The SKX 16699 phalanx (Fig. 1) is a complete pedal phalanx, with minimal erosion to its proximal and distal ends. It was associated with sediments from the Lower Bank of Member 1 of the Swartkrans deposits and thus dates to ≈ 1.8 Ma (Balter et al., 2008). It is likely to be from the right side based a lateral orientation (7°) of the distal trochlea. A right side designation is further supported by a small tubercle just distal of the proximal capsular attachment on the right side in dorsal view. If this tubercle represents the insertion of the abductor digiti minimi tendon, it would be from the fifth digit. Yet, the proximal epiphysis is virtually round (breadth/height ratio of 1.03 [1.01 with the extra plantar growth – see below]), which places it below the same ratio for all other Pliocene and Pleistocene hominin fifth proximal phalanges (1.07–1.34; n = 22) but within the range of second to fourth ones (0.89–1.28; n = 93). The tubercle could be for the tendon of the third or fourth dorsal interosseus muscle, but it is unlikely to be from the second one given the absence of a medial proximal tubercle for the first dorsal interosseus tendon. The phalanx is therefore most likely from the third, fourth or fifth digit.

There is a small bulge on the lateral side of the plantar proximal epiphysis, which has a slightly irregular and discrete border (Fig. 1). It is a small growth of bone, associated with the lateral plantar ligament, but expanded beyond the usual small tubercle for the ligament. The proximal maximum height has been adjusted (from 8.8 to 8.6 mm) to account for this extra growth.

The SKX 16699 phalanx is qualitatively and quantitatively compared to samples of proximal lateral pedal phalanges of (frequently shod) Upper Paleolithic modern humans (n = 80), habitually unshod Middle Paleolithic early modern and late archaic humans (n = 85), the Dinaledi (H. naledi) sample (n = 9), and australopith (n = 18) remains referred to Australopithecus afarensis, Australopithecus africanus and Paranthropus robustus, plus the taxonomically uncertain Burtele Pliocene foot (BRT-VP-2/73) (the individual specimens are listed in Table S1). The Australopithecus afarensis phalanges are from Hadar (one from A.L. 288-1y, four from A.L. 333-115, and seven isolated ones from the A.L. 333 locality). The Australopithecus africanus and P. robustus phalanges are from Sterkfontein (StW 355) and Drimolen (DNH-117), respectively. Phalanges from digits 2 to 5 are included, given the difficulties in assigning isolated ones to digits 2 to 4. Recent human reference data for diaphyseal robusticity are from the habitually unshod Pecos Pueblo, New Mexico late prehistoric Amerindian sample (Trinkaus, 2005). The SKX 16699, StW 355, most Late Pleistocene and Holocene data are from personal measurement of the original specimens. The other phalangeal data derive from the literature and personal communications (Table S1).

The assessment of SKX 16699 (Fig. 1) is based on the original specimen housed in the Ditsong National Museum of Natural History (Pretoria, South Africa), combined with high-resolution micro-CT scans (23.5 μm voxel resolution) of the bone obtained with a Nikon Metrology XTH 225/320 LC dual source industrial CT system located at the Microfocus X-ray CT facility of the University of the Witwatersrand. The lengths and diameters (Table S1) were measured with calipers (and verified on the digital scan data) following Trinkaus (1983). An ‘articular angle’ was measured in the midline parasagittal plane of the phalanx, between the tangent to the metatarsal facet and the interarticular axis of the phalanx (Fig. S1); a proximodorsal orientation of the facet provides an angle > 90°. This angle is modestly less than the more commonly measured angle between the metatarsal facet and the plantar plane of the phalanx (i.e., the ‘dorsal canting’ angle; Duncan et al., 1994, Haile-Selassie et al., 2012 and Ward et al., 2012). The articular angle better represents the functional (joint reaction force) orientation of the proximal phalanx in the latter portions of stance phase, given the dorsiflexion of the metatarsophalangeal joint and the associated plantarflexion of the interphalangeal joints at heel-off (Martin, 2011). Published ‘dorsal canting’ angles have been converted to ‘articular angles’ by subtracting the angles, in lateral view, between the plantar tangent and the interarticular axis from the available ‘dorsal canting’ angles (Fig. S1). All of these angles are likely to have measurements errors of ± 1°–2°, given irregularities in the articular facet margins unrelated to the predominant orientations of the facets.

