bioRxiv preprint doi: https://doi.org/10.1101/2021.01.08.425868; this version posted January 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. 1 Towards the restoration of ancient hominid craniofacial anatomy: 2 Chimpanzee morphology reveals covariation between craniometrics 3 and facial soft tissue thickness 4 5 Ryan M Campbell1,*, Gabriel Vinas2, Maciej Henneberg1,3 6 7 1 Adelaide Medical School, Biological Anthropology and Comparative Anatomy Research Unit, The 8 University of Adelaide, Helen Mayo North, Floor 2, Room 24, Frome Road, Adelaide, South 9 Australia,5000, Australia 10 11 2 Herberger Institute for Design and the Arts, Sculpture Department, Arizona State University, Art 12 Building, 900 S Forest Mall, Tempe, Arizona, 85281, United States 13 14 3 Institute of Evolutionary Medicine, Faculty of Medicine, University of Zurich, Building 42, Floor G, 15 Room 70, Winterthurerstr. 190, 8057, Zurich, Switzerland 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 * Corresponding author. 34 E-mail address: [email protected] 35 Telephone: +61 412281013 1 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.08.425868; this version posted January 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. 36 Abstract 37 In modern humans, facial soft tissue thicknesses have been shown to covary with craniometric 38 dimensions. However, to date it has not been confirmed whether these relationships are shared with 39 non-human apes. In this study, we analyze these relationships in chimpanzees (Pan troglodytes) with 40 the aim of producing regression models for approximating facial soft tissue thicknesses in Plio- 41 Pleistocene hominid individuals. Using CT scans of 19 subjects, 637 soft tissue, and 349 craniometric 42 measurements, statistically significant multiple regression models were established for 26 points on 43 the face and head. Examination of regression model validity resulted in minimal differences between 44 observed and predicted soft tissue thickness values. Assessment of interspecies compatibility using a 45 bonobo (Pan paniscus) and modern human (Homo sapiens) subject resulted in minimal differences 46 for the bonobo but large differences for the modern human. These results clearly show that (1) soft 47 tissue thicknesses covary with craniometric dimensions in P. troglodytes, (2) confirms that such 48 covariation is uniformly present in both extant Homo and Pan species, and (3) suggests that chimp- 49 derived regression models have interspecies compatibility with hominids who have similar 50 craniometric dimensions to P. troglodytes. As the craniometric dimensions of early hominids, such as 51 South African australopithecines, are more similar to P. troglodytes than those of H. sapiens, 52 chimpanzee-derived regression models may be used for approximating their craniofacial anatomy. It 53 is hoped that the results of the present study and the reference dataset for facial soft tissue thicknesses 54 of chimpanzees it provides will encourage further research into this topic. 55 56 57 58 59 60 61 62 Keywords: Evolution; Facial reconstruction; Forensic anthropology; Primate morphology. 2 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.08.425868; this version posted January 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. 63 Introduction 64 The primate family of Hominidae is comprised of the African apes, humans and all 65 ancestors leading to these clades. Reconstructing soft tissue characters of extinct members of 66 the Hominidae, called here hominids, has become an increasingly popular practice with many 67 approximations of their faces presented in museum exhibitions, popular science publications 68 and at conference presentations worldwide (1-3). In these contexts, reconstructions of the 69 face and body have proven to be an effective vehicle for the dissemination of scientific 70 information about human evolution. However, there is a recognized problem of variability 71 among reconstructions of the same individual. A recent study comparing approximations of 72 LB1, the holotype of Homo floresiensis (4, 5), reported that they vary significantly among 73 one another (6). Similarly, in a systematic survey of 860 hominid reconstructions presented in 74 71 museums across Australia and Europe, it was found that inconsistencies are clearly 75 prevalent in all other approximations of extinct hominid species (7). If practitioners were 76 using reliable methods, this variability would not have occurred. Either the practitioners 77 relied on unspecified sets of assumptions and/or the methods they used were unreliable. 78 Furthermore, in its present state, the practice of hominid reconstructions is particularly 79 vulnerable to attack from critics of human evolution who have used discrepancies between 80 reconstructions of the same individual to undermine the reliability of evolutionary theory. 81 Therefore, it is important to strengthen existing methods of reconstruction as much as 82 possible to reduce this variability and avoid such criticisms. 83 The term used to describe the process of building a face over a skull varies in the 84 literature between disciplines. In forensics, the name of the process most commonly referred 85 to as ‘facial reconstruction’ was updated to ‘facial approximation’ because it is a more 86 accurate description of the results, whereas in paleoanthropology the term ‘facial 87 reconstruction’ is still being used (2). Regardless of what term is preferred, the results are 3 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.08.425868; this version posted January 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. 88 always approximate and therefore we agree with previous authors and prefer the term ‘facial 89 approximation’ (6, 8, 9). Scientific testing of facial approximation methods has been a major 90 focus in craniofacial identification of human remains with research dating back at least to 91 Welcker (10), with important contributions by Gerasimov (11, 12), Prag and Neave (13), and 92 Wilkinson (14). Methods using means of soft tissues of the face have received the most 93 attention (15-18), however, there is a recognized flaw in extrapolating means to individuals. 94 As statistically robust as means may be, they only express means for specific populations. 95 For reconstructing individuals, population means are not appropriate because they completely 96 ignore variation among individuals. Regarding the approximation of extinct hominids, 97 interspecies extrapolation of means derived from either modern humans or the extant great 98 apes (Gorilla, Pan, and Pongo), as suggested in Hanebrink (17), is equally inappropriate 99 because it also ignores variation among individuals. 100 One possible solution to this problem is to identify approximation methods that are 101 compatible across all members of the Hominoidea superfamily. If a consistent pattern in 102 covariation between soft tissue and craniometric measurements can be identified in extant 103 hominids, then extinct hominids can reasonably be assumed to have followed suite. Such 104 covariations were first explored in human material by Sutton (19) and extended in Simpson 105 and Henneberg (20). Correlations were found and multiple linear regression models were 106 used to generate equations for improving estimations of soft tissue thickness from 107 craniometrics alone in modern humans, though this covariance has rarely been used in facial 108 approximations. Reactions to the results of these studies are mixed. Stephan and Sievwright 109 (21), using data measured with substantial random errors, report that regression models have 110 low correlation coefficients that do not improve soft tissue thickness estimates above 111 population means. However, Dinh, Ma (22) repeated the use of linear regression models and 112 produced favorable results that encourage further exploration. Thus, for the purpose of 4 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.08.425868; this version posted January 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. 113 hominid facial approximation hominids, the possibility of generating soft tissue thickness 114 values that are individualized to a specific hominid specimen is undoubtably better than 115 extrapolation of species-specific means. 116 The present study is motivated by the aforementioned concerns and while we hold 117 that the findings reported here are valuable we raise three caveats at the outset: 1) As in 118 previous studies of chimpanzee soft tissues (23-25), this study includes only a small sample 119 of chimpanzees and therefore the conclusions from the results are subject to further testing on 120 larger samples; 2) We also do not include other members of the African ape clade and so we 121 cannot expand our findings to the entire Hominoidea superfamily; and 3) We do not claim to 122 eliminate the need for informed speculation in hominid facial approximation entirely. Not all 123 soft tissue characters of ancient hominids are addressed here, such as the facial features (eyes, 124 nose, mouth, and ears) that arguably have a much greater impact on the variability between 125 reconstructions of the same individual than soft tissue thicknesses alone.