidimensional data space transforms confidence “intervals” into confidence “spheroids” (or ellipses), which are equidistant with regard to the within-group dispersion. Finally, there are usually several canonical variates, independently, holding biological information. However, it is the earlier variates that will contain information such as differences in overall shape and size.

For this research, the five crania were designated A12022, A22022, A42022, A52022, and A62022, respectively (Figure 3). A12022 and A22022 are the only crania with mandibles.

Figure 3. Five crania used in analysis to estimate race and sex. Each cranium is photographed in four views: frontal, left lateral, posterior, and basal (from top to bottom: A12022, A22022, A42022, A52022, A62022).

Results and Discussion

Analysis of Cranium A12022

After 3D osteometric landmark coordinates were collected for this cranium and recorded by 3 Skull, the data were imported to FORDISC 3.1. All cranial, mandibular, and FORDISC 3.1-calculated measurements (i.e., NAA, PRA, BAA, NBA, BBA, BRA) were used in the FDB (Table 1). Since there was no clear information on ancestry and sex for this cranium, all female reference groups (i.e., White females, Black females, Hispanic females, American Indian females, and Japanese females) and all male reference groups (i.e., White males, Black males, Hispanic males, Guatemalan males, American Indian males, Japanese males, Vietnamese males, and Chinese males) were used in the analysis. The name “American Indian” (as opposed to Native American) is the language used in the FORDISC 3.1 FDB.

On the first run (or initial processing), FORDISC classified cranium A12022 into the Japanese female (JF) reference group with a posterior probability of 0.572 (Table 2). The typicality probabilities were 0.394 (Typ F, which is dependent on sample size), 0.346 (Typ Chi-which is not dependent on sample size), and 0.221 (Typ R-where the cranium was ranked 88th out 113 individuals within the group). In essence, this cranium is as typical as 78% of Japanese females. However, other typicality probabilities show that this cranium is within the range of variation of the following reference groups: Chinese males, Japanese males, Hispanic females, Guatemalan males, Vietnamese males, and Hispanic males, all having typicality probabilities above .05. The graph of the results depicted in 3D canonical space showed this variation (Figure 4). Cranium A12022, the unknown (indicated by the bold ‘X’ in the graph) is closest to the Japanese female group centroid but within the ellipses of the aforementioned groups. See Supplement 2 for additional FORDISC descriptive data. In this analysis, 62 percent of the reference groups in FORDISC were classified correctly. A second run was performed using only Chinese males, Japanese males, Japanese females, and Vietnamese males reference groups and there was no change in the Japanese female classification for this cranium.

Figure 4. Graph of cranium A12022 FORDISC (FDB) classification results in canonical space (all male and female groups).

Analysis of Cranium A22022

All cranial, mandibular, and FORDISC 3.1-calculated measurements (i.e., NAA, PRA, BAA, NBA, BBA, BRA) were used in the FDB (Table 3). As in the previous cranium A12022, there was no clear information on ancestry and sex. Therefore, all female reference groups (i.e., White females, Black females, Hispanic females, American Indian females, and Japanese females) and all male reference groups (i.e., White males, Black males, Hispanic males, Guatemalan males, American Indian males, Japanese males, Vietnamese males, and Chinese males) were used in the analysis.

On the first run, FORDISC classified cranium A22022 into the Guatemalan (GTM) reference group with a posterior probability of 0.523 (Table 4). The typicality probabilities were 0.137 (Typ F), 0.105 (Typ Chi), and 0.119 (Typ R-where the cranium was ranked 59th out 67 individuals within the group). In essence, this cranium is as typical as 88% of Guatemalan males. But, FORDISC also showed that this cranium is within the range of variation of Hispanic males with all typicality probabilities greater than .05. The graph of the results depicted in 3D canonical space showed the unknown cranium A22022 (indicated by a bold ‘X’) close to the Guatemalan and Hispanic male group centroids but also in the ellipses of Japanese and Chinese male reference groups (Figure 5). See Supplement 3a for additional FORDISC descriptive data. In this analysis, 61 percent of the reference groups in FORDISC were classified correctly.

