This is the accepted version of the article: Wennemann, Sophie E.; Lewton, Kristi L.; Orr, Caley M.; [et al.]. «A geometric morphometric approach to investigate primate proximal phalanx diaphysis shape». American Journal of Biological Anthropology, (December 2021), p. 1-22. DOI 10.1002/ajpa.24460 This version is available at https://ddd.uab.cat/record/250225 under the terms of the license Accepted Manuscript for American Journal of Biological Anthropology - December 2021 1 A geometric morphometric approach to investigate primate proximal phalanx diaphysis 2 shape. 3 4 Running title: Phalangeal curvature shape 5 6 Sophie E. Wennemann 1, Kristi L. Lewton 1, 2, Caley M. Orr 3, 4, Sergio Almécija 5, 6, 7, Matthew 7 W. Tocheri 8, 9, 10, William L. Jungers 11, 12, Biren A. Patel 1, 2, * 8 1 9 Department of Integrative Anatomical Sciences, Keck School of Medicine, University of 10 Southern California, Los Angeles, CA 90033, USA 2 11 Human and Evolutionary Biology Section, Department of Biological Sciences, University of 12 Southern California, Los Angeles, CA 90089, USA 3 13 Department of Cell and Developmental Biology, University of Colorado School of Medicine, 14 Aurora, CO 80045, USA 4 15 Department of Anthropology, University of Colorado Denver, Denver, CO 80217, USA 5 16 Division of Anthropology, American Museum of Natural History, Central Park West at 79th 17 Street, New York, NY 10024, USA 6 18 New York Consortium in Evolutionary Primatology, New York, NY, USA 7 19 Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona, c/ 20 Columnes s/n, Campus de la UAB, 08193, Cerdanyola del Vallès, Barcelona, Spain 8 21 Department of Anthropology, Lakehead University, Thunder Bay, Ontario P7B 5E1, Canada 9 22 Human Origins Program, National Museum of Natural History, Smithsonian Institution, 23 Washington DC 20013, USA 10 24 Australian Research Council Centre of Excellence for Australian Biodiversity and Heritage, 25 University of Wollongong, Wollongong, New South Wales, 2522, Australia 11 26 Department of Anatomical Sciences, Stony Brook University School of Medicine, Stony 27 Brook, NY 11794, USA 12 28 Association Vahatra, BP 3972, Antananarivo 101, Madagascar 29 30 * Corresponding author: Biren A. Patel, 1333 San Pablo Street, BMT 404, Keck School of 31 Medicine, University of Southern California, Los Angeles CA, 90033, USA; Email address: 32 birenpat@usc.edu 33 34 Keywords: phalanx; curvature; geometric morphometrics; included angle; suspensory 35 36 Funding Statement 37 Funding for this research was generously provided by: National Science Foundation (BCS 38 #1317047 to B.A.P.; BCS #1539741 to C.M.O.); the L.S.B. Leakey Foundation (to B.A.P.); the 39 Wenner-Gren Foundation (to C.M.O.); AEI/FEDER EU (CGL2017-82654-P to S.A.), 40 Generalitat de Catalunya (CERCA Programme to S.A.); the University of Southern California’s 41 URAP program (to B.A.P.), and the University of Southern California’s Office of the Provost (to 42 S.E.W.). 43 44 Abstract 45 Current approaches to quantify phalangeal curvature assume that the long axis of the bone's 46 diaphysis approximates the shape of a portion of a circle (included angle method) or a parabola 47 (second-degree polynomial method). Here we developed, tested, and employed an alternative 48 geometric morphometrics-based approach to quantify diaphysis shape of proximal phalanges in 49 humans, apes and monkeys with diverse locomotor behaviors. 100 landmarks of the central 50 longitudinal axis were extracted from 3D surface models and analyzed using 2DGM methods, 51 including Generalized Procrustes Analyses. Principal components analyses were performed and 52 PC1 scores (>80% of variation) represented the dorsopalmar shape of the bone’s central 53 longitudinal axis and separated taxa consistently and in accord with known locomotor behavioral 54 profiles. The most suspensory taxa, including orangutans, hylobatids and spider monkeys, had 55 significantly lower PC1 scores reflecting the greatest amounts of phalangeal curvature. In 56 contrast, bipedal humans and the quadrupedal cercopithecoid monkeys sampled (baboons, 57 proboscis monkeys) exhibited significantly higher PC1 scores reflecting flatter phalanges. 58 African ape (gorillas, chimpanzees and bonobos) phalanges fell between these two extremes and 59 were not significantly different from each other. PC1 scores were significantly correlated with 60 both included angle and the a coefficient of a second-degree polynomial calculated from the 61 same landmark dataset, but had a significantly higher correlation with included angles. Our 62 alternative approach for quantifying diaphysis shape of proximal phalanges to investigate 63 dorsopalmar curvature is replicable and does not assume a priori either a circle or parabola 64 model of shape, making it an attractive alternative compared with existing methodologies. 65 66 1. Introduction 67 The hand skeletons of living primates are morphologically diverse and reflect the unique 68 evolutionary histories of various primate lineages as well as the ways that primates use their 69 hands for manipulation and locomotion (e.g., Napier, 1993; Kivell et al., 2016). For example, 70 proximal phalanges vary greatly within and among extant primate taxa in terms of how curved 71 the shafts are (hereafter referred to as phalangeal curvature; Fig. 1). Highly suspensory primates 72 (e.g., Pongo, Hylobates, Symphalangus, and Ateles) have very curved proximal phalanx shafts 73 (i.e., high phalangeal curvatures), whereas habitually terrestrial pronograde species (e.g., Gorilla 74 and Papio) along with bipeds (i.e., humans) have straighter shafts (i.e., low phalangeal 75 curvatures) (e.g., Stern, Jungers & Susman, 1995; Dean & Begun, 2008; Rein, 2011). Primates 76 that are more generalist in their substrate preference (e.g., Pan and Macaca) tend to have 77 phalangeal curvature that is intermediate or more variable in grade compared with these two 78 extremes (e.g., Jungers et al., 1997). These observed differences in phalangeal curvature are 79 behaviorally linked to how primates use their hands when moving. Furthermore, theoretical 80 models (Preuschoft, 1973b), along with both in vitro and in silico experimental biomechanical 81 studies (Richmond, 2007; Nguyen et al., 2014) have shown that phalangeal curvature in primates 82 is in large part an adaptation to the habitual mechanical loading environment that arises from 83 substrate reaction forces, muscle forces from the contraction of extrinsic digital flexors, and joint 84 reaction forces at the both the metacarpophalangeal and interphalangeal joints when the hands 85 are used during locomotor and positional behaviors. Additionally, there is a phylogenetic 86 component to phalangeal curvature variation among living primates which is likely a result of a 87 long history of strong selection for attenuating these mechanical loads on relatively mechanically 88 weak bones that are also important for manipulation in addition to locomotion (e.g., Wallace, 89 Burgess & Patel, 2020; but see Richmond, 1998 and Jungers et al., 2002 for alternative 90 hypotheses). Because of the strong form-function relationship between phalangeal curvature and 91 locomotion in extant primates, paleoanthropologists typically use phalangeal curvature to infer 92 the positional and locomotor behavior of fossil primate species including hominins (Preuschoft, 93 1973a; Stern & Susman, 1983; Susman, Stern & Jungers, 1984; Hamrick, Meldrum & Simons, 94 1995; Susman, de Ruiter & Brain, 2001; Kivell et al., 2011, 2015; Domínguez-Rodrigo et al., 95 2015; Stratford et al., 2016; Prang et al., 2021). 96 97 Figure 1 Here 98 99 Different methods have been used to quantitatively assess phalangeal curvature in extant 100 and fossil primates. The most prevalent approach has been to calculate the bone’s included angle 101 (IA), which assumes that the central longitudinal (i.e., proximodistal) axis of a phalanx 102 represents a shape that corresponds to a portion of an arc on the perimeter of a circle (Fig. 2A; 103 Stern, Jungers & Susman, 1995). Larger values of IA, measured in degrees, indicate bones that 104 are more curved whereas smaller values indicate bones that are flatter. In general, this method 105 does well to distinguish extant species based on habitual locomotor behavior, with the most 106 suspensory primate species having larger IAs and more terrestrial species (both bipedal and 107 pronograde quadrupedal) having smaller IAs (e.g., Rein, 2011). Despite the relatively high 108 accuracy of this form-function relationship in extant primates, some authors (e.g., Patel & 109 Maiolino, 2016) have noted that published IA values can differ substantially between studies 110 (i.e., high inter-study variation). For example, IA mean (and one standard deviation) values 111 reported for Pan troglodytes by Stern, Jungers & Susman (1995), Matarazzo (2008), and Rein 112 (2011) are 42.4° (4.8°), 55.0° (4.4°) and 38.4° (5.3°), respectively. Even more variable are IA 113 mean (and one standard deviation) values reported for Macaca fascicularis, where Matarazzo 114 (2008), Rein (2011), and Patel & Maiolino (2016) reported values of 45.0° (6.6°), 16.2° (6.6°) 115 and 22.0° (7.0°), respectively. With such a large amount of variation exhibited in IA values for a 116 single species between studies, and researchers often being unable to independently collect IA 117 data for a comparative sample or when describing new fossils (e.g., Stratford et al., 2016), it is 118 unclear on which dataset researchers should rely, if any, when trying to make functional 119 inferences about fossil specimens. 120 121 Figure 2 Here 122 123 Reasons for such inter-study discrepancies are unclear, but they could simply be related 124 to different sample compositions (i.e., different museum collections). Alternatively, they could 125 derive from the different tools used to acquire raw data such as handheld coordinate calipers 126 (Susman, Stern & Jungers, 1984; Stern, Jungers & Susman, 1995), Microscribe point digitizers 127 (Rein, 2011), three-dimensional (3D) landmarking software (Patel & Maiolino, 2016; Wallace, 128 Burgess & Patel, 2020), or lateral view photographs (e.g., Richmond, 1998; Jungers et al., 2002; 129 Burgess, 2018; Prang et al., 2021). In fact, some of these tools might be more prone to error than 130 others. For example, there can be movement artifacts when applying a Microscribe point 131 digitizer on small specimens (B.A.P. personal observation) and photographs can suffer from 132 parallax issues if the specimen is not positioned properly relative to the lens of the camera 133 (Spencer & Spencer, 1995; Richmond, 1998). Another overlooked problem of lateral view 134 photographs of proximal phalanges is that important morphological features needed to calculate 135 IA of the central longitudinal axis can be obscured in the field of view (Fig. 1). One example of 136 this is the presence of well-developed flexor sheath ridges (as found in Gorilla) that can 137 incorrectly inflate measurements of midshaft dorsopalmar height thus resulting in smaller 138 calculated IA values (see equations and figures in Stern, Jungers & Susman, 1995; Jungers et al., 139 1997). In fact, low values of IA for Gorilla calculated from lateral view photographs often 140 overlap extensively with human IA values (e.g., Prang et al., 2021), which would not be 141 predicted based on biomechanical differences in hand use during locomotion between humans 142 and gorillas. Another example is the presence of a very concave proximal articular surface (as 143 seen in many monkeys like Papio that have a large range of dorsiflexion at the 144 metacarpophalangeal joint) that when not fully visible can lead to erroneous longer 145 proximodistal interarticular length measurements resulting in larger IAs. Hence, it is important to 146 acknowledge that end users (e.g., paleoanthropologists) relying on published IA data for 147 comparisons with fossil phalanges should proceed with caution (e.g., always use a single source 148 of comparative data and follow its same methodology for taking measurements) or should only 149 use original self-collected data. 150 Some researchers have argued against using IAs for assessing phalangeal curvature 151 claiming that it may be theoretically flawed. Specifically, Deane, Kremer & Begun (2005) 152 argued that the diaphyses of primate proximal phalanges, in particular the dorsal surfaces, look 153 more like parabolas or a second-degree (2°) polynomial1 rather than portions of a circle and 154 emphasized that the IA is an inappropriate metric to quantify phalangeal curvature (but see 155 contrary arguments made by Stern, Jungers & Susman, 1995). A strong visual case for this 156 opinion can be made by looking at the mid-sagittal plane of a proximal phalanx from taxa that 157 have relatively straight bones like cercopithecine monkeys (e.g., Papio and Nasalis) and humans 158 (Fig. 1). Accordingly, Deane, Kremer & Begun (2005) devised a High-Resolution Polynomial 159 Curve Fitting method (hereafter referred to as PCF) that uses a lateral view photograph of a 160 proximal phalanx to fit a second-degree polynomial function (defined as y=ax2+bx+c) to an 161 extensive number of two-dimensional (2D) landmarks placed on the dorsal surface of the 162 diaphysis (Fig. 2B). The resulting a coefficient (or quadratic coefficient) is then used as a proxy 163 metric for curvature where a larger a value is equivalent to a narrower parabola and signifies a 164 more curved bone. In general, this method can distinguish extant species based on habitual 165 locomotor behavior, with suspensory orangutans having larger a values and bipedal humans 166 having smaller values for a (Deane, Kremer & Begun, 2005; Deane & Begun, 2008). A second 167 argument subsequently made by Deane & Begun (2008) against using IA was that it has limited 168 paleoanthropological utility because a large number of fossil phalanges are broken and only 169 complete bones can be used to calculate IA correctly. These authors suggested that a benefit of 170 using PCF is that it allows for the inclusion of more fossils in comparative analyses than would 171 be possible if using IA alone. Although the use of this method has not been widespread in 172 paleoanthropological research, it has recently been applied in studies of early hominin and 173 Miocene hominoid fossil hand bones (Kivell et al., 2011, 2015; Böhme et al., 2019) (Fig. 2C). 174 Despite the reasonable arguments against implementing the IA method, however, it 175 continues to be used (e.g., Stratford et al., 2016; Wallace, Burgess & Patel, 2020; Prang et al., 176 2021), most likely because the PCF method also has some practical disadvantages. The first is 177 that as currently implemented PCF measures the curvature of the dorsal surface of the bone 178 (Deane, Kremer & Begun, 2005) rather than the more biomechanically relevant central 179 longitudinal axis that better reflects overall adaptation for attenuating mechanical stresses 180 experienced by the bone during grasping (Preuschoft, 1973b; Richmond, 2007; Nguyen et al., 1 Note that the merits of a second-degree polynomial specifically for proximal phalanges over some other higher order polynomial function, such as third-sixth-degree, are discussed in detail by Deane, Kremer & Begun (2005) and are not repeated here. See also Zhang, Harrison & Ji (2020) who discuss the merits of a fourth-degree polynomial compared to a second-degree polynomial for proximal phalangeal curvature. 181 2014). Second, more taxonomically inclusive studies using PCF have shown inconsistent results 182 regarding known form-function relationships in extant primates. Most notably, gibbons, which 183 are among the most suspensory living primates, have a mean and range of a values that indicate a 184 degree of curvature that is equal to some generalized arboreal quadrupedal monkeys (e.g., 185 Cercopithecus and Alouatta) and overlap terrestrial quadrupedal knuckle-walking African apes 186 (Gorilla and Pan) (Böhme et al., 2019). 187 In light of these issues with using both IA and PCF, the current study investigates the 188 suitability of an alternative method to assess diaphysis shape of primate proximal phalanges. 189 Specifically, we implemented geometric morphometrics (GM) methods to determine bone shape 190 that is both biomechanically informative and reflects the dorsopalmar curvature of the central 191 longitudinal (proximodistal) phalangeal axis. The use of GM to assess shape reflective of 192 diaphysis curvature is not new and has been applied to other postcranial elements such as the 193 femur, clavicle, and ulna (De Groote, Lockwood & Aiello, 2010; Squyres & DeLeon, 2015; 194 Milne & Granatosky, 2021). In the current study, our alternative approach uses 3D digital 195 surface models of proximal phalanges on which a semi-automated software workflow with 196 minimal user input is applied to obtain landmarks that can be subjected to established GM 197 methods. The utility of this GM approach is then assessed by comparing 2D shape by means of 198 Principal Components (PC) scores among different extant anthropoid species with known 199 locomotor behaviors. Finally, resulting PC scores reflecting diaphysis dorsopalmar curve shape 200 are compared with curvature assessed from the same landmark datasets using the IA and PCF 201 methods. 202 203 2. Materials and Methods 204 2.1 Comparative Sample 205 The comparative sample of 378 manual proximal phalanges of the third digit (mPP3) 206 used in this study includes modern humans, chimpanzees, bonobos, gorillas, orangutans, 207 siamangs, gibbons, baboons, proboscis monkeys, and spider monkeys (Table 1). Phalanges from 208 digits 2, 4, and 5 (mPP2, mPP4, and mPP5) were also studied for a smaller subset of these 209 primates (see below). These taxa were chosen because their habitual locomotor behaviors are 210 relatively well known and have been quantified by primate behaviorists. Moreover, they 211 represent taxa that engage in differing amounts of suspensory and quadrupedal locomotion, as 212 well as differing preferences for arboreal and terrestrial substrates (e.g., Rein [2011] and 213 Granatosky [2018] and references therein). All specimens except for one gorilla and one spider 214 monkey were collected in the wild based on museum records. The sample includes females and 215 males as well as some where the sex is unknown; all of these were combined to maximize 216 intraspecific variation, and to make the final datasets more applicable for future broader 217 paleoanthropological studies. Most specimens were skeletal adults and had fully fused proximal 218 epiphyses; a handful of non-human subadult specimens for some taxa were included if the 219 epiphysis was in the process of fusion and could not be manually separated from the diaphysis 220 (see Supplemental Information and Figure S1 for additional considerations on adult vs. subadult 221 sampling). The studied specimens derive from the following collections: AMNH, American 222 Museum of Natural History (New York, NY, USA); ANSP, Academy of Natural Sciences of 223 Drexel University (Philadelphia, PA, USA); LACM, Natural History Museum of Los Angeles 224 County (Los Angeles, CA, USA); MCZ, Harvard University Museum of Comparative Zoology 225 (Cambridge, MA, USA); RMCA, Royal Museum for Central Africa (Tervuren, Belgium); SBU, 226 Department of Anatomical Sciences of Stony Brook University (Stony Brook, NY, USA); USC, 227 Department of Integrative Anatomical Sciences of the Keck School of Medicine of the 228 University of Southern California (Los Angeles, CA, USA); USNM, Smithsonian Institution’s 229 National Museum of Natural History (Washington, DC, USA); WITS, University of the 230 Witwatersrand (Johannesburg, South Africa), and the Mountain Gorilla Skeletal Project 231 (MGSP). Additional details of the comparative sample are found in Table 1. 232 233 Table 1 Here 234 235 2.2 Surface Data 236 3D surface models of mPP3s were generated from computed tomography scans (CT), 237 micro-CT scans (µCT), or laser surface scans following protocols detailed elsewhere (see 238 Fernandez et al., 2015; Patel et al., 2017; Fig. 1). A number of studies have previously shown 239 that different scanning modalities has little to no effect on most morphometric results (e.g., 240 Tocheri et al., 2011; Robinson & Terhune, 2017; Shearer at et al., 2017). CT scans were obtained 241 with slice intervals up to 0.187 mm (with scanning field of views that produced near isotropic 242 voxels). µCT scans were obtained with isometric voxel dimensions up to 0.080 mm. The CT and 243 µCT scan data were segmented on unfiltered 16-bit DICOM (.dcm) or TIFF (.tif) images using 244 the SEGMENTATION EDITOR in Amira v.5.6 (or later) software (Thermo Fisher Scientific) by 245 means of a combination of automated and slice-by-slice manual thresholding techniques, after 246 which triangular surface meshes were created using the SURFACE GEN [or GENERATE SURFACE in 247 more recent versions of Amira] function (with ‘unconstrained smoothing’ and ‘compactify’ 248 options implemented) and saved as POLYGON mesh files (.ply). All laser scans were obtained 249 with a NextEngine 3D laser scanner (NextEngine, Inc.) using a resolution of >10,000 points per 250 square inch; 6 to 12 scans were taken at different positions and then merged using ScanStudio 251 HD PRO software (NextEngine, Inc.) and saved as .ply files. 252 The analytical procedures developed and employed in this study can potentially be 253 sensitive to the quality of the 3D surface meshes used. Therefore, any scanning defects (e.g., 254 spikes, floating triangles, holes) in the mesh surface models created from any of the scanning 255 modalities were corrected using Geomagic Wrap software (3D Systems, Inc.). Additionally, any 256 biological holes (i.e., foramina) in the diaphysis from nutrient arteries were also virtually 257 removed. Each surface mesh was then made hollow to include only an external watertight shell 258 of triangles, which was then either decimated or refined to the same approximate size of 250,000 259 triangles (see Supplemental Information and Figure S2 for additional considerations about 260 preferred surface mesh triangle counts). All subsequent procedures and data acquisition were 261 performed on right side proximal phalanges; if the original specimen was from the left side, it 262 was first mirror-imaged in Amira or Geomagic Wrap to create an antimere. 263 264 2.3 Landmark Data Acquisition 265 The data acquisition methodology developed and used in this study relied on Amira and 266 ImageJ software (Schneider, Rasband & Eliceire, 2012); therefore, the terminology used here 267 reflects these software packages only. When possible, we provide names of specific functions, 268 modules, and plugins for each program that we implemented so that other users who prefer 269 different software (e.g., Avizo, Meshlab, MATLAB, Python, R, etc.) can replicate these methods 270 in the software package of their choice. 271 It was necessary to first position all surface models in the same anatomical orientation to 272 acquire equivalent landmarks from near the center of the diaphysis of a phalanx (Fig. 3A). 273 Rather than using previously published computational routines that facilitate automated 274 alignment of 3D surfaces (e.g., Boyer et al., 2015), we opted for a manual approach using a 275 standardized protocol in Amira such that: 1) a line connecting the centers of the proximal and 276 distal articular surfaces in dorsal view (XZ) was aligned vertically; 2) the palmar-most aspect of 277 the proximal palmar tubercles and the palmar-most aspect of the distal trochlea in lateral view 278 (YZ) were aligned on the same plane and positioned vertically; and 3) the palmar-most aspect of 279 the proximal palmar tubercles in proximal view (XY) were aligned on the same plane and 280 positioned horizontally. The initial orientation of specimens was facilitated using the ALIGN 281 PRINCIPAL AXES module (which aligns the virtual bone according to its three principal axes; Axis 282 1 was set to ‘x’ and Axis 2 was set to ‘y’), and this was followed by minor manual movements 283 (facilitated by a checkerboard background) using the TRANSFORM EDITOR function (see below 284 for error analyses for intra-observer differences in determining a standardized anatomical 285 orientation). We refer to this as our ‘gold standard’ orientation from here forward. 286 287 Figure 3 Here 288 289 The gold standard orientation allows for subsequent re-slicing of the virtual phalanx to 290 ultimately obtain cross sections that are oriented orthogonal to its long (proximodistal) axis. To 291 re-slice a surface mesh of triangles, the mesh must first be converted to a volumetric 292 representation with user defined voxel dimensions. This was accomplished using the SCAN 293 CONVERT SURFACE (or SCAN SURFACE TO VOLUME in more recent versions of Amira) module, a 294 tool that voxelizes (i.e., 3D pixilation) the surface mesh resulting in a binary label field (i.e., set 295 of slices) of known dimensions. To create a new binary label field with enough resolution, we 296 used a z-pixel dimension of 1000 while keeping the model’s original x- and y-pixel dimensions 297 unchanged (thereby yielding non-isotropic voxels in the resulting volume representation). For 298 our purposes, this procedure effectively creates 1000 new binary image slices from proximal to 299 distal ends of the bone which can be further cropped and further resampled (as discussed below) 300 (Fig. 3B). 301 Because our goals are to identify shape reflecting dorsopalmar curvature of the most 302 biomechanically relevant region of the diaphysis (i.e., non-articular regions), a new region of 303 interest of the diaphysis (dROI) was defined between the palmar-most aspect of the proximal 304 palmar tubercles and the inflection point between the distal trochlea and the diaphysis on the 305 palmar side (Fig. 3C; see Supplemental Information and Figure S1 for further discussion on 306 dROIs of subadults with partial fusion of the proximal articular surfaces). The region proximal to 307 the palmar tubercles and the entire distal trochlea was then cropped out using the CROPEDITOR 308 function from the binary label field thereby resulting in a new image stack with less than 1000 309 slices. Due to differences in absolute phalangeal length as well as the size and shape of the 310 trochlea and proximal ends among and within primate species (Fig. 1; see also Patel & Maiolino, 311 2016), the cropped binary label field was further resampled so that each specimen in the analysis 312 yielded the same number of binary images; this step could not simply be implemented 313 simultaneously with the previous one because of technical limitations in Amira. This second 314 phase in resampling was accomplished by first generating a new 3D surface (using the SURFACE 315 GEN function with the default ‘add border’ option not implemented), and then again using the 316 SCAN CONVERT SURFACE function to voxelize and resample with a z-pixel dimension of 100 317 while again keeping the model’s original x- and y-pixel dimensions unchanged. The new binary 318 image stack of 100 cross section images was then saved in .tif format. Note that outputting a 319 stack of 100 TIFF images within Amira also produces an important accompanying metadata file 320 (“.info”) that contains real world dimensions (in mm) for pixel size (x and y) and slice position 321 (z) from proximal to distal for the resulting set of images. This resolution information is needed 322 after raw landmark coordinates are obtained to keep proper scaling of the raw data in real world 323 units (see below). 324 The new stack of 100 binary images, each representing a cross section from proximal to 325 distal of the reoriented phalanx's dROI, was then imported into ImageJ in order to find the cross- 326 section’s approximate center (Fig. 3D). While a number of approaches are possible to 327 accomplish this, we chose a method that allows for both automation and replication with no user 328 input, and an approach that we inferred to allow us to identify the approximate center between 329 the dorsal and palmar surfaces of all slices independently. Specifically, we applied a modified 330 looped version of the MAX INSCRIBED CIRCLES plugin written by Burri & Guiet (2016) and 331 available from the PTBIOP Update Site for ImageJ, which implements the largest inscribed 332 circle algorithm following code initially written by Birdal (2021) in MATLAB software 333 (MathWorks). The looped plugin runs sequentially through the stack of 100 dROI cross sections 334 finding a specific circle-shaped region of interest in each image, from which the x- and y- 335 coordinates of their centers can be calculated using the CENTER OF MASS measurement function 336 in ImageJ (Fig. 3D). Using the known pixel resolution and slice spacing information from the 337 .info file outputted from Amira (see above), these center x-, y-, and z-coordinates are then scaled 338 back into real world dimensions (in mm). These 100 semi-landmarks are then subjected to 339 geometric morphometric analyses. 340 341 2.4 Geometric Morphometrics 342 Phalanges have rather complex shapes with varying amounts of shaft torsion, flexor 343 sheath ridge development, and general mediolateral asymmetry (Patel & Maiolino, 2016; see 344 also Figs. 1 and 3D). Therefore, the position of a cross section’s center can vary in the 345 mediolateral (x-) direction when taking these features into account, whereas the dorsopalmar (y-) 346 position of the center changes negligibly (i.e., less noisy). Accordingly, while the resulting 347 output of the previous steps produced 3D landmarks, we chose to pursue a 2D geometric 348 morphometric (2DGM) analysis since our primary concern is determining the dorsopalmar 349 longitudinal shape of the phalanx’s diaphysis (i.e., its dROI). Moreover, using a 2DGM approach 350 also has the added benefit of being able to combine samples of right and left side bones, which is 351 particularly useful when studying fragmentary fossil assemblages, or when reliable side 352 determinations of isolated phalanges cannot be made by the researcher. 353 Landmark data files for each specimen exported from Amira consisted of y- and z- 354 coordinates (i.e., dorsopalmar and proximodistal coordinates, respectively). These landmark data 355 files were first manually configured to Morphologika format to facilitate import into R software 356 (R Core Team, 2019) for analysis using the GEOMORPH package (Adams, Collyer & 357 Kaliontzopoulou, 2019). Raw landmark configurations were scaled, translated, and rotated using 358 Generalized Procrustes Analysis (GPA), which produced a new set of coordinates (Procrustes 359 coordinates) that were then used for subsequent statistical analyses. A number of intra- and inter- 360 observer error analyses were first performed using Procrustes distances resulting from the 361 GPAs. Principal component analyses (PCA) were then performed to assess variance in shape via 362 resulting principal component (PC) scores. It was determined that the first two PCs typically 363 accounted for >80% of the variation in all of the 2DGM analyses implemented. Moreover, PC1 364 accounted for nearly all the shape variation of the phalanx’s central longitudinal dorsopalmar 365 curvature (see Figure S3) and thus we placed our primary emphasis on these values in the results 366 of subsequent analytical tests of utility and comparability. All statistical analyses discussed in the 367 following paragraphs were performed in R software, JMP v.14 software (SAS Institute), or 368 PAST software (Hammer, Harper & Ryan, 2001). The R script as implemented for all analyses is 369 available in the Supplemental Information. Landmark data files in Morphologika format for the 370 comparative sample of humans, apes, and monkeys are available on the journal's website. 371 372 3. Results and Discussion 373 3.1 Error 374 To assess the effects of intra- and inter-observer error, we performed three analyses. The 375 first analysis for intra-observer error focused specifically on the specimen orientation protocol. 376 The second analysis for intra- and inter-observer error examined the specimen dROI 377 determination protocol. The final analysis for both intra- and inter-observer error evaluated all 378 steps of the protocol. 379 3.1.1 Specimen orientation: To systematically investigate the effect of differences in 380 specimen orientation in a realistic way, one observer (S.E.W.) performed a controlled sensitivity 381 analysis in which one orangutan specimen (Pongo abelii AMNH-143598) was used to create 60 382 model variants that deviated 10 degrees (in 1 degree increments) from our gold standard model 383 in six possible directions about this specimen’s centroid: dorsal, palmar, medial, lateral, 384 clockwise, counterclockwise (see insert in Fig. 4). We then performed all the landmark 385 acquisition and 2DGM procedures outlined above on these model variants without altering any 386 other parameters or inputs. Procrustes distances between the gold standard specimen and the 60 387 model variants were plotted against change in degrees in reference to the gold standard specimen 388 to visually assess any relationship between model orientation and diaphysis shape (Fig. 4). 389 Procrustes distances appear to differ inconsequentially when the specimen moved between 1 and 390 10 degrees from the gold standard in medial, lateral, clockwise, and counterclockwise directions. 391 In contrast, Procrustes distances increased almost linearly when the oriented bone was 392 incorrectly positioned in the dorsal and palmar directions from 1 to 10 degrees. Based on these 393 findings, we emphasize the importance to orient specimens carefully prior to subsequent steps 394 for landmark acquisition, especially in the dorsal and palmar directions about the bone’s 395 centroid. 396 397 Figure 4 Here 398 399 3.1.2 Specimen dROI determination: The cropping procedure to determine specimen 400 dROI is also subject to both intra- and inter-observer error. Therefore, we performed an error 401 study to test for both in which two observers (S.E.W. and B.A.P.) took one orangutan specimen 402 (Pongo abelii AMNH-143598) and performed the cropping steps described above 10 times over 403 a period of several days. We then performed all the landmark acquisition procedures outlined 404 above without altering any other parameters or inputs. A 2DGM analysis was then performed on 405 the resulting 20 landmark datasets (i.e., 10 landmark set replicates from two different observers), 406 along with 16 additional Pongo abelii specimens (for a total n=36 individuals in the analysis; the 407 latter 16 were quantified only by S.E.W.). The additional 16 Pongo abelii specimens were 408 included to compare variation in Procrustes distances within each observer's and among the two 409 observers' replicates (n=45 and n=100 comparisons, respectively), and to determine how these 410 compare to variation in Procrustes distances between other specimens of the same species 411 (n=136 comparisons). Box-and-whiskers plots (Fig. 5) demonstrate that intra- and inter-observer 412 Procrustes distances for the test specimen (Pongo abelii AMNH-143598) are significantly 413 smaller than the Procrustes distances among all specimens of Pongo abelii.2 These findings 414 indicate that any minor differences in cropping within and between observers is fundamentally 415 negligible in larger scale comparisons within taxa (Fig. 5) and among taxa (see below). 416 417 Figure 5 Here 418 419 3.1.3 Complete protocol: Intra- and inter-observer error for the entire protocol was 420 evaluated by two observers (S.E.W. and B.A.P.) using one orangutan specimen (Pongo abelii 421 AMNH-143598). Both observers obtained the necessary landmark datasets 10 times over a 422 period of one month. A 2DGM analysis was then performed on the resulting 20 landmark 423 datasets (i.e., 10 landmark set replicates from two different observers), along with 16 additional 424 Pongo abelii specimens (for a total n=36 individuals in the analysis; the latter 16 were quantified 425 only by S.E.W.). As noted above, the additional 16 Pongo abelii specimens were included to 426 compare variation in Procrustes distances within each observer's and among the two observers' 2 Post-hoc analyses show that variation between and among observers was not significant (p>0.05) but was significant (p<0.05) when compared to total species variation. 427 replicates (n=45 and n=100 comparisons, respectively), and to determine how these compare to 428 variation in Procrustes distances between other specimens of the same species (n=136 429 comparisons). Box-and-whiskers plots (Fig. 6) demonstrate that intra- and inter-observer 430 Procrustes distances for the test specimen (Pongo abelii AMNH-143598) are significantly 431 smaller than the Procrustes distances among all specimens of Pongo abelii.3 Thus, these findings 432 indicate that any minor differences in executing the entire protocol within and between observers 433 is minimal in larger scale comparisons within taxa (Fig. 6) and among taxa (see below). 434 435 Figure 6 Here 436 437 3.2 Inter-digit variation 438 Because an ultimate objective for implementing this method is to evaluate shape of 439 primate fossil phalanges, and not all primate fossil hands have a mPP3 preserved, we performed 440 an analysis examining inter-digit variation by comparing diaphysis dorsopalmar shape across 441 non-pollical proximal phalanges (i.e., mPP2–5). Specifically, we compared the mPP2-5s within 442 a sample of orangutans (Pongo abelii; n=10) that have very curved proximal phalanges and 443 within a sample of humans (Homo sapiens; n=12) that have straighter proximal phalanges (Fig. 444 1; Rein, 2011). The two samples were analyzed separately (i.e., subjected to their own 2DGM 445 analyses) rather than being combined so that their resulting PC1 scores would not be influenced 446 by the presence of different taxa with different levels of curvature. For each taxonomic sample, a 447 non-parametric Kruskal-Wallis test and non-parametric Mann-Whitney pairwise comparisons 448 were used to compare digit PC1 scores for the entirety of each species. In addition, we compared 449 PC1 scores by individual (i.e., matched pairs: mPP2 vs. mPP3, mPP2 vs. mPP4, etc.) within each 450 species sample using non-parametric Wilcoxon Signed Rank tests. Despite some dissimilarities 451 across digits, there were no significant differences (p>0.05) in any of the comparisons in either 452 species (Fig. 7). These results are consistent with findings reported for both IA and PCF methods 453 that also observed no significant differences in phalangeal curvature across hand digits in both 454 hominoid and cercopithecoid primates (Stern, Jungers & Susman, 1995; Deane & Begun 2008). 455 More importantly, and for practical purposes, the lack of significant differences among digits 3 Post-hoc analyses show that variation between and among observers was not significant (p>0.05) but was significant (p<0.05) when compared to total species variation. 456 (when analyzed among individuals or analyzed within an individual) suggests that if necessary, 457 non-pollical proximal phalanges can be combined to increase sample sizes so that the 2DGM 458 methods presented in this study can be employed more broadly. This is particularly useful when 459 studying isolated fossil phalanges that may not necessarily come from the same digit as an 460 existing comparative sample (e.g., from digit 3 as emphasized in this study). 461 462 Figure 7 Here 463 464 3.3 Utility 465 The utility of any novel quantitative method needs to be assessed and we would interpret 466 our method to be useful if the PCAs resulted in group separations along known locomotor 467 behaviors (i.e., with a shape morphocline from most curved to least curved corresponding to 468 most suspensory/arboreal to least suspensory/arboreal and most terrestrial). Therefore, we 469 performed a comprehensive 2DGM analysis on the entire comparative sample of 378 mPP3s of 470 humans, apes, and monkeys (Table 1). The resulting PC scores were visualized using both 471 bivariate scatter plots and box-and-whiskers plots, and interspecific statistical difference was 472 assessed using ANOVA and post-hoc Tukey’s tests for pairwise comparisons. We predicted that 473 the more suspensory orangutans, gibbons, siamangs, and spider monkeys would plot near each 474 other on a given PC score distribution, thereby reflecting proximal phalanges with greater 475 curvature. In contrast, we predicted that humans and baboons would plot near each other on the 476 same PC score distribution reflecting proximal phalanges with less curvature. Finally, we 477 predicted that chimpanzees would be closer to the suspensory taxa (i.e., with more curved shape) 478 whereas gorillas and proboscis monkeys would be closer to the non-suspensory taxa (i.e., less 479 curved shape) of the resulting PC score distributions. In addition to comparing PC scores, species 480 average diaphysis shapes based on Procrustes coordinates were visualized and compared to the 481 sample mean shape. 482 The first two principal components accounted for ~92% of the total observed variation 483 (85.3% for PC1 and 6.6% for PC2). Variation along PC1 tracks degree of suspensory behavior: 484 highly suspensory primates were found on the negative side of the PC1 distribution and 485 demonstrate more curved shapes, bipedal humans and pronograde quadrupedal monkeys were 486 found on the positive side of the PC1 distribution and demonstrate flatter shapes, and African 487 apes were in the middle (Figs. 8 & 10; the mean, maximum and minimum shapes for PC1 and 488 PC2 resulting from the 2DGM analysis are illustrated in Figure S3). ANOVAs reveal that there 489 are significant differences in PC1 scores (F=123.00; df=12,365; p<0.0001) among the studied 490 taxa (Fig. 9). Post-hoc analyses (Table 2) of PC1 scores also reveal the following: both Pongo 491 species are significantly different from each other and from all other primates; hylobatids and 492 Ateles do not differ from each other; African apes differ from all other primates examined but not 493 each other; humans and Nasalis are not significantly different from each other; Papio is 494 significantly different from all other primates except Nasalis. 495 To investigate the potential effects of variation in size on shape differences among taxa, 496 we subsequently calculated the correlation between species mean PC1 scores and the natural log 497 (ln) of dROI length (which was computed from the most proximal [#1] and distal [#100] 498 landmarks). While this correlation was significant (r=-0.607; p=0.028), further review reveals 499 that three groups, Pongo abelii and Pongo pygmaeus with absolutely the longest dROIs and 500 Papio spp. with absolutely the shortest dROIs were the primary influencers driving significance 501 in this analysis (see Figure S4A). When these three taxa are removed, and the correlation is 502 performed again (see Figure S4B), it is found to be not significant (r=0.163, p=0.653). Thus, the 503 latter demonstrates that species with similar mPP3 dROI lengths can exhibit quite different PC1 504 scores and therefore levels of phalangeal curvature, further supporting the general hypothesis 505 that differences in mPP3 curvature shape relate to locomotor behavior. 506 ANOVA reveals that there are significant differences among taxa in PC2 scores 507 (F=61.13; df=12,365; p<0.0001, Fig. 9). Post-hoc analyses of PC2 scores also reveal the 508 following: all three Gorilla species have significantly higher PC2 scores than all other taxa; the 509 smaller bodied hylobatids, Ateles, Papio and Nasalis have significantly lower PC2 scores, but do 510 not generally differ among each other; humans do not differ from either species of Pongo or Pan. 511 512 Figure 8 Here 513 Figure 9 Here 514 Figure 10 Here 515 Table 2 Here 516 517 The 2DGM method to quantify proximal phalanx diaphysis shape produces results that 518 reflect dorsopalmar curvature and correspond with variation in known locomotor behaviors 519 among the studied primate taxa. Hence, this novel alternative approach will be useful for future 520 functional morphology studies. The results also show patterns of morphological variation that are 521 generally consistent, but not exactly the same, with other methods (but see below for additional 522 comparisons with IA and PCF). Despite the relatively intuitive pattern in species mean shape 523 differences revealed by the PCA (Figs. 8-10) from more suspensory orangutans on one end of the 524 spectrum of curvature shape variation to terrestrial quadrupedal baboons on the other, there were 525 a few surprising results that did not match predictions, and which will require further inquiry. 526 Specifically, the following five unexpected patterns were found: African apes do not 527 significantly differ from each other in PC1 scores; the two species of orangutans are significantly 528 different from each other in their PC1 scores; Ateles which uses forelimb suspension and 529 quadrupedalism does not significantly differ from brachiating hylobatids; the two quadrupedal 530 monkeys (Papio and Nasalis) rather than bipedal humans have the flattest phalangeal curve 531 shapes; and all three gorilla taxa differ in their PC2 scores from all other primates studied here. 532 Because chimpanzees, bonobos, and gorillas all have similar diaphysis shapes to each 533 other, and because they are significantly flatter in shape compared to the more habitually 534 suspensory taxa, the dorsopalmar curvature observed here in the African apes does not 535 necessarily reflect a suspensory adaptation alone or the significant differences in proportion of 536 time spent suspensory behaviors (see Carlson [2005] and referenced therein). What is more 537 likely is that their intermediate levels of curvature (as reflected by PC1 scores) better represents 538 adaptations for more diverse hand postures used in arboreal behaviors more generally, such as 539 vertical climbing, and may also reflect adaptations for quadrupedal locomotion with a knuckle- 540 walking hand posture, among others (e.g., Tuttle, 1970l Thompson et al., 2018). This possibility 541 is further supported by the fact that the quadrupedal cercopithecoid monkeys that use palmigrade 542 and digitigrade hand postures (e.g., Patel, 2010, and references therein) and are not stereotypical 543 orthograde vertical climbers have even flatter shapes than African apes. The large overlap among 544 the African apes (i.e., no significant differences among genera or species), however, may further 545 indicate that their phalangeal shapes may be due more to the lack of significant differences in the 546 proportion of time spent vertical climbing and its associated hand kinematics (e.g., Neufuss et 547 al., 2017), rather than the dramatic kinematic differences in knuckle-walking postures used 548 among African apes (e.g., Tuttle, 1970; Matarazzo, 2013). 549 Sumatran orangutans (Pongo abelii) were found to have diaphysis shapes that are more 550 curved than Bornean orangutans (Pongo pygmaeus) (Figs. 9 and 10). Although previous studies 551 have not demonstrated differences in phalangeal curvature between orangutan species, this may 552 be because these studies either only included one orangutan species (e.g., Jungers et al., 1997; 553 Dean & Begun, 2008; Rein, 2011), or both species may have been combined into a more general 554 orangutan sample (e.g., Stern, Jungers & Susman, 1995). While the biological significance of 555 this difference in phalangeal curvature between orangutan species is uncertain, it may be related 556 to their different ecologies. Specifically, the Borneo orangutans (males in particular) frequent the 557 ground as part of their normal behaviors (Ancrenaz et al., 2014), whereas Sumatran orangutans 558 may use terrestrial substrates less frequently due to the presence of tigers on the island of 559 Sumatra (there are no known tigers on the island of Borneo as they likely have gone extinct; 560 Delgado & van Schaik, 2000). Thus, if the Bornean orangutan has had a longer recent 561 evolutionary history of descending to the ground compared to Sumatran orangutans, it may have 562 secondarily evolved flatter proximal phalanges (assuming high phalangeal curvature is ancestral 563 for the genus Pongo). 564 Among the monkeys studied here, spider monkeys (Ateles) are the only species that 565 habitually use forelimb suspensory locomotion and postural behaviors (e.g., Rein, 2011). Thus, it 566 is not surprising that Ateles phalanges are similar in diaphysis shape to those of gibbons and 567 siamangs (Figs. 9 and 10). Furthermore, considering that they do not significantly differ from 568 either hylobatid group, it confirms that suspensory behaviors have a greater influence on 569 increasing proximal phalangeal curvature than does a more generalized arboreal locomotor 570 repertoire (e.g., as seen in Nasalis). In contrast, the diaphysis shapes seen in Nasalis (an above- 571 branch arboreal quadruped) and Papio (a primarily terrestrial quadruped) are similar to each 572 other in this study and both represent the flattest phalangeal curvatures (of the taxa studied here). 573 It is not entirely clear why these two large bodied monkeys have similar shapes despite 574 differences in preferred substrates and quadrupedal hand postures (i.e., palmigrade for Nasalis 575 vs. digitigrade for Papio; Patel, 2010), but it may be possible that these levels of proximal 576 phalangeal curvatures are common to all cercopithecoids or simply because they do not engage 577 in any significant amount of suspensory locomotion. It is also equally likely that the curvature of 578 proximal phalanges is not sensitive enough for selection to act on subtle locomotor differences 579 among habitually pronograde quadrupeds like cercopithecoids, regardless of their degree of 580 arboreality. Additional cercopithecoid taxa that use some form of suspensory locomotion (e.g., 581 Pygathrix nameus; Byron and Covert, 2004) should be included in future analyses to assess this 582 possibility further. It is important to note that the similarity between Papio and Nasalis seen 583 here does contrasts previous studies using IA (e.g., Jungers et al., 1997) and PCF (Deane and 584 Begun, 2008) where both found Papio to have significantly flatter proximal phalanges than 585 Nasalis. Again, the addition of more data from monkeys is needed to better understand this 586 outcome. 587 The fact that both cercopithecoids included here also overlap with diaphysis shapes 588 exhibited by bipedal humans (which rarely use their hands for locomotion) was unanticipated 589 (Fig. 9). A number of plausible, but not mutually exclusive, reasons may account for this 590 overlap. First, there is enough fossil evidence demonstrating that the earliest hominins had 591 greater phalangeal curvature than modern humans thereby making it a primitive feature for the 592 group (e.g., Stern & Susman, 1983; Tocheri et al., 2008; Kivell et al., 2011; Stratford et al., 593 2016; Prang et al., 2021). Thus, modern humans may simply have retained some degree of 594 phalangeal curvature from more ancestral hominin species that were adapted for arboreality. 595 Likewise, there may have been continued selection for moderate levels of phalangeal curvature 596 throughout human evolution as an adaptation to attenuate higher mechanical loads while using 597 powerful grips with large forces during stone tool production and use (e.g., Tocheri et al., 2008; 598 Harmand et al., 2015). Further studies of curvature shape of fossil proximal phalanges from the 599 Miocene, Pliocene and Pleistocene are needed to test these hypotheses further (e.g., Deane & 600 Begin, 2008; Stratford et al., 2016). Alternatively, it may be that the large cercopithecoid 601 monkeys (like Papio and Nasalis) have evolved to have flatter phalanges because their terrestrial 602 or above branch arboreal quadrupedal locomotion, respectively, may not benefit from greater 603 curvature (i.e., different mechanical loading regimes compared to below branch or vertical 604 climbing locomotion). Quadrupedal monkeys of varying size, especially in the smaller body 605 mass range, and substrate preference are needed to test this hypothesis further. 606 Finally, all three Gorilla taxa separated on PC2 in the full taxonomic analysis of 607 diaphysis shape (Fig. 9). Upon closer inspection of the maximum and minimum shapes relative 608 to the mean shape on PC2 (see Figure S3), and species mean shape relative to the sample mean 609 shape (Fig. 10), it appears that gorillas differ from the other primates studied here in the location 610 of the dROI’s curvature main inflection point. For gorillas with higher PC2 scores, the curvature 611 inflection point appears to be slightly distal to the midpoint whereas in other taxa it is closer to 612 the midpoint or slightly more proximal. It is difficult to say why the dROI curvature inflection 613 point would differ between gorillas and other primates, but we do not think this is related to 614 habitual locomotor behavior, substrate preference, or even body size of these great apes. 615 Moreover, since PC2 accounts for ~6% of the total variation, it is unclear how, if at all, 616 biologically and biomechanically meaningful the unusual gorilla pattern really is. But this may 617 warrant further consideration in the future especially if it successfully separates African ape 618 genera (Gorilla vs. Pan) from each other. 619 620 3.4 Comparability 621 The primary motivation for this project was to find an alternative to both IA and PCF 622 methods to quantify proximal phalangeal curvature based on 3D morphology of the bone’s 623 longitudinal central axis. One way to assess comparability between the different methods is with 624 correlation analyses. Here we correlated PC1 scores from the 2DGM analysis of the entire 625 comparative sample of 378 specimens with each specimen’s IA and PCF a coefficient. 626 Specifically, both IA and a were calculated in Excel software (Microsoft) from the same set of 627 100 2D landmarks identified from the dROI used in the 2DGM shape analysis (see Figure S5). 628 Three of the landmarks – proximal [#1], middle [#50], and distal [#100] – were used to compute 629 IA following modified equations presented in Jungers et al. (1997). Note that IAs calculated 630 from our landmark data will not yield the same IAs as those obtained in other studies based on 631 different protocols (Stern, Jungers & Susman, 1995; Matarazzo, 2008; Rein, 2011) and thus 632 should not be directly compared. All 100 landmarks were used to fit a second-degree polynomial 633 in order to determine the a coefficient following Deane, Kremer & Begun (2005). One important 634 difference between Deane, Kremer & Begun’s (2005) study and the present one is that the 635 second-degree polynomial was applied to the dorsal surface of the phalanx’s diaphysis in the 636 former study whereas the second-degree polynomial was applied to the bone’s longitudinal 637 central axis in the present study (compare Figure 2B with Figure S5). Pearson correlations were 638 assessed for two whole-sample comparisons: PC1 vs. IA and PC1 vs. a. Both correlations were 639 significant (p<0.05), but the r value for the correlation between PC1 and IA (r=-0.940; Fig. 11) 640 was much higher than that for the correlation between PC1 and a (r=-0.753; Fig. 12). 641 These results demonstrate that the 2DGM approach presented here to quantify diaphysis 642 shape of proximal phalanges reflecting longitudinal dorsopalmar curvature produces results that 643 are more in line with measures of included angle (and thus a portion of a circle) as championed 644 by Stern, Susman & Jungers (1995) rather than a second-degree polynomial (and hence a 645 parabola) as proposed by Deane, Kremer & Begun (2005). Despite the higher correlation, it is 646 particularly important to emphasize that our assessment of curvature shape in PC scores is not 647 the same as IA and thus does not specifically reflect a portion of a circle. In fact, the approach 648 presented here using 2DGM is not based on any a priori shape for phalangeal curvature. In 649 comparing PC1 scores, IA values, and a coefficients for each specimen by taxon (see Figure S6), 650 it is clear that the PC1 scores from the 2DGM analysis and IA values yield better results in 651 differentiating extant taxa by their habitual locomotor modes, whereas the a coefficient primarily 652 works well to discriminate only highly suspensory primates (i.e., Pongo, hylobatids, and Ateles) 653 from all others. That is to say, using the a coefficient derived from PCF as implemented in this 654 study does a relatively poorer job distinguishing among habitually arboreal and terrestrial 655 quadrupeds, as well as bipedal humans. We therefore recommend not using PCF methods in 656 more diverse phylogenetic and functional sampling of primates. 657 658 Figure 11 Here 659 Figure 12 Here 660 661 But which of these three methods correlates best with quantified levels of suspensory 662 locomotion (i.e., with real-world behavioral data)? To assess this, we performed non-parametric 663 Spearman correlations of mean PC1 scores, IA values and a coefficients (again from the same 664 100 landmark datasets obtained in this study) with mean time spent suspensory as a proportion of 665 total locomotor behavior (reported in Rein, 2011) at the genus level for a subset of our sample 666 (Table 3). Recognizing the small sample size in this comparison, results did show that the 667 proportion of time a genus has been reported to be suspensory was not significantly correlated 668 with the a coefficient (rs=0.618; p=0.153), but was with IA (rs=0.946; p=0.005) and PC1 scores 669 (rs=-0.982; p=0.002). Note that the correlation with PC1 scores was slightly higher than the 670 correlation with IA values, again suggesting that the 2DGM analysis does a better job than the 671 IA method in identifying suspensory taxa. The latter finding in particular may further influence 672 functional morphologists and paleoanthropologists to adopt this novel alternative method to 673 quantitatively assess phalangeal curvature in their future research. 674 675 Table 3 Here 676 677 4. Summary and Conclusions 678 A number of different metrics have been employed to assess and compare phalangeal 679 curvature in living and fossil primates. The most common are the included angle (measured in 680 degrees; Stern, Jungers & Susman, 1995) which assumes that the proximal phalanx represents a 681 portion of an arc of a circle (Fig. 2A), and the a coefficient of a second-degree polynomial (via 682 polynomial curve fitting or PCF; Deane, Kremer & Begun, 2005) which assumes that the bone 683 represents the shape of a parabola (Fig. 2B). Although not identical, both methods and metrics 684 generally produce similar results (Fig. 2C) with an acceptable amount of discriminatory power 685 between primate taxa that are highly suspensory (i.e., with more curved proximal phalanges) and 686 taxa that are more terrestrial and quadrupedal (i.e., with flatter proximal phalanges) (Fig. 1; Rein, 687 2011; Deane & Begin, 2008). However, both methods also have a number of drawbacks. One is 688 the high degree of inter-study variability in published included angle values for several 689 commonly studied primates like chimpanzees and macaques (Patel & Maiolino, 2016). Also 690 notable is the fact that as currently implemented, PCF quantifies only the dorsal contour of the 691 proximal phalanx’s diaphysis (Deane, Kremer & Begun, 2005) which is less biomechanically 692 meaningful than the curvature of its central longitudinal axis (e.g., Richmond, 2007). 693 Additionally, both approaches often yield some counterintuitive results based on predicted form- 694 function relationships. For instance, calculating included angles from lateral view photographs, 695 often results in no significant differences between gorillas (terrestrial quadrupeds/climbers) and 696 bipedal humans (e.g., Prang et al., 2021). Similarly, brachiating gibbons are not significantly 697 different from arboreal quadrupedal monkeys and African apes (e.g., Böhme et al., 2019) when 698 calculating the a coefficient with PCF from lateral view photographs of the diaphysis’ dorsal 699 contour. Because phalangeal curvature has been an important variable to assess locomotor and 700 positional behavior in paleoanthropological research (e.g., Stern & Susman, 1983; Jungers et al., 701 1997; Deane & Begun, 2008; Zhang, Harrison, & Ji, 2020), especially in fossil hominins and 702 fossil non-human apes, it is necessary to consistently and more accurately (i.e., track known 703 behavioral lines) assess curvature across a diverse range of primates. 704 The present study was conducted to develop and test a new alternative method - that is 705 repeatable and biomechanically meaningful using geometric morphometric (GM) techniques - to 706 quantify proximal phalanx diaphysis shape that reflects dorsopalmar curvature. The 2DGM 707 method presented here is not sensitive to inter- and intra-observer error or the digit studied 708 (mPP2 vs. mPP3 vs. mPP4 vs. mPP5). Even the overall quality of the 3D model did not have a 709 major impact on the results (see Supplemental Information) as long as it has a surface mesh of at 710 least 100,000 triangles that does not contain any erroneous spikes, depressions, or holes 711 (including the removal of any naturally occurring nutrient foramina). Accordingly, the source of 712 the 3D surface data (e.g., photogrammetry, laser scan, CT scan, µCT scan) matters little, thus 713 making the method applicable to many more users in the research community (see also Tocheri 714 et al., 2011; Robinson & Terhune, 2017; Shearer at et al., 2017). However, users of this 2DGM 715 method should be mindful to orient the virtual phalanx prior to any subsequent steps using our 716 ‘gold standard’ orientation (as described above; Fig. 3A). The latter is particularly important if 717 users want to use the complete set of landmark data provided in the online Supplemental 718 Information. 719 Overall, the results of our 2DGM analyses demonstrate that differences in diaphysis 720 shape reflective of dorsopalmar curvature across taxa derive from real biological inter- and 721 intraspecific variation. Specifically, assessing curvature shape of the dROI (i.e., the bone’s 722 central longitudinal axis excluding proximal and distal articular ends) proves to be useful as PC1 723 scores (which accounted for >80% of total sample variation) distinguish extant taxa along known 724 locomotor behavioral lines. The most suspensory primates such as orangutans, gibbons, 725 siamangs, and spider monkeys still clearly have more curved proximal phalanges compared to 726 stereotypical quadrupedal species (both arboreal and terrestrial) such as proboscis monkeys and 727 baboons, as well as bipedal humans. However, in the new 2DGM approach, different orangutan 728 species can be differentiated, gorilla proximal phalanges are now observed to be significantly 729 more curved than humans, and gibbon proximal phalanges are now observed to be significantly 730 more curved than African apes and monkeys (except suspensory spider monkeys). Therefore, we 731 contend that the 2DGM approach offers even greater discriminating power compared to using IA 732 and PCF especially with regards to degree of forelimb suspension used by living primates. 733 Whether this approach can better differentiate habitually arboreal vs. terrestrial quadrupedal 734 monkeys, however, needs further testing. Additionally, while PC1 scores were significantly 735 correlated with both the included angle and the a coefficient of a second-degree polynomial 736 fitted to the same landmark datasets, PC1 scores had a significantly higher correlation with 737 included angles suggesting that a circle model for phalangeal curve shape might be more 738 appropriate than a parabola model. But, we do emphasize that PC scores are not equivalent to IA 739 (measured in degrees) and cannot be used interchangeably. 740 Compared to using IA as is currently implemented (e.g., lateral view photographs, 741 Microscribe point digitizer), some readers of this technical note may consider the 2DGM method 742 presented here too complex and time intensive given that it yields results generally similar to 743 those obtained from other methods. We acknowledge that our approach has more steps and using 744 IA as currently implemented still has some value. For example, limiting comparisons between 745 closely related taxa that are behaviorally different like bipedal humans, quadrupedal 746 chimpanzees and suspensory orangutans (e.g. Stern, Susman & Jungers, 1995; Wallace, Burgess 747 & Patel, 2020) produces results that could be used to generalize the locomotor behaviors of some 748 fossil hominins (e.g., Stratford et al., 2016). But similar comparisons that include more 749 phylogenetic and behavioral diversity of extant and fossil primates often produce results that do 750 not clearly reflect with known or inferred locomotor behaviors (Prang et al., 2021). 751 In summary, we maintain that our 2DGM approach is a worthwhile alternative to assess 752 dorsopalmar curvature because: 1) 2DGM does not assume a priori portion of an arc of a circle 753 or parabola model; 2) the developed method has very low intra- and inter-observer variation 754 relative to the variation seen within and among taxa; 3) resulting PC scores show improved 755 extant group separation along known behavioral lines; and 4) resulting PC scores have higher 756 correlation with quantified behavioral data than other metrics. Finally, one other benefit of our 757 landmark-based approach is that the derived 100 landmarks can still be used to quantify IA and a 758 coefficient if a researcher still prefers these metrics over PCA on shape coordinates, but with the 759 added benefit that now both IA and a metrics will be derived from the more biomechanically 760 relevant central longitudinal axis of the phalanx. To conclude, the outcomes of this technical 761 study demonstrate that proximal phalanx diaphysis shape reflecting longitudinal dorsopalmar 762 curvature determined from 2DGM methods should prove to be useful for addressing additional 763 questions on the functional morphology of primate hands, as well as future paleoanthropological 764 research that involves reconstructing behavior in the fossil record. 765 766 Acknowledgements 767 We thank the curators and collections managers at the following museums that facilitated 768 access to specimens in their care: American Museum of Natural History; Academy of Natural 769 Sciences of Drexel University; Natural History Museum of Los Angeles County; Harvard 770 University’s Museum of Comparative Zoology; Royal Museum for Central Africa; Smithsonian 771 Institution’s National Museum of Natural History; University of the Witwatersrand. Andrew 772 Deane also generously shared Gorilla b. beringei data obtained with assistance from the 773 Mountain Gorilla Skeleton Project (MGSP); more details and funding information for the MGSP 774 is available at: www.gwprimatelifehistorylab.org/mountain-gorilla-skeletal-project. 775 We are also grateful to the staff at the following institutions that assisted us with CT and 776 µCT scanning: University of Southern California’s Molecular Imaging Center which receives 777 funding from the National Institutes of Health (NIH S10 RR027665 01); Department of 778 Biomedical Engineering at Stony Brook University; Department of Radiology at Stony Brook 779 University School of Medicine; and the Smithsonian Institution’s National Museum of Natural 780 History. 781 Funding for this research was generously provided by: National Science Foundation 782 (BCS #1317047 to B.A.P.; BCS #1539741 to C.M.O.); the L.S.B. Leakey Foundation (to 783 B.A.P.); the Wenner-Gren Foundation (to C.M.O.); AEI/FEDER EU (CGL2017-82654-P to 784 S.A.), Generalitat de Catalunya (CERCA Programme to S.A.); the University of Southern 785 California’s URAP program (to B.A.P.), and the University of Southern California’s Office of 786 the Provost (to S.E.W.). 787 788 Author contributions 789 B.A.P. designed the study; B.A.P., C.M.O., S.A. and M.W.T. obtained the raw image and scan 790 data and created the 3D surface models; S.E.W. collected the landmark data; K.L.L. 791 implemented the R script for the 2DGM analyses; S.E.W., K.L.L. and B.A.P. performed the 792 analyses; B.A.P. and S.E.W. drafted the manuscript; all authors reviewed and approved the final 793 version of the manuscript. 794 795 Conflict of Interest 796 The authors declare no conflict of interests. 797 798 Data availability statement 799 R script for performing 2D geometric morphometrics on 100 landmarks of primate proximal 800 phalanges to assess longitudinal curvature shape used in this study is available in the online 801 Supplemental Information. Morphologika data files for all specimens (n=378) used in this study 802 is also available in the online Supplemental Information. 3D surface models illustrated in Figure 803 1 are available on MorphoSource.org as part of project "Anthropoid Primate Hands", the details 804 of which are provided in Appendix Table 1. 805 806 References 807 Adams, D. C., Collyer, M. L., & Kaliontzopoulou, A. (2019). Geomorph: software for geometric 808 morphometric analyses. https://cran.r-project.org/package=geomorph 809 810 Birdal, T. (2021). Maximum Inscribed Circle using Distance 811 Transform (https://www.mathworks.com/matlabcentral/fileexchange/30805-maximum- 812 inscribed-circle-using-distance-transform), MATLAB Central File Exchange. Retrieved March 813 1, 2021. 814 815 Böhme, M., Spassov, N., Fuss, J., Tröscher, A., Deane, A. S., Prieto, J., Kirscher, U., Lechner, 816 T., & Begun, D. R. (2019). A new Miocene ape and locomotion in the ancestor of great apes and 817 humans. Nature. 575, 1–20. https://doi.org/10.1038/s41586-019-1731-0 818 819 Boyer, D. M., Puente, J., Gladman, J. T., Glynn, C., Mukherjee, S., Yapuncich, G. S., & 820 Daubechies, I. (2015). A new fully automated approach for aligning and comparing shapes. 821 Anatomical Record. 298, 249–276. https://doi.org/10.1002/ar.23084 822 823 Burgess, M. L. (2018). Ontogenetic Changes in Limb Bone Structural Properties and Locomotor 824 Behavior in Pan. Ph.D. Dissertation, Johns Hopkins University, Baltimore. 825 826 Burri, O., & Guiet, R. (2016). https://imagej.net/Max_Inscribed_Circles. 827 828 Byron, C., & Covert, H. (2004). Unexpected locomotor behaviour: brachiation by an Old World 829 monkey (Pygathrix nemaeus) from Vietnam. Journal of Zoology. 263, 101–106. 830 http://doi.org/10.1017/S0952836904004935 831 832 Carlson, K. J. (2005). Investigating the form-function interface in African apes: Relationships 833 between principal moments of area and positional behaviors in femoral and humeral diaphyses. 834 American Journal of Physical Anthropology. 127, 312-334. http://doi.org/10.1002/ajpa.20124. 835 836 Deane, A. S., & Begun, D. R. (2008). Broken fingers: retesting locomotor hypotheses for fossil 837 hominoids using fragmentary proximal phalanges and high-resolution polynomial curve fitting 838 (HR-PCF). Journal of Human Evolution. 55, 691–701. 839 https://doi.org/10.1016/j.jhevol.2008.05.005 840 841 Deane, A. S., Kremer, E. P., & Begun, D. R. (2005). New approach to quantifying anatomical 842 curvatures using high-resolution polynomial curve fitting (HR-PCF). American Journal of 843 Physical Anthropology. 128, 630–638. https://doi.org/10.1002/ajpa.20202 844 845 De Groote, I., Lockwood, C. A., & Aiello, L. C. (2010). Technical note: A new method for 846 measuring long bone curvature using 3D landmarks and semi-landmarks. American Journal of 847 Physical Anthropology. 141, 658-664. https://doi.org/10.1002/ajpa.21225 848 849 Delgado, R. A., & van Schaik, C. P. (2000). The behavioral ecology and conservation of the 850 orangutan (Pongo pygmaeus): a tale of two islands. Evolutionary Anthropology. 9, 201–218. 851 https://doi.org/10.1002/1520-6505(2000)9:5%3C201::AID-EVAN2%3E3.0.CO;2-Y 852 853 Domínguez-Rodrigo, M., Pickering, T. R., Almécija, S., Heaton, J. L., Baquedano, E., Mabulla, 854 A., & Uribelarrea, D. (2015). Earliest modern human-like hand bone from a new >1.84-million- 855 year-old site at Olduvai in Tanzania. Nature Communications. 6, 7987. 856 https://doi.org/10.1038/ncomms8987 857 858 Fernández, P. J., Almécija, S., Patel, B. A., Orr, C. M., Tocheri, M. W., & Jungers, W. L. (2015). 859 Functional aspects of metatarsal head shape in humans, apes, and Old World monkeys. Journal 860 of Human Evolution. 86, 136–146. https://doi.org/10.1016/j.jhevol.2015.06.005 861 862 Granatosky, M. C. (2018). A review of locomotor diversity in mammals with analyses exploring 863 the influence of substrate use, body mass and intermembral index in primates. Journal of 864 Zoology. 306, 207–216. https://doi.org/10.1111/jzo.12608 865 866 Hammer, Ø., Harper, D. A. T., & Ryan, P. D. (2001). PAST: Paleontological statistics software 867 package for education and data analysis. Palaeontologia Electronica. 4, art. 4. 868 869 Hamrick, M. W., Meldrum, D. J., & Simons, E. L. (1995). Anthropoid phalanges from the 870 Oligocene of Egypt. Journal of Human Evolution. 28, 121–145. 871 https://doi.org/10.1006/jhev.1995.1011 872 873 Harmand, W., Lewis, J. E., Feibel, C. S., Lepre, C. J., Prat, S., Lenoble, A., Boës, X., Quinn, R. 874 L., Brenet, M., Arroyo, A., Taylor, N., Clément, S., Daver, G., Brugal, J., Leakey, L., Mortock, 875 R. A., Wright, J. D., Lokorodi, S., Kirwa, C., Kent, D. V., & Roche, H. (2015). 3.3-million-year 876 old stone tools from Lomekwi 3, West Turkana, Kenya. Nature. 521, 310-315. 877 https://doi.org/10.1038/nature14464 878 879 Jungers, W. L., Godfrey, L. R., Simons, E. L., & Chatrath, P. S. (1997). Phalangeal curvature 880 and positional behavior in extinct sloth lemurs (Primates, Palaeopropithecidae). Proceedings of 881 the National Academy of Sciences. 94, 11998–12001. https://doi.org/10.1073/pnas.94.22.11998 882 883 Jungers, W. L., Godfrey, L. R., Simons, E. L., Wunderlich, R. E., Richmond, B. G., & Chatrath, 884 P.S. (2002). Ecomorphology and behavior of giant extinct lemurs from Madagascar. In: Plavcan, 885 J.M., Kay, R.F., Jungers, W.L., van Schaik, C.P. (Eds.), Reconstructing Behavior in the Primate 886 Fossil Record. Kluwer Academic/Plenum Publishers, New York, pp. 371–411. 887 https://doi.org/10.1007/978-1-4615-1343-8_10 888 889 Kivell, T. L., Deane, A. S., Tocheri, M. W., Orr, C. M., Schmid, P., Hawks, J., Berger, L. R., & 890 Churchill, S. E. (2015). The hand of Homo naledi. Nature Communications. 6, 8431. 891 https://doi.org/10.1038/ncomms9431 892 893 Kivell, T. L., Kibii, J. M., Churchill, S. E., Schmid, P., & Berger, L. R. (2011). Australopithecus 894 sediba hand demonstrates mosaic evolution of locomotor and manipulative abilities. Science. 895 333, 1411–1417. https://doi.org/10.1126/science.1202625 896 897 Kivell, T. L., Lemelin, P., Richmond, B. G., & Schmitt, D. (2016). The Evolution of the Primate 898 Hand: Anatomical, Developmental, Functional, and Paleontological Evidence. Springer, New 899 York. http://doi.org/10.1007/978-1-4939-3646-5 900 901 Matarazzo, S. (2008). Knuckle walking signal in the manual digits of Pan and Gorilla. American 902 Journal of Physical Anthropology. 135, 27–33. https://doi.org/10.1002/ajpa.20701 903 904 Matarazzo, S. (2013). Manual pressure distribution patterns of knuckle-walking apes. American 905 Journal of Physical Anthropology. 152, 44–50. https://doi.org/10.1002/ajpa.22325 906 907 Milne, N., & Granatosky, M. C. (2021). Ulna curvature in arboreal and terrestrial primates. 908 Journal of Mammalian Evolution. 28, 897-909. http://doi.org/10/1007/s10914-021-09566-5 909 910 Napier, J. R. (1993). Hands [Revised by Russell H. Tuttle]. Princeton University Press, 911 Princeton. 912 913 Neufuss, J., Robbins, M. M., Baeumer, J., Humle, T., & Kivell, T. L. (2017). Comparison of 914 hand use and forelimb posture during vertical climbing in mountain gorillas (Gorilla beringei 915 beringei) and chimpanzees (Pan troglodytes). American Journal of Physical Anthropology. 164, 916 651–664. https://doi.org/10.1002/ajpa.23303 917 918 Nguyen, N. H., Pahr, D. H., Gross, T., Skinner, M. M., & Kivell, T. L. (2014). Micro-finite 919 element (µFE) modeling of the siamang (Symphalangus syndactylus) third proximal phalanx: the 920 functional role of curvature and the flexor sheath ridge. Journal of Human Evolution. 67, 60–75. 921 https://doi.org/10.1016/j.jhevol.2013.12.008 922 923 Patel, B. A. (2010). Functional morphology of cercopithecoid primate metacarpals. Journal of 924 Human Evolution. 58, 320–337. https:/doi.org/10.1016/j.jhevol.2010.01.001 925 926 Patel, B. A., & Maiolino, S. A. (2016). Morphological diversity in the digital rays of primate 927 hands. In: Kivell, T. L., Lemelin, P., Richmond, B. G., & Schmitt, D. (Eds.), The Evolution of 928 the Primate Hand: Anatomical, Developmental, Functional, and Paleontological Evidence. 929 Springer, New York, pp. 55–100. https://doi.org/10.1007/978-1-4939-3646-5_4 930 931 Patel, B. A., Yapuncich, G. S., Tran, C., & Nengo, I. O. (2017). Catarrhine hallucal metatarsals 932 from the early Miocene site of Songhor, Kenya. Journal of Human Evolution. 108, 176–198. 933 https://doi.org/10.1016/j.jhevol.2017.03.013 934 935 Prang, T. C., Ramirez, K., Grabowski, M., & Williams, S. A. (2021). Ardipithecus hand provides 936 evidence that humans and chimpanzees evolved from an ancestor with suspensory adaptations. 937 Sciences Advances. 7, eabf2474. http://doi.org/10.1126/sciadv.abf2474 938 939 Preuschoft, H. (1973a). Body posture and locomotion in some east African Miocene 940 Dryopithecinae. In: Day, M. H. (Ed.), Human Evolution (Symposium of the Society for the Study 941 of Human Biology, Volume XI). Taylor and Francis, London, pp. 13–46. 942 943 Preuschoft, H. (1973b). Functional anatomy of the upper extremity. In: Bourne, G. H. (Ed.), The 944 Chimpanzee, Volume 6. Karger, Basel, pp. 34–120. 945 946 R Core Team. (2019). R: A language and environment for statistical computing. R Foundation 947 for Statistical Computing, Vienna, Austria. https://www.R-project.org/. 948 949 Rein, T. R. (2011). The correspondence between proximal phalanx morphology and locomotion: 950 implications for inferring the locomotor behavior of fossil catarrhines. American Journal of 951 Physical Anthropology. 146, 435–445. https://doi.org/10.1002/ajpa.21599 952 953 Richmond, B. G. (1998). Ontogeny and biomechanics of phalangeal form in primates. Ph.D. 954 Dissertation, State University of New York at Stony Brook, Stony Brook. 955 956 Richmond, B. G. (2007). Biomechanics of phalangeal curvature. Journal of Human Evolution. 957 53, 678–690. https://doi.org/10.1016/j.jhevol.2007.05.011 958 959 Robinson, C., & Terhune, C. E. (2017). Error in geometric morphometric data collection: 960 Combining data from multiple sources. American Journal of Physical Anthropology. 164, 62-75. 961 https://doi.org/10.1002/ajpa.23257 962 963 Schneider, C. A., Rasband, W. S., & Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of 964 image analysis. Nature Methods. 9, 671–675. https://doi.org/10.1038/nmeth.2089 965 966 Shearer, B. M., Cooke, S. B., Halenar, L. B., Reber, S. L., Plummer, J. E., Delson, E., & 967 Tallman, M. (2017). Evaluating causes of error in landmark-based data collection using scanners. 968 PLOS ONE 12, e0187452. https://doi.org/10.1371/journal.pone.0187452 969 970 Spencer, M. A., & Spencer, G. S. (1995). Technical note: video-based three-dimensional 971 morphometrics. American Journal of Physical Anthropology 96, 443–453. 972 https://doi.org/10.1002/ajpa.1330960409 973 974 Stern, J. T., Jungers, W. L., & Susman, R. L. (1995). Quantifying phalangeal curvature: an 975 empirical comparison of alternative methods. American Journal of Physical Anthropology. 97, 976 1–10. https://doi.org/10.1002/ajpa.1330970102 977 978 Stern, J. T., & Susman, R.L. (1983). The locomotor anatomy of Australopithecus afarensis. 979 American Journal of Physical Anthropology. 60, 279–317. 980 https://doi.org/10.1002/ajpa.1330600302 981 982 Stratford, D., Heaton, J. L., Pickering, T. R., Caruana, M. V., & Shadrach, K. (2016). First 983 hominin fossils from Milner Hall, Sterkfontein, South Africa. Journal of Human Evolution. 91, 984 167–173. https://doi.org/10.1016/j.jhevol.2015.12.005 985 986 Squyres, N., & DeLeon, V. B. (2015). Clavicular curvature and locomotion in anthropoid 987 primates: A 3D geometric morphometric analysis. American Journal of Physical Anthropology. 988 158, 257-268. https://doi.org/10.1002/ajpa.22785 989 990 Susman, R. L., de Ruiter, D., & Brain, C. K. (2001). Recently identified postcranial remains of 991 Paranthropus and early Homo from Swartkrans Cave, South Africa. Journal of Human 992 Evolution. 41, 607–629. https://doi.org/10.1006/jhev.2001.0510 993 994 Susman, R. L., Stern, J. T., & Jungers, W. L. (1984). Arboreality and bipedality in the Hadar 995 hominids. Folia Primatologica. 43, 113–156. https://doi.org/10.1159/000156176 996 997 Thompson, N. E., Ostrofsky, K. R., McFarlin, S. C., Robbins, M. M., Stoinski, T. S., & 998 Almécija, S. (2018). Unexpected terrestrial hand posture diversity in wild mountain gorillas. 999 American Journal Physical Anthropology. 166, 84-94. https://doi.org/10.1002/ajpa.23404 1000 1001 Tocheri, M. W., Orr, C. M., Jacofsky, M. C., & Marzke, M. W. (2008). The evolutionary history 1002 of the hominin hand since the last common ancestor of Pan and Homo. Journal of Anatomy. 212, 1003 544-562. https://doi.org/10.1111/j.1469-7580.2008.00865.x 1004 1005 Tocheri, M. W., Solhan, C. R., Orr, C. M., Femiani, J., Frohlich, B., Groves, C. P., Harcourt- 1006 Smith, W. E., Richmond, B. G., Shoelson, B., & Jungers, W. L. (2011). Ecological divergence 1007 and medial cuneiform morphology in gorillas. Journal of Human Evolution. 60, 171-184. 1008 https://doi:10.1016/j.jhevol.2010.09.002 1009 1010 Tuttle, R. H. (1970). Postural, propulsive, and prehensile capabilities in the cheiridia of 1011 chimpanzees and other great apes. In: Bourne, G. H. (Ed.), The Chimpanzee, Volume 2. Karger, 1012 Basel, pp. 167–253. 1013 1014 Wallace, I. J., Burgess, M. L., & Patel, B. A. (2020). Phalangeal curvature in a chimpanzee 1015 raised like a human: Implications for inferring arboreality in fossil hominins. Proceedings of the 1016 National Academy of Sciences. 117, 11223–11225. https://doi.org/10.1073/pnas.2004371117 1017 1018 Zhang, Y. Q., Harrison, T., & Ji, X. P. (2020). Inferring the locomotor behavior of fossil 1019 hominoids from phalangeal curvature using a novel method: Lufengpithecus as a case study. Acta 1020 Anthropologica Sinica. 39, 532-554. https://doi.org/10.16359/j.cnki.cn11-1963/q.2020.0061 1021 1022 1023 1024 Figure Legends 1025 1026 Figure 1. Primate proximal phalanges. Representative 3D surface models (top; silver color) 1027 and mid-sagittal planes (bottom; black color) of manual proximal phalanges of digit 3 (mPP3) in 1028 medial views. The proximal ends of the phalanges are to the left and the distal ends are to the 1029 right. All bones are from the right side or have been mirror-imaged to depict right side elements 1030 (as noted in Table S1). One hundred landmarks (yellow filled spheres) within the mid-sagittal 1031 plane show the central proximodistal axis and illustrates diaphysis shape reflecting its 1032 dorsopalmar curvature. Abbreviations: POA, Pongo abelii (USNM-143598; female); POP, 1033 Pongo pygmaeus (USNM-145306; female; mirror-imaged); SS, Symphalangus syndactylus 1034 (USNM-143579; female); HY, Hylobates agilis (AMNH-103667, male); PT, Pan troglodytes 1035 (USNM-176228; male); PP, Pan paniscus (RMCA-20881; male; mirror-imaged); GG, Gorilla 1036 gorilla (AMNH-54355; male); GBB, Gorilla beringei beringei (USNM-545041; male; mirror- 1037 imaged); HS, Homo sapiens (USNM-437R; female); AT, Ateles paniscus (USNM-361027; male; 1038 mirror-imaged); PAP, Papio cynocephalus (USNM-452509; male); NL, Nasalis larvatus 1039 (AMNH-106275; male). All individuals have been scaled to the same interarticular length of the 1040 mid-sagittal plane. Scale bars for each individual are below its surface model and equal 1 cm. 1041 The 3D surface models of each specimen illustrated are available for further study on 1042 www.MorphoSource.org (see Appendix Table 1 for more details). 1043 1044 Figure 2. Current methods to measure phalanx curvature. A) Included angles - an arc of a 1045 circle model of curvature - shown for a highly curved proximal phalanx of Pongo abelii (top; Θ1; 1046 red dashed line half-circle) and a straight proximal phalanx of Homo sapiens (bottom; Θ2; red 1047 solid line half-circle); note that Θ1 > Θ2. B) Second-degree polynomials – a parabola model of 1048 curvature - shown for a highly curved proximal phalanx of Pongo abelii (top; a1; blue dashed 1049 line parabola) and a straight proximal phalanx of Homo sapiens (bottom; a2; blue solid line 1050 parabola); note that a1 > a2. C) Bivariate scatter plot of mean included angles (from Rein, 2011) 1051 vs. median polynomial curve fitting a coefficients (from Dean & Begun, 2008) in a sample of 1052 extant apes and monkeys showing close, but less than ideal, correlation. 1053 1054 Figure 3: 2D landmark acquisition. A) 3D surface model of a manual proximal phalanx of 1055 digit 3 in Pongo abelii (USNM-143598) shown in the standardized anatomical orientation (i.e., 1056 the ‘gold’ standard orientation) as defined in the text. B) Illustration depicting a hollow surface 1057 mesh being voxelized (i.e., virtually filled) and resliced yielding 1000 evenly spaced binary 1058 slices along the length of the entire bone. C) The 1000 binary slices are cropped to create a new 1059 region of interest of the diaphysis (dROI; pink shaded region) as defined in the text, from which 1060 a new hollow surface mesh is created of the dROI, and subsequently the dROI is again voxelized 1061 and resliced yielding 100 evenly spaced binary slices. D) The centroid of each of the 100 binary 1062 slices of the dROI is determined using the Max Inscribed Circle plugin for ImageJ software as 1063 defined in the text. The y- and z-coordinates of the centroids ultimately serve as the landmarks 1064 for the 2DGM analyses to quantify diaphysis shape. 