Palaeotragus inexspectatus ( Samson & Radulesco, 1966 )

Palaeotragus inexspectatus ( Samson & Radulesco, 1966)

( Figs 2–13 View Figure 2 View Figure 3 View Figure 4 View Figure 5 View Figure 6 View Figure 7 View Figure 8 View Figure 9 View Figure 10 View Figure 11 View Figure 12 View Figure 13 )

Holotype: distal part of left m3 (№ 5227, Institute of Speleology ‘Emil Racoviţă’, Bucharest).

Synonymy: As in Athanassiou (2014) and:

Mitilanotherium martinii Mayda et al. (2016) .

Mitilanotherium ineospectatum Terhune et al. (2020): p. 20 , fig 4.

Palaeotragus ineopectatum Sianis (2023): p. 65 , fig. 19.

Mitilanotherium ineospectatum Croitor et al. (2024): p. 513 .

Amended diagnosis: Palaeotragus of large size. Skull with very elongated postorbital cranial part. Ossicones tilting rostrally and positioned strongly laterally, right above the orbits. Frontals inflated and possibly rugose in the peri-ossicone area, and concave in between the ossicones. Postorbital part of the skull slightly descending. Parietals concave laterally. Temporal fossa very elongated and wide; temporal lines strong and dorsally located. Shortened premolar rows, compared with the molars; p3 and p4 similar to each other and semi-molarized; dp3 similar to p3. Radius with an enlarged, rugose, and rectangular radial tuberosity; furrow for the extensor carpi radialis relatively narrow; articular facet for the semilunate more dorsally located than the others. Long and moderately robust metacarpals, with a dorsal medial tuberosity in the medial epicondyle, which becomes thinner along the shaft, forming the medial ridge of a short dorsal trough. Rather square-shaped astragalus in comparison to other giraffids, with sharp and protruding medial bulge of the collum tali and with tilted medial ridge. Ventral head of the lateral epicondyle of the metatarsal more distally located than the dorsal head; metatarsal articular facets for the intermediolateral cuneiform and for the cubonavicular of roughly the same size and shape; metatarsal articular facets for the intermediolateral and medial cuneiforms very close to each other; metatarsals with very pronounced dorsal trough that reaches the distal epiphysis.

Differential diagnosis: Toothrow comparable in size to those of Palaeotragus berislavicus Korotkevitch, 1957 and Palaeotragus coelophrys (Rodler and Weithofer, 1890) , and postcranials comparable to those of P. coelophrys . The temporal lines are stronger, longer, and more dorsally located than in other Palaeotragus species. The lower premolar/molar (p/m) length ratio is possibly lower than in known Miocene Palaeotragus species. Metapodials are of comparable length to P. rouenii , but more robust.

Descriptions

Dafnero-1 fossil site

Material: A right partial upper toothrow with P4–M3 (DFN-28; Fig. 2 View Figure 2 ); a right radius (DFN 69); a left tibia (DFN-150; see Supporting Information, Appendix S1; Fig. 3 View Figure 3 A-D); and a left metatarsal (DFN-68; Fig. 3 View Figure 3 E-G).

Partial upper toothrow: The P4 of the partial toothrow ( Fig. 2 View Figure 2 ) is sub-rectangular in occlusal view. The enamel is finely rugose lingually. The fossette is wide and simple. The paracone is very prominent and blunt, owing to wear. The parastyle and metacone are faint. A faint buccal cingulum is observed at the base of the tooth. Lingually, the hypocone cusp is blunt. Distally of the hypocone there is a well-developed hypoconal fold.

The M1 and M2 are very similar to each other; they both have finely rugose enamel lingually, more rugose than in P4. Moreover, they both have a weak buccal and lingual cingulum at the base of the tooth. In lingual view, the most prominent feature is the paracone, whereas the metacone is weak. The parastyle and the mesostyle are moderately developed, and the metastyle is fairly weak. The praehypocrista is more developed than the neocrista and protrudes in the mesial fossette; thus, the neocrista tilts mesially. In occlusal view, the hypocone is slightly less prominent lingually than the protocone.