Given differences in Australopithecus versus Homo in relative toe lengths (Latimer and Lovejoy, 1990, Susman et al., 1984 and White and Suwa, 1987), relative phalanx length was assessed using an index of phalangeal length to available femoral head diameters (see Table S2 for Pliocene and Early Pleistocene femoral head diameters). Femoral head diameter is employed as a reflection of body mass (Auerbach and Ruff, 2004); it scales similarly across Pliocene and Early/Middle Pleistocene hominin samples to other skeletal reflections of body mass, and it is insignificantly proportionately smaller in the earlier samples than it is among Late Pleistocene and recent humans (Grabowski et al., 2015; Trinkaus, unpub. data). Associated femoral head diameters are used as available for A.L. 288-1 and the Late Pleistocene Homo phalanx samples. An average of 37.4 mm (± 4.1 mm; from A.L. 152-2, 333-3 and 827-1) is used for the other Australopithecus afarensis phalanges (data from Ward et al., 2012). A mean Australopithecus africanus value of 33.2 mm (± 2.4 mm, n = 7) is used for StW 355, and a mean P. robustus value of 32.2 mm (± 3.3 mm, n = 6) is employed for DNH-117 (data from Ruff, 2010). The H. naledi lengths are compared to the mean of the two available femoral head diameters (35.5 mm) (Marchi et al., 2016). SKX 16699 is compared to both the P. robustus mean and an early Homo average of 42.8 mm (± 3.3 mm, n = 4; data from Ruff, 2010) given its taxonomic uncertainty. Femoral head diameters are not available for the Burtele foot. To assess phalangeal robusticity, the diaphysis is modeled as a solid beam and the diaphyseal polar moment of area (J) calculated from the midshaft diameters using ellipse formulae (O’Neill and Ruff, 2004). Polar moments of area are compared to interarticular length since body mass estimates are unavailable for all of the Pliocene and Earlier Pleistocene pedal phalanges, except for A.L. 288-1y. The raw residuals from the reduced major axis line of the polar moment of area versus interarticular length for the recent human habitually unshod digits 2 to 5 sample are then compared (Ln J = 4.243 × Ln length – 8.43, r = 0.616, n = 171).

Given that variable numbers of phalanges are available by individual, and that it is not always possible to reliably assign phalanges to a digit, SKX 16699 is compared to values for proximal phalanges from digits 2 to 5 individually. To avoid duplication by individual for the associated sets of phalanges, the comparisons employ repeated measure ANOVAs [calculations done with NCSS 8.0.16 (Hintze, 2012)]. Phalanx length is not different across the reference samples (P = 0.453), but the other comparisons are all significant at P < 0.0001.

SKX 16699's interarticular length is 18.6 mm, which is shorter than those of 96.6% (n = 179) of the other fossil hominin pedal phalanges (Fig. 2A). It is among the smaller of the later Homo phalanges, and modestly longer than one of H. naledi (U.W. 101-1034). It falls well below those of BRT-VP-2/73, Australopithecus afarensis, Australopithecus africanus and P. robustus; among the australopiths, its length is approached only by those of A.L. 288-1y (21.0 mm), StW 355 (23.4 mm) and DNH-117 (22.2 mm).

When phalanx length is scaled to femoral head diameter as an index (Fig. 2B), the australopiths separate from the Late Pleistocene hominins. The smallest Australopithecus afarensis index (67.6) is for the A.L. 333-154 phalanx; the largest Late Pleistocene indices are for the Qafzeh 8 (65.2) and Veneri 2 (65.0) second phalanges. The StW 355 and DNH-117 phalanges, scaled against Australopithecus africanus and P. robustus mean femoral head diameters respectively, provide indices of 70.5 and 68.9, respectively, at the bottom of the Australopithecus afarensis range. The H. naledi indices fall among the relatively longer of the Late Pleistocene humans.

The values for SKX 16699 as a P. robustus (57.8) or an early Homo (45.5) bracket the interquartile ranges for Late Pleistocene Homo and are distinct from the australopith range. If the three shortest australopith phalanges (A.L. 288-1y, StW 355, DNH-117) are scaled to the smallest known femoral head diameters for their respective species (A.L. 288-1: 28.0 mm; Sts 14: ≈ 30.8 mm; SK 3121: 28.8 mm) and SKX 16699 is scaled to the smallest of them, assuming that they all derive from small individuals, their resultant indices are 78.6, 76.0, 77.1 and 66.4 respectively. Thus, SKX 16699 remains separate from the australopith specimens, if slightly above the highest later Homo values. SKX 16699 therefore has both an absolutely and, to the extent assessable, relatively short length, similar to those of Homo.

Comparing each phalanx length separately to each of the femoral head diameters within each taxonomic group (Table S2) modestly increases the ranges of indices for each sample or specimen. However, it does not change the overall pattern (Fig. S2), and the comparative samples remain significantly different from each other. The larger indices for SKX 16699 using P. robustus femoral head diameters overlap the smallest values for Australopithecus afarensis, StW 355 and DNH-117, but the majority of its indices using P. robustus femoral heads and all of the ones using early Homo values separate it from the australopith specimens.

The robusticity of the SKX 16699 diaphysis is assessed by comparing its midshaft polar moment of area (modeled as a solid beam) to those of the other hominin fossil samples (Fig. 3A). The two Late Pleistocene samples follow the pattern of habitually unshod and shod recent humans (Middle and Upper Paleolithic remains, respectively) (Trinkaus, 2005). The Australopithecus afarensis phalanges (including the small A.L. 288-1y) and the two phalanges from BRT-VP-2/73 fall along the gracile margin of the habitually shod human sample. Their phalangeal “gracility” is likely a result of their phalangeal lengths, given that they are relatively long and australopith postcrania are otherwise robust (Ruff et al., 1999). The H. naledi phalanges align with the Middle Paleolithic sample. The two later australopith phalanges, StW 355 and DNH-117, are close to the H. naledi phalanges and among the Late Pleistocene ones. SKX 16699 is within the range of both the Middle and Upper Paleolithic samples, close to those of H. naledi, distinct from the Australopithecus afarensis ones, yet in line with the two later australopiths from South Africa (StW 355 and DNH-117).

The distributions of their residuals from a recent human unshod sample line (Fig. 3B) places SKX 16699 between the medians of the two later Homo phalanx samples and among the more gracile of the H. naledi and Middle Paleolithic phalanges. It is similar to the values for the two later South African australopith specimens, above the Burtele values, and well separated from the Australopithecus afarensis sample. It is unclear to what extent the differences in earlier hominin phalangeal robustness are driven by shaft diameters versus length (Fig. 2); the similarity of SKX 16699 to StW 355 and DNH-117 may be due in part to their modest lengths, yet A.L. 288-1y remains gracile along with the other Australopithecus afarensis phalanges.

The dorsal contour of SKX 16699 in lateral view exhibits a modest convexity, one that is evenly distributed proximodistally along the diaphysis (Fig. 1); its plantar contour is only slightly concave. It is therefore more curved dorsally than are those of most Late Pleistocene and recent humans, which have a straight or slightly convex dorsal diaphyseal profile (Fig. 4). Yet, its dorsal curvature is substantially less than those of Australopithecus afarensis, Australopithecus africanus and BRT-VP-2/73 (Fig. 4; cf. Haile-Selassie et al., 2012, Latimer et al., 1982 and Ward et al., 2012). It is similar to those of the H. naledi phalanges, despite their phalanges being described as having “significantly greater curvature than those of modern humans” (Harcourt-Smith et al., 2015:4). Yet, the slight plantar concavity of SKX 16699 closely resembles those of later Homo phalanges and is close to those of the H. naledi phalanges. It contrasts with the distinctly plantarly concave proximal phalanges of Australopithecus afarensis, StW 355 and DNH-117 (Fig. 4; cf. Latimer et al., 1982, Vernon, 2013 and Ward et al., 2012).

Susman et al. (2001) provided an “included angle” of 22° for SKX 16699, a measure of its midline curvature and hence one that combines both the dorsal convexity and the degree of dorsoplantar expansion of the proximal diaphysis (Stern et al., 1995 and Susman et al., 1984). This angle is substantially below the value of 36° for both A.L. 288-1y and StW 355, which are similar to other Australopithecus afarensis proximal pedal phalanges, but among the more curved of a recent human sample (Susman et al., 1984 and Susman et al., 2001). Yet, the diaphysis of SKX 16699 expands more proximally than do those of Australopithecus (albeit not as much as some later Homo phalanges) (Fig. 4), such that the “included angle” likely minimizes the difference in dorsal curvature between it and those of Australopithecus. SKX 16699 also contrasts with DNH-117, whose curvature is similar to StW 355 and the Australopithecus afarensis phalanges (Vernon, 2013).

Hominin pedal phalanges have variably developed medial and lateral sulci, with or without small crests, for the attachments of the flexor digitorum longus and brevis muscles’ fibrous sheaths. The sulci extend from the proximal metatarsophalangeal capsular attachment to an area close to the distal trochlea. In later Pleistocene humans, they are marked, sometimes with small crests, on the proximal-to-middle shaft and in the dorsoplantar middle of the medial and lateral diaphysis (Fig. 4). On the Australopithecus afarensis and StW 355 phalanges (Fig. 4; cf. Ward et al., 2012), they are variably evident along the full diaphysis and are more developed on the distal half to two-thirds of the diaphysis. Their sulci are commonly bordered medially and laterally by small crests extending along the middle to distal diaphysis, located medioplantarly and lateroplantarly as opposed to the more directly medial and lateral positions of midshaft crests on later Homo diaphyses. The DNH-117 phalanx resembles the earlier Australopithecus ones in this respect. The H. naledi configurations are similar to the pattern seen in later Homo.

The fibrous flexor sheath attachments of SKX 16699 (Fig. 1) resemble those of Late Pleistocene Homo and H. naledi; they are distinct only on the proximal two-thirds of the diaphysis with the deepest portions of the sulci and the associated medial and lateral projections centered approximately one-third of the distance from the proximal epiphysis. The rounded crests are dorsoplantarly located mid-medial and mid-lateral.

As a result of the more distal development of the fibrous sheath attachments on Australopithecus pedal phalanges, the widest portions of most of their diaphyses are the distal halves and the narrowest points are close to the proximal epiphysis (Fig. 4). The exceptions are A.L. 333-168 and DNH-117, which have largely parallel medial and lateral diaphyseal margins (Vernon, 2013 and Ward et al., 2012). The H. naledi phalanges largely have a midshaft bulge for the flexor sheath crests, with similar mediolateral constrictions proximal and distal of the crests. Later Homo pedal proximal phalanges, in contrast, have the narrow portion of the diaphysis either at midshaft or towards the distal shaft (Fig. 4). SKX 16699 has its narrowest portion slightly distal of midshaft, thereby giving it an “hour-glass” profile (Susman et al., 2001).

Investigations of the orientation of the metatarsal facet on early hominin lateral proximal pedal phalanges (Duncan et al., 1994, Fernández, 2016, Haile-Selassie et al., 2012, Latimer and Lovejoy, 1990, Susman et al., 1984, Vernon, 2013 and Ward et al., 2012) have noted the predominantly proximodorsal orientation of the proximal phalangeal facet (or ‘dorsal canting’), and these observations are combined with metatarsal indications of some degree of metatarsophalangeal dorsiflexion (Latimer and Lovejoy, 1990 and Zipfel et al., 2009). On the proximal phalanges, assessment of the angle between the mid-facet tangent and the articular axis (the articular angle; Fig. 5) shows that the BRT-VP-2/73, StW 355 and most of the Australopithecus afarensis facets (the angle is not available for DNH-117) are angled slightly dorsal of directly proximal, with the more angled of the phalanges minimally overlapping the range of variation of the Late Pleistocene humans. The habitually unshod human sample has slightly higher angles than the habitually shod one (P = 0.102) (cf. Griffin and Richmond, 2010). The H. naledi sample exhibits angles well within the later Homo range. The SKX 16699 phalanx, with an angle of 110°, is among the more angled of the later Pleistocene humans. It is distinct from the earlier hominin range and above the values for the H. naledi sample.

In lateral view, the medial and lateral sides of the phalangeal base (i.e., proximal epiphysis) of Late Pleistocene and recent humans project most proximally at the plantar articular margin because the collateral (and plantar) metatarsophalangeal ligaments insert plantarly on the phalanx (Martin, 2011). The epiphyses of BRT-VP-2/73 and Australopithecus project proximally near the dorsoplantar middles of articulations (Fig. 4; cf. Haile-Selassie et al., 2012, Latimer et al., 1982, Susman et al., 1984 and Ward et al., 2012), whereas that of DNH-117 has the proximal projection slightly plantar of the dorsoplantar middle of the base (Vernon, 2013). In this feature, SKX 16699 is intermediate between Australopithecus and later Homo phalanges, in that its proximal projection is ≈ 35% of the dorsoplantar distance from the plantar margin [assessed on the medial side (Fig. 1) to avoid the plantar growth]. The H. naledi phalanges are mostly damaged on the sides of their bases, but they (e.g., U.W. 101-1395) appear to exhibit a pattern closer to that of later Pleistocene humans.