Figure 5. Graph of cranium A22022 FORDISC (FDB) classification results in canonical space (all male and female groups).

Due to the ambiguous graph in Figure 5, a second run of the data was performed using only male reference groups (White males, Black males, Hispanic males, Guatemalan males, American Indian males, Japanese males, Vietnamese males, and Chinese males). FORDISC classified cranium A22022 into the Hispanic male group with a posterior probability of 0.683. But the typicality Chi was below .05. Measurements PRA, BAA, NLH, and UFHT were one to two standard deviations lower or higher than all group means. NLH and UFHT were instrumentally checked and, along with PRA and BAA, not used in the analysis. On a third run, cranium A22022 was again classified into Hispanic males with posterior probability of 0.736 (Table 5). The typicality probabilities were 0.102 (Typ F), 0.076 (Typ Chi), and 0.166 (Typ R-where the cranium was ranked 131st out 157 individuals within the group). In essence, this cranium is as typical as 83% of Hispanic males.

This cranium is in the range of Guatemalan males, and the graph shows it closest to the Guatemalan and Hispanic male group centroids (Figure 6). See Supplement 3b for additional FORDISC descriptive data. In this analysis, 62 percent of the reference groups in FORDISC were classified correctly.

Figure 6. Graph of cranium A22022 FORDISC (FDB) classification results in canonical space (male groups only).

Analysis of Cranium A42022

All cranial and FORDISC 3.1-calculated measurements (i.e., NAA, PRA, BAA, NBA, BBA, BRA) were used in the FDB (Table 6). The shape transformation option was used because there was excessive left occipital sloping of the cranium. As in the previous crania, there was no clear information on ancestry and sex. Therefore, all female reference groups (i.e., White females, Black females, Hispanic females, American Indian females, and Japanese females) and all male reference groups (i.e., White males, Black males, Hispanic males, Guatemalan males, American Indian males, Japanese males, Vietnamese males, and Chinese males) were used in the analysis. There was no mandible available for this cranium.

On the first run, FORDISC classified cranium A42022 into the Guatemalan male (GTM) reference group with a posterior probability of 0.487. The typicality probabilities were 0.171 (Typ F), 0.135 (Typ Chi), and 0.119 (Typ R-where the cranium was ranked 59th out 67 individuals within the group). Therefore, this cranium is as typical as 88% of Guatemalan males (Table 7). Additionally, this cranium is in the range of variation of Hispanic and Japanese male and female reference groups. The graph of the results depicted in 3D canonical space showed this fact, but cranium A42022 is closest to the Guatemalan and Hispanic male group centroids (Figure 7). See Supplement 4a for additional FORDISC descriptive data. In this analysis, 61.2 percent of the reference groups in FORDISC were classified correctly.

Figure 7.Graph of cranium A42022 FORDISC (FDB) classification results in canonical space (all male and female groups).

A second run was performed using the Guatemalan male group with the Hispanic, Japanese, and American Indian male and female groups. Again, FORDISC classified the cranium into the Guatemalan male group with a greater posterior probability of 0.679, and its range of variation, based on the typicality probabilities, was within the same Hispanic and Japanese reference groups (Table 8). The graph of the results depicted in 3D canonical space showed the same results as in the first analysis: cranium A42022 is closest to the Guatemalan and Hispanic male group centroids (Figure 8). See Supplement 4b for additional FORDISC descriptive data. In this analysis, 63.7 percent of the reference groups in FORDISC were classified correctly.

Figure 8.Graph of cranium A42022 FORDISC (FDB) classification results in canonical space (Guatemalan, Hispanic, Japanese, and American Indian male and female groups).

Analysis of Cranium A52022

All cranial and FORDISC 3.1-calculated measurements (i.e., NAA, PRA, BAA, NBA, BBA, BRA) were used in the FDB (Table 9). As in the other crania in this research, there was no clear information on ancestry and sex. Therefore, all female reference groups (i.e., White females, Black females, Hispanic females, American Indian females, and Japanese females) and all male reference groups (i.e., White males, Black males, Hispanic males, Guatemalan males, American Indian males, Japanese males, Vietnamese males, and Chinese males) were used in the analysis. There was no mandible available for this cranium.

On the first run, FORDISC classified cranium A52022 into the American Indian male (AM) reference group with a strong posterior probability of 0.845. The typicality probabilities were 0.180 (Typ F), 0.132 (Typ Chi), and 0.388 (Typ R-where the cranium was ranked 30th out 49 individuals within the group). Therefore, this cranium is as typical as 61% of American Indian males (Table 10). Other typicality probabilities indicate that this cranium is not confidently in the range of variation in any other sample groups. This is shown very clearly in 3D canonical space. This cranium (indicated by a bold ‘X’) is near the centroid of the American Indian male group and barely in the spheroids of the other reference groups (Figure 9). A second run using only males did not change the results. See Supplement 5 for additional FORDISC descriptive data. In this analysis, 60.8 percent of the reference groups in FORDISC were classified correctly.

Figure 9.Graph of cranium A52022 FORDISC (FDB) classification results in canonical space (all male and female groups).

Analysis of Cranium A62022

All cranial and FORDISC 3.1-calculated measurements (i.e., NAA, PRA, BAA, NBA, BBA, BRA) were used in the FDB (Table 11). The shape transformation option was used because there was excessive left sloping and left and right parietal bossing when the cranium was viewed from the back. As in the previous crania, there was no clear information on ancestry and sex. Therefore, all female reference groups (i.e., White females, Black females, Hispanic females, American Indian females, and Japanese females) and all male reference groups (i.e., White males, Black males, Hispanic males, Guatemalan males, American Indian males, Japanese males, Vietnamese males, and Chinese males) were used in the analysis. There was no mandible available for this cranium.

On the first run, FORDISC classified cranium A62022 into the Chinese male (CHM) sample group with a posterior probability of 0.359 (Table 12). The typicality probabilities were 0.450 (Typ F), 0.393 (Typ Chi), and 0.311 (Typ R-where the cranium was ranked 51st out 74 individuals within the group). In essence, this cranium is as typical as 69% of Chinese males. However, based on the other typicality probabilities, FORDISC indicated that this cranium was in the range of variation of all reference groups except Black and White male and female groups. The graph of the results depicted in 3D canonical space showed the cranium (indicated by a bold ‘X’) in the ellipses of the aforementioned reference groups but closest to the Chinese male group centroid (Figure 10). See Supplement 6a for additional FORDISC descriptive data. In this analysis, 62 percent of the reference groups in FORDISC were classified correctly.

Figure 10.Graph of cranium A62022 FORDISC (FDB) classification results in canonical space (all male and female groups).

A second run was performed using only males with no change in the results. A third run was completed using the three sample groups with the highest posterior probabilities: Chinese males, American Indian males, and Japanese males (Table 13). The cranium was, again, classified in the Chinese male reference group with a higher posterior probability of 0.695 with typicality probabilities of 0.450 (Typ F), 0.393 (Typ Chi), and 0.311 (Typ R—where the cranium was ranked 44th out 74 individuals within the group) (Table 14). The graph of the results depicted in 3D canonical space showed a strong separation of the Chinese and Japanese male groups from the American Indian male group. But cranium A62022 is very close to the Chinese male group centroid (Figure 11). See Supplement 6b for additional FORDISC descriptive data. In this analysis, 76 percent of the reference groups in FORDISC were classified correctly.

Figure 11.Graph of cranium A62022 FORDISC (FDB) classification results in canonical space (Chinese, American Indian, and Japanese male groups).

Howells’ global Craniometric data set was used to pinpoint this cranium to a region in Asia; only East Asian males were used in the analysis. On the first run, FORDISC did not classify the cranium because some measurements deviated 1 or 2 standard deviations lower or higher than all group means. These measurements were not used (i.e., DKS, DKA, FRA, NAR, ZMR, ZOR). On the second run, FORDISC classified cranium A62022 in the Atayal male sample group with a very high posterior probability of 0.911. The typicality probabilities were 0.335 (Typ F), 0.067 (Typ Chi), and 0.233 (Typ R-where the cranium was ranked 23rd out 30 individuals within the group (Table 14).

The graph of the results depicted in 3D canonical space showed that cranium A62022 is in the range of variation of all groups except the Andaman Island and Buriat reference groups (Figure 12). See Supplement 6c (S6c) for additional FORDISC descriptive data. In this analysis, 66.3 percent of the reference groups in FORDISC were classified correctly.

Figure 12.Graph of cranium A62022 FORDISC Howells global craniometric data set classification results in canonical space (East Asian male groups).

Conclusion

After analysis of the osteometric coordinate landmark data for the five crania, collected using a MicroScribe G-2X along with 3 Skull to record 3D landmarks and import them into FORDISC 3.1 for processing, the following results were obtained:

1. The odds are very high that cranium A12022 is Japanese female. Despite the fact that this cranium has posterior probabilities in the Chinese male group, using only Asian males and females in a second analysis did not change the result of Japanese female. Overall, cranium A12022 belonged to an East Asian—most likely female.

2. The odds are very high that cranium A22022 is a Guatemalan male. This cranium also had posterior probabilities in the Hispanic male group. In fact, when all males were used in a second analysis, this cranium was classified in the Hispanic male group. There is tremendous complexity in knowing what is “Hispanic” vs. what is “Spanish” because of the Americas’ tremendous admixture in population historically. In short, this cranium belonged to a Guatemalan/Hispanic male.

3. The odds are very high that cranium A42022 belonged to a Guatemalan/Hispanic male (for similar reasons as cranium A22022).

4. The odds are very high that cranium A52022 is an American Indian male. This researcher will dig deeper into the details of the original acquisition with the eventual goal of repatriating this cranium to the lineal Native American descendants for reburial.

5. Finally, the odds are very high that cranium A62022 is a Chinese male. Furthermore, in a global comparison with East Asian reference groups, this cranium classifies into the Atayal or Taiwanese reference group.

As indigenous and marginalized peoples in other parts of the world become awakened to the plight of their cultural and biological remains, in time, all skeletal materials-after respectful analysis-will be repatriated.

Conflict of Interest

None.

References

1. “Native American Graves Protection Act.”(2021).
2. Campanacho, Vanessa, Francisca Alves Cardoso, and Douglas H. Ubelaker. "Documented skeletal collections and their importance in forensic anthropology in the United States." Forensic Sci 1 (2021): 228-239.

Google Scholar, Crossref

3. Wade, Lizzie. "Ghost in the museum: Anthropologists are reckoning with collections of human remains—and the racism that built them." Sci 373 (2021): 148-152.

Google Scholar, Crossref

4. Renschler, Emily S and Janet Monge. "The Samuel George Morton Cranial Collection." Expedition 50 (2008): 30-38.

Google Scholar

5. Ross, Ann H., Ashley H. McKeown and Lyle W. Konigsberg. "Allocation of crania to groups via the “new morphometry”." J Forensic Sci 44 (1999): 584-587.

Google Scholar, Crossref, Indexed at

6. Ousley, Stephen, and Ashley McKeown. "Three dimensional digitizing of human skulls as an archival procedure." Bar Int Series 934 (2001): 173-186.

Crossref

7. Ousley, Stephen D. "How to record landmark coordinates from a cranium using 3 Skull and a Micro Scribe digitizer." (2013).
8. Fleischman, Julie M., and C. M. Crowder. "Standard operating procedure for microscribe 3-dimensional digitizer and craniometric data." Washington, DC: US Department of Justice (2018).

Google Scholar

9. Howells, William White. "Cranial variation in man: a study by multivariate analysis of patterns of difference among recent human populations." Peabody Museum Archaeology Ethnology, Harvard Univ (1973).

Google Scholar, Crossref

10. Ousley, Stephen, and R. Eric Hollinger. "A forensic analysis of human remains from a historic conflict in North Dakota." Hard Evidence (2015): 91-102.

Google Scholar

11. Jantz, Richard L., and Stephen D. Ousley. "FORDISC 3.0: Personal computer forensic discriminant functions." University of Tennessee, Knoxville (2005).

Google Scholar

12. Albrecht, Gene H. "Multivariate analysis and the study of form, with special reference to canonical variate analysis." Am Zool 20 (1980): 679-693.

Google Scholar, Crossref