1065 1066 Figure 4: Error analysis – specimen orientation. Procrustes distances between the gold 1067 standard specimen and 60 variants of different orientation plotted against change in degrees in 1068 reference to the gold standard specimen. The insert shows a manual proximal phalanx of digit 3 1069 in Pongo abelii (USNM-143598) rotated about its central axis of rotation between 1-10 degrees 1070 from its ‘gold standard’ orientation (as defined in the text and shown in Figure 3A) in the dorsal, 1071 palmar, medial, lateral, clockwise, counterclockwise directions. Among the 60 variants, 1072 Procrustes distances differed inconsequentially when the specimen moved from the gold 1073 standard in medial, lateral, clockwise, and counterclockwise directions, but Procrustes distances 1074 increased linearly when the oriented bone was incorrectly positioned in the dorsal and palmar 1075 directions. 1076 1077 Figure 5: Error analysis – dROI determination. Box-and-whiskers plots of Procrustes 1078 distances of intra- and inter-observer error evaluating specimen dROI determination. Blue and 1079 red boxes show intra-observer Procrustes distances (n=45 comparisons for each) and the purple 1080 box shows inter-observer Procrustes distances (n=100 comparisons) for the tested Pongo abelii 1081 specimen (Pongo abelii USNM-143598). The gray box shows Procrustes distances between all 1082 Pongo abelii specimens (n=136 comparisons). Horizontal lines within each box illustrate the 1083 median of the distribution. Boxes envelop the inter quartile range (50% of values) of the sample 1084 distribution, and whiskers encompass the range excluding outliers. Filled circles beyond 1085 whiskers indicate outliers. Overall species variation (gray box) indicates that any minor 1086 differences in cropping (as defined in the text and shown in Figure 3) within and between 1087 observers of a single specimen is negligible in larger scale comparisons within and among taxa. 1088 1089 Figure 6. Error analysis – complete protocol. Box-and-whiskers plots of Procrustes distances 1090 of intra- and inter-observer error evaluating the complete protocol. Blue and red boxes show 1091 intra-observer Procrustes distances (n=45 comparisons for each) and the purple box shows inter- 1092 observer Procrustes distances (n=100 comparisons) for the tested Pongo abelii specimen (Pongo 1093 abelii USNM-143598). The gray box shows Procrustes distances between all Pongo abelii 1094 specimens (n=136 comparisons). Horizontal lines within each box illustrate the median of the 1095 distribution. Boxes envelop the inter quartile range (50% of values) of the sample distribution, 1096 and whiskers encompass the range excluding outliers. Filled circles beyond whiskers indicate 1097 outliers. Overall species variation (gray box) indicates that any minor differences in 1098 implementing the complete protocol (as defined in the text and shown in Figure 3) within and 1099 between observers of a single specimen is trivial in larger scale comparisons within and among 1100 taxa. 1101 1102 Figure 7. Inter-digit variation. Box-and-whiskers plots of PC1 scores for the analysis 1103 evaluating curvature shape across manual proximal phalanges of digits 2-5 (PP2-PP5) in Pongo 1104 abelli (left, filled boxes) and Homo sapiens (right, open boxes), two species with curved and flat 1105 phalanges, respectively. Horizontal lines within each box illustrate the median of the distribution. 1106 Boxes envelop the inter quartile range (50% of values) of the sample distribution, and whiskers 1107 encompass the range excluding outliers. Filled circles beyond whiskers indicate outliers. Despite 1108 some variation across digits there were no significant differences (p>0.05) in any of the 1109 comparisons in either species. 1110 1111 Figure 8: Evaluating 2DGM variation. Bivariate scatter plot of PC1 vs. PC2 scores for the 1112 2DGM analysis evaluating diaphysis shape of manual proximal phalanges in the comparative 1113 sample. PC1 accounted for 85.3% of the total variation and represents the overall curvature 1114 shape of the bone’s diaphysis. Specimens that are more curved are negative on the PC1 axis and 1115 those that are flatter are positive (see also Figure 10). The maximum and minimum shapes for 1116 PC1 are shown on the left and right sides, respectively (black filled circles are the max and min 1117 shapes; gray filled circles are mean shapes). The maximum and minimum shapes for PC2 are 1118 shown in Figure S3. Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan paniscus; 1119 GG, Gorilla gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; POA, 1120 Pongo abelii; POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; PAP, 1121 Papio spp.; NL, Nasalis larvatus; AT, Ateles spp. See Table 1 for additional information about 1122 the comparative sample. 1123 1124 Figure 9: Evaluating 2DGM utility. Box-and-whiskers plots of PC1 scores (top panel) and PC2 1125 scores (bottom panel) for the 2DGM analysis evaluating diaphysis shape of manual proximal 1126 phalanges in the comparative sample. Specimens that are more curved have smaller (and 1127 negative) PC1 scores and those that are flatter have higher (and positive) PC1 scores. Horizontal 1128 lines within each box illustrate the median of the distribution. Boxes envelop the inter quartile 1129 range (50% of values) of the sample distribution, and whiskers encompass the range excluding 1130 outliers. Filled circles beyond whiskers indicate outliers. Specimens ordered from left to right 1131 based on mean PC1 score. Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan 1132 paniscus; GG, Gorilla gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; 1133 POA, Pongo abelii; POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; 1134 PAP, Papio spp.; NL, Nasalis larvatus; AT, Ateles spp. See Table 1 for additional information 1135 about the comparative sample. 1136 1137 Figure 10. Species mean shapes. Plots of species mean shapes relative to the sample mean for 1138 the 2DGM analysis evaluating manual proximal phalanx diaphysis shape. Black filled circles are 1139 species mean Procrustes coordinates; gray filled circles represent the sample mean Procrustes 1140 coordinates. The proximal ends of the phalanges are to the left and distal ends are to the right. 1141 The dorsal sides of the phalanges are to the top and the palmar sides are to the bottom. 1142 Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan paniscus; GG, Gorilla gorilla; 1143 GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; POA, Pongo abelii; POP, 1144 Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; PAP, Papio spp.; NL, 1145 Nasalis larvatus; AT, Ateles spp. See Table 1 for additional information about the comparative 1146 sample. 1147 1148 Figure 11. 2DGM vs. included angle. Bivariate scatter plot of PC1 scores vs. included angles 1149 (in degrees) evaluating the correlation between different methods of assesses proximal 1150 phalangeal curvature in a sample of humans, non-human apes, and monkeys. The correlation is 1151 significant (r=-0.940) and higher than the correlation observed between PC1 scores and PCF a 1152 coefficients (see Figure 12). Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan 1153 paniscus; GG, Gorilla gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; 1154 POA, Pongo abelii; POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; 1155 PAP, Papio spp.; NL, Nasalis larvatus; AT, Ateles spp. See Table 1 for additional information 1156 about the comparative sample. 1157 1158 Figure 12. 2DGM vs. PCF. Bivariate scatter plot of PC1 scores vs. PCF a coefficients 1159 evaluating the correlation between different methods of assessing proximal phalangeal curvature 1160 in a sample of humans, non-human apes, and monkeys. The correlation is significant (r=-0.753) 1161 but is lower than the correlation observed between PC1 scores and included angles (see Figure 1162 11). Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan paniscus; GG, Gorilla 1163 gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; POA, Pongo abelii; 1164 POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; PAP, Papio spp.; 1165 NL, Nasalis larvatus; AT, Ateles spp. See Table 1 for additional information about the 1166 comparative sample. Table 1: Comparative sample. € Taxon Code Taxon Common Name Sample size Collection Acquisition Modality n (f, m, u, sa) HS Homo sapiens Human 70 (20, 23, 27, 0) SBU, USNM, USC, WITZ µCT, CT, laser scan PT Pan troglodytes$ Chimpanzee 57 (19, 31, 7, 3) AMNH, LACM, RMCA, USNM µCT, CT, laser scan PP Pan paniscus Bonobo 21 (12, 9, 0, 0) AMNH, RMCA µCT, laser scan GG Gorilla gorilla Western Gorilla 27 (8, 17, 2, 1) AMNH, MCZ, RMCA, USNM µCT, CT, laser scan GBB Gorilla beringei beringei Mountain Gorilla 43 (19, 23, 1, 1) AMNH, MGSP, RMCA, USNM µCT, CT, laser scan GBG Gorilla beringei graueri Eastern Lowland Gorilla 18 (6, 12, 0, 0) AMNH, RMCA µCT, laser scan POA Pongo abelii Sumatran Orangutan 17 (9, 8, 0, 7) USNM µCT, CT POP Pongo pygmaeus Bornean Orangutan 20 (12, 7, 1, 1) AMNH, USNM µCT, CT SS Symphalangus syndactylus Siamang 9 (5, 2, 2, 2) AMNH, USNM µCT HY Hylobates spp.# Gibbon 33 (12, 18, 3, 4) AMNH, MCZ, USNM µCT NL Nasalis larvatus Proboscis Monkey 23 (7, 11, 5, 0) AMNH, ANSP, MCZ µCT PAP Papio spp.† Baboon 19 (9, 5, 5, 0) AMNH, LACM, SBU, USNM µCT AT Ateles spp.§ Spider Monkey 21 (12, 8, 1, 2) AMNH, USNM µCT € Abbreviations: n = total sample size; f = females; m = males; u = unknown sex; sa = subadults; AMNH = American Museum of Natural History; ANSP = Academy of Natural Sciences of Drexel University; MGSP = Mountain Gorilla Skeletal Project; LACM, Natural History Museum of Los Angeles County; MCZ = Harvard University’s Museum of Comparative Zoology; RMCA = Royal Museum for Central Africa; SBU = Stony Brook University; USC = University of Southern California; USNM = Smithsonian Institution’s National Museum of Natural History; WITS, University of the Witwatersrand. $ Includes the following subspecies: Pan troglodytes schweinfurthii; P. t. troglodytes; P. t. verus; and specimens without subspecies information. # Includes the following species: Hylobates agilis; H. klossii; H. lar; H. moloch; H. muelleri; Hoolock hoolock; Nomascus concolor; N. leucogenys; N. gabriellae. † Includes the following species: Papio anubis; P. cynocephalus; P. hamadryas; P. ursinus. § Includes the following species: Ateles belzebuth; A. fusciceps; A. geoffroyi; A. marginatus; A. paniscus. Table 2: Interspecific variation in proximal phalangeal curvature shape. Results of ANOVA and Tukey’s Pairwise Comparisons for PC1 scores (top) and PC2 scores (bottom). Above the diagonal are p values; below the diagonal are Tukey’s Q values. Taxon codes as in Table 1. PC1 Scores - ANOVA Results: F=123.00; df=12,365; p<0.0001 HS PT PP GG GBB GBG POA POP SS HY NL PAP AT HS --- 0.0065 0.0000 0.0000 0.0041 0.0241 0.0000 0.0000 0.0000 0.0000 0.9996 0.0420 0.0000 PT 5.5090 --- 0.9501 0.3655 1.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0000 0.0000 PP 7.6700 2.1610 --- 0.9989 0.9732 0.8032 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 GG 9.0630 3.5540 1.3930 --- 0.4476 0.1700 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 GBB 5.6770 0.1676 1.9930 3.3870 --- 1.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0000 0.0000 GBG 5.0010 0.5083 2.6690 4.0620 0.6758 --- 0.0000 0.0000 0.0000 0.0000 0.0007 0.0000 0.0000 POA 31.7100 26.2000 24.0400 22.6400 26.0300 26.7100 --- 0.0003 0.0000 0.0000 0.0000 0.0000 0.0000 POP 25.1300 19.6200 17.4600 16.0600 19.4500 20.1200 6.5810 --- 0.1103 0.0001 0.0000 0.0000 0.0000 SS 20.8200 15.3200 13.1500 11.7600 15.1500 15.8200 10.8800 4.3000 --- 0.7650 0.0000 0.0000 0.6539 HY 18.0700 12.5600 10.4000 9.0030 12.3900 13.0700 13.6400 7.0590 2.7590 --- 0.0000 0.0000 1.0000 NL 1.2630 6.7730 8.9330 10.3300 6.9400 6.2640 32.9700 26.3900 22.0900 19.3300 --- 0.3915 0.0000 PAP 4.7630 10.2700 12.4300 13.8300 10.4400 9.7640 36.4700 29.8900 25.5900 22.8300 3.5000 --- 0.0000 AT 17.8300 12.3300 10.1600 8.7720 12.1600 12.8300 13.8700 7.2910 2.9900 0.2315 19.1000 22.6000 --- PC2 Scores - ANOVA Results: F=61.13; df=12,365; p<0.0001 HS PT PP GG GBB GBG POA POP SS HY NL PAP AT HS --- 0.8968 1.0000 0.0000 0.0000 0.0000 0.9909 1.0000 0.8731 0.0014 0.0160 0.0073 0.6320 PT 2.3980 --- 0.8850 0.0000 0.0000 0.0000 1.0000 0.9620 0.0325 0.0000 0.0000 0.0000 0.0080 PP 0.0406 2.4390 --- 0.0000 0.0000 0.0000 0.9889 1.0000 0.8857 0.0016 0.0177 0.0081 0.6528 GG 12.0900 9.6890 12.1300 --- 0.9999 0.9836 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 GBB 13.1500 10.7600 13.1900 1.0660 --- 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 GBG 13.9600 11.5700 14.0000 1.8760 0.8102 --- 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 POA 1.7520 0.6461 1.7930 10.3400 11.4000 12.2100 --- 0.9986 0.1260 0.0000 0.0001 0.0000 0.0400 POP 0.3141 2.0840 0.3547 11.7700 12.8400 13.6500 1.4380 --- 0.7503 0.0005 0.0070 0.0030 0.4678 SS 2.4770 4.8760 2.4370 14.5700 15.6300 16.4400 4.2300 2.7910 --- 0.3640 0.7948 0.6541 1.0000 HY 6.0350 8.4330 5.9940 18.1200 19.1900 20.0000 7.7870 6.3490 3.5570 --- 1.0000 1.0000 0.6480 NL 5.1670 7.5650 5.1260 17.2500 18.3200 19.1300 6.9190 5.4810 2.6890 0.8678 --- 1.0000 0.9545 PAP 5.4670 7.8650 5.4270 17.5500 18.6200 19.4300 7.2190 5.7810 2.9900 0.5674 0.3004 --- 0.8864 AT 3.0330 5.4310 2.9920 15.1200 16.1900 17.0000 4.7850 3.3470 0.5556 3.0020 2.1340 2.4340 --- Table 3: Mean PC1 score, included angle (IA), PCF a coefficient, and proportion of time suspensory (SUS). Taxon codes as in Table 1. Taxon Code PC1 score IA (degrees) a coefficient SUS# HS 0.02214 22.26 0.00600 0.00 PT 0.00743 30.07 0.00597 0.01 GG -0.00206 33.30 0.00691 0.01 POP -0.04455 50.12 0.00878 0.59 HY -0.02610 40.88 0.01151 0.51 PAP 0.03486 17.20 0.00720 0.00 AT -0.02548 41.09 0.01243 0.28 # See citations reported in Rein (2011). Appendix Table 1. MorphoSource ID numbers for specimens illustrated in Figure 1 of the main text. Species Side# Museum ID$ MorphoSource Media Homo sapiens R USNM-437R ark:/87602/m4/M154307 Pan troglodytes R USNM-176228 ark:/87602/m4/M154311 Pan paniscus LàR RMCA-20881 ark:/87602/m4/M154310 Gorilla b. beringei LàR USNM-545041 ark:/87602/m4/M154305 Gorilla gorilla R AMNH-54355 ark:/87602/m4/M154306 Pongo abelii R USNM-143598 ark:/87602/m4/M154313 Pongo pygmaeus LàR USNM-145306 ark:/87602/m4/M154314 Symphalangus syndactylus R USNM-143579 ark:/87602/m4/M154315 Hylobates agilis R AMNH-103667 ark:/87602/m4/M154308 Papio cynocephalus R USNM-452509 ark:/87602/m4/M154312 Nasalis larvatus R AMNH-106275 ark:/87602/m4/M154309 Ateles paniscus LàR USNM-361027 ark:/87602/m4/M154304 # Original side studied (àR = specimen mirror imaged from left to right for analysis). $ Museum Abbreviations: AMNH = American Museum of Natural History – Department of Mammalogy; USNM = Smithsonian Institution’s National Museum of Natural History – Division of Mammals or Anthropology; RMCA = Royal Museum of Central Africa – Department of Mammalogy. SUPPLEMENTAL INFORMATION: ADDITIONAL TEXT & FIGURES A geometric morphometric approach to investigate primate proximal phalanx diaphysis shape. Sophie E. Wennemann 1, Kristi L. Lewton 1, 2, Caley M. Orr 3, 4, Sergio Almécija 5, 6, 7, Matthew W. Tocheri 8, 9, 10, William L. Jungers 11, 12, Biren A. Patel 1, 2, * 1 Department of Integrative Anatomical Sciences, Keck School of Medicine, University of Southern California, Los Angeles, CA 90033, USA 2 Human and Evolutionary Biology Section, Department of Biological Sciences, University of Southern California, Los Angeles, CA 90089, USA 3 Department of Cell and Developmental Biology, University of Colorado School of Medicine, Aurora, CO 80045, USA 4 Department of Anthropology, University of Colorado Denver, Denver, CO 80217, USA 5 Division of Anthropology, American Museum of Natural History, Central Park West at 79th Street, New York, NY 10024, USA 6 New York Consortium in Evolutionary Primatology, New York, NY, USA 7 Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona, c/ Columnes s/n, Campus de la UAB, 08193, Cerdanyola del Vallès, Barcelona, Spain 8 Department of Anthropology, Lakehead University, Thunder Bay, Ontario P7B 5E1, Canada 9 Human Origins Program, National Museum of Natural History, Smithsonian Institution, Washington DC 20013, USA 10 Australian Research Council Centre of Excellence for Australian Biodiversity and Heritage, University of Wollongong, Wollongong, New South Wales, 2522, Australia 11 Department of Anatomical Sciences, Stony Brook University, Stony Brook, NY 11794, USA 12 Association Vahatra, BP 3972, Antananarivo 101, Madagascar * Corresponding author: Biren A. Patel, 1333 San Pablo Street, BMT 404, Keck School of Medicine, University of Southern California, Los Angeles CA, 90033, USA; birenpat@usc.edu 1. Adults vs. subadults As indicated in the main text, the majority of all mPP3 specimens in this study were considered skeletally adult and had fully fused proximal epiphyses. However, in order to augment sample sizes, a handful of non-human sub-adult specimens were also included in this study if the proximal epiphysis was in the process of fusion and could not be manually separated from its diaphysis. The taxa that included skeletally sub-adult specimens were Pongo abelii (n=7), Pongo pygmaeus (n=1), Gorilla gorilla (n=1), Gorilla beringei beringei (n=1), Pan troglodytes (n=3), Symphalangus syndactylus (n=2), Hylobates spp. (n=4), and Ateles spp. (n=2) (see Table 1 in main text for additional details). For adult specimens, the most biomechanically relevant region of interest of the diaphysis (dROI) was defined between the palmar-most aspect of the proximal palmar tubercles and the inflection point between the distal trochlea and the diaphysis on the palmar side. For sub-adult specimens, the proximal end of the dROI was defined as the point just distal to the zone of epiphyseal fusion which is always just slightly distal to the palmar tubercles. The distal end of the dROI was the same for adults and subadults. Specimens characterized as subadults were combined with adult specimens in all the 2DGM analyses. As seen in the box- and-whiskers plots, PC1 scores of sub-adults for each taxon are within the ranges of variation observed for their adult counterparts (Fig. S1). Therefore, for those taxa studied here, skeletal subadults (as defined here) do not have proximal phalangeal diaphysis shapes that reflect greater or lesser curvature than skeletal adults with fully fused proximal epiphyses. It is important to acknowledge that the current study did not look at shape in very young individuals including infants and juveniles that lack any fusion of the proximal epiphyses (e.g., Richmond, 1998), therefore additional testing and sensitivity analyses would be needed to confirm if the methods implemented in this study can be used on these specimens and combined with older individuals. The same applies for incomplete and broken proximal phalanges which can be of interest to paleoanthropologists given the fragmentary nature of the fossil record (e.g., Deane & Begun, 2008). Figure S1: Adults vs. subadults. Box-and-whiskers plots of PC1 scores for the 2DGM analysis evaluating manual proximal phalanx diaphysis shape in the comparative sample. Adult specimens (with complete epiphyseal fusion) are shown in solid black circles and subadults (with incomplete and partial epiphyseal plate fusion) are shown in solid red diamonds. Note that subadult specimens for each taxon have PC1 scores, and thus diaphysis shapes, that are within their respective total adult variation observed here. Horizontal lines within each box illustrate the median of the distribution. Boxes envelop the inter quartile range (50% of values) of the sample distribution, and whiskers encompass the range excluding outliers. Filled circles beyond whiskers indicate outliers. Specimens ordered from left to right based on mean PC1 score. Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan paniscus; GG, Gorilla gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; POA, Pongo abelii; POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; PAP, Papio spp.; NL, Nasalis larvatus; AT, Ateles spp. See Table 1 in main text for additional information about the comparative sample. 2. Surface mesh triangle count As noted in the main text, mPP3 surface data derived from a variety of surface scanning and X- ray technologies, as well as researchers, can yield different quality 3D meshes, including different numbers of triangles in the final surface intended for analyses. Ultimately, all our surfaces were either decimated or refined to contain a mesh triangle count of approximately 250,000 triangles. To some readers, this could appear as arbitrary, or possibly as either too high or not enough in number, to ensure accurate results. To assess this issue, we took one orangutan specimen (Pongo abelii AMNH-143598) and created 18 variants of its surface mesh ranging from 2,500 to 2,000,000 triangles (Fig. S2A), thereby mimicking outputs from low-resolution laser surface scans to high-resolution µCT scans, respectively. We then performed all the landmark acquisition and 2DGM procedures outlined in the main text on these variants without altering any other parameter or inputs (i.e., same specimen orientation, same dROI dimensions, etc.). The resulting PC1 (31.1% of the variation) and PC2 (22.4% of the variation) scores of each of the 18 variants of the 2DGM analysis were then subjected to a visualization of convergence to assess the relationship with surface mesh triangle count. Among the 18 variants, those with a surface mesh of >100,000 triangles begin to converge in both PC1 and PC2 scores (Fig. S2B and S2C). Based on these observations, we concluded that our ‘gold standard’ defined as a surface mesh with a count of ~250,000 triangles is sufficiently higher than is necessary, but also without being too excessive. Figure S2: Surface mesh triangle count. A) 3D surface models of the manual proximal phalanx of digit 3 (mPP3) in medial view of a Pongo abelii (USNM-143598) with varying numbers of surface mesh triangles (top to bottom: 2,500 triangles; 20,000 triangles; 250,000 triangles; 1,000,000 triangles). Notice that the quality of mesh visually improves after reaching >20,000 triangles. B) Bivariate scatter plot of surface mesh triangle count vs. PC1 score. Notice that variation in PC1 scores begin to stabilize after a model has 100,000 triangles. C) Bivariate scatter plot of surface mesh triangle count vs. PC2 score. Notice that variation in PC2 scores begin to stabilize after a model has 100,000 triangles. The number of surface mesh triangles for all specimens of all taxa ultimately used in this study approximated 250,000 (seen here as the gold diamond for USNM-143598). 3. Shape variation along PC1 and PC2 Figure S3: Maximum and minimum manual proximal phalanx diaphysis shapes (black filled circles) derived from the Principal Component Analysis conducted on Procrustes coordinates for the comparative sample (see Figure 8 in the main text). Sample mean shapes (gray filled circles) compared to A) maximum PC1 shape, B) minimum PC1 shape, C) maximum PC2 shape, and D) minimum PC2 shape. 4. Size Correlation Figure S4. Size Correlation. Bivariate scatter plot of mean natural log (ln) dROI length vs. mean PC1 score evaluating possible size correlation. A) All taxa analysis (n=13) has a significant correlation (r=-0.607; p=0.028). B) With Pongo and Papio removed (n=10) the correlation is not significant (r=0.163; p=0.653). Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan paniscus; GG, Gorilla gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; POA, Pongo abelii; POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; PAP, Papio spp.; NL, Nasalis larvatus; AT, Ateles spp. See Table 1 for additional information about the comparative sample. 5. Method to calculate IA and a from 2DGM landmarks PC1 scores from the 2DGM analysis of the entire comparative sample of 378 specimens were correlated with each specimen’s included angle (IA; measured in degrees) and its second-degree polynomial a coefficient. Both IA and a were determined from the same set of 2D landmarks of the dROI needed for the 2DGM shape analysis (Fig. S5). Three of the landmarks – proximal [#1], middle [#50], and distal [#100] – were used to calculate IA from overall length (L) and midshaft height (H) following modified calculations presented in Jungers et al. (1997). All 100 landmarks were used to fit a second-degree polynomial to determine the a coefficient following Deane, Kremer & Begun (2005). Calculations for both IA and a were performed in Excel software. One important difference between Deane, Kremer & Begun’s (2005) study and the present one is that the second-degree polynomial was applied to the dorsal surface of the phalanx’s diaphysis in the former study whereas the second-order polynomial was applied to the bone’s longitudinal proximodistal central axis in present study (compare Fig. S5 below with Fig. 2B in the main text); the approach used here in the present study is more biomechanically appropriate. Figure S5: Method to calculate IA and a from 2DGM landmarks. Methodology to quantify the included angle and the second-degree polynomial a coefficient from the 100 2D landmarks used in the 2DGM analysis described in the main text. The included angle and a coefficient for each specimen were then correlated with its PC1 score to assess comparability between methods used to assess proximal phalangeal curvature and proximal phalanx diaphysis shape (see also Figures 11 and 12 in the main text; see also Fig. S6). 6. 2DGM vs. IA vs. PCF Figure S6: 2DGM vs. IA vs. PCF. Box-and-whiskers plots of PC1 scores (top, blue), included angles in degrees (middle, red), and PCF a coefficients (bottom, black) derived from the same 100 landmark datasets as described in the main text (see also Fig. S5). Note that PC1 scores and included angles do a better job in discriminating between taxa based on habitual locomotor modes, whereas PCF a coefficients are only good at distinguishing highly suspensory taxa (i.e., Pongo, hylobatids, and Ateles) from all other groups. Horizontal lines within each box illustrate the median of the distribution. Boxes envelop the inter quartile range (50% of values) of the sample distribution, and whiskers encompass the range excluding outliers. Filled circles beyond whiskers indicate outliers. Specimens ordered from left to right based on mean PC1 score. Abbreviations: HS, Homo sapiens; PT, Pan troglodytes; PP, Pan paniscus; GG, Gorilla gorilla; GBB, Gorilla beringei beringei; GBG; Gorilla beringei graueri; POA, Pongo abelii; POP, Pongo pygmaeus; SS, Symphalangus syndactylus; HY, Hylobates spp.; PAP, Papio spp.; NL, Nasalis larvatus; AT, Ateles spp. See Table 1 in main text for additional information about the comparative sample. 7. R code for 2DGM analysis library(geomorph) #Read morphologika files #Morphologika files should be in one folder with the read.multi.morphologika support code available on Github: https://github.com/EmSherratt/MorphometricSupportCode/blob/master/read.multi.morphologika.R filelist <- list.files(pattern = "*.txt") mPP <- read.morphologika(filelist) dimnames(mPP)[[3]] <- gsub(".txt","",filelist) #Perform Generalized Procrustes Analysis mPP.gpa <- gpagen(mPP, PrinAxes=TRUE, max.iter=NULL, Proj=TRUE, print.progress=TRUE) #Examine outliers plotOutliers(mPP.gpa$coords) #Perform Principal Component Analysis & export table of PC scores PCA <- gm.prcomp(mPP.gpa$coords) summary(PCA) write.table(PCA$x, "PCScores.txt", sep="\t") #Visualize shapes at extremes of principal component axes ref<-mshape(mPP.gpa$coords) PC1min <- plotRefToTarget(ref,PCA$shapes$shapes.comp1$min, method="points") PC1max <- plotRefToTarget(ref,PCA$shapes$shapes.comp1$max, method="points") PC2min <- plotRefToTarget(ref,PCA$shapes$shapes.comp2$min, method="points") PC2max <- plotRefToTarget(ref,PCA$shapes$shapes.comp2$max, method="points") 8. Morphologika Data Landmark data files in Morphologika format for the entire comparative sample of humans, apes, and monkeys (n=378) are available on journal's website to download as Zip file. The Zip file also contains a single .cvs file with specimen classifier information and the "read.multi.morphologika" as noted above in #7. 9. References cited in Supplemental Information Deane, A. S., Kremer, E. P., & Begun, D. R. (2005). New approach to quantifying anatomical curvatures using high-resolution polynomial curve fitting (HR-PCF). American Journal of Physical Anthropology. 128, 630–638. https://doi.org/10.1002/ajpa.20202 Deane, A. S., & Begun, D. R. (2008). Broken fingers: retesting locomotor hypotheses for fossil hominoids using fragmentary proximal phalanges and high-resolution polynomial curve fitting (HR-PCF). Journal of Human Evolution. 55, 691–701. https://doi.org/10.1016/j.jhevol.2008.05.005 Jungers, W. L., Godfrey, L. R., Simons, E. L., & Chatrath, P. S. (1997). Phalangeal curvature and positional behavior in extinct sloth lemurs (Primates, Palaeopropithecidae). Proceedings of the National Academy of Sciences. 94, 11998–12001. https://doi.org/10.1073/pnas.94.22.11998 Richmond, B. G. (1998). Ontogeny and biomechanics of phalangeal form in primates. Ph.D. Dissertation, State University of New York at Stony Brook, Stony Brook.