The distal lobe of M3 is ~20% less wide than the mesial one, in contrast to M1 and M2, in which the two lobes are of almost equal width. The enamel is as rugose as in M1 and M2. The parastyle and paracone are strong. In contrast to the M1 and M2, the mesostyle is very thin. The metacone is also much weaker than those of M1 and M2, and the metastyle is extremely weak. Buccally, at the height of the metastyle there is a stylid, probably created by a distal buccal cingulum, that tilts mesially at almost half of the crown of the tooth. The praehypocrista is more developed than the neocrista, obliquely oriented and protruding into the mesial fossette, as in the mesial molars. The posthypocrista is very thin and touches the postmetacrista at about the middle of the crown height. The hypocone is narrower than the protocone.

Radius: The only available radius, DFN-69, has an RI of 10.91. It is slightly dorsopalmarly curved. The proximal epiphysis is somewhat damaged; however, the lateral articular facet for the humerus is concave and has the shape of half an arch, while the medial is rectangular and subdivided into two surfaces. Traces of the coronoid process are barely noticeable. Dorsally, the most prominent feature is the radial tuberosity, which is, however, blunt. Distally on the dorsal side, the furrow for the extensor carpi radialis is almost parallel to the axis of the bone, and its medial border tilts slightly. The furrow for the extensor digitorum lateralis is shallow. The radial styloid process is vertical and blunt in medial view. The ulnar styloid process is more pointed in lateral view. The palmar surface of the radius is slightly convex. The glenoid cavities for the carpal bones are relatively deep. The glenoid cavity for the semilunate is transversely placed compared with the mediolateral axis, and it is more dorsally located than the other two articular facets. The facet for the scaphoid is obliquely oriented and slightly smaller. The articular facet for the triquetrum has the form of a tilting (almost transverse to the mediolateral axis) furrow, which is located more palmarly than the articular facet for the semilunate. The crest that separates the two latter articular facets is strong.

Metatarsal: The metatarsal DFN-68 ( Fig. 3E–G View Figure 3 ) is slender, with an RI of 7.5. The ventral head of the lateral epicondyle is located more distally than the rest of the heads of the epicondyles. The articular facet for the cubonavicular and that for the intermediolateral cuneiform occupy equal areas. The articular facet for the cubonavicular is somewhat kidney shaped, wide and flat.The articular facet for the intermediolateral cuneiform is narrower, semicircular in shape, and it is curved dorsally, ending in a bony tip at the point where it touches the cubonavicular articular facet. The articular facet for the medial cuneiform is elliptical, located laterally to the dorsoplantar axis of the bone, forming an angle of 45° with it. It is almost in contact with the facet for the intermediolateral cuneiform. The synovial fossa is large and elliptical, and it is aligned with the dorsoplantar axis of the bone. The pygmaios is relatively small and laterally located. The central trough is very deep and narrow at the proximal one-quarter of the bone; it becomes wider at the middle of the shaft. The lateral and medial ridges are sharp. The medial ridge is stronger and extends more plantarly than the lateral one. In medial view, the lateral ridge is curved, whereas the medial one is straight; they both disappear towards the fifth distalmost part of the bone, where the central trough flattens. At that point of the shaft there is a blunt and narrow pyramidal rise. The dorsal trough is well developed and runs along the whole shaft, down to the distal condyles. It is wide towards the proximal part of the shaft, and it becomes narrower more distally. The medial ridge of that trough is more pronounced than the lateral one. The keels of the distal condyles are equally developed plantarly, distally, and dorsally.

Dafnero-3 fossil site

Material: Partial cranium bearing the ossicones (DFN3-351a; Fig. 4 View Figure 4 ) and articulated atlas (DFN3-351b); left distal humerus (see Supporting Information, Appendix S1; Fig. 5 View Figure 5 A-B) articulated with a radius–ulna (DFN3-347a, b); right radius–ulna (DFN3-350; Fig. 5 View Figure 5 C-F); left distal radius (DFN3-344); right scaphoid and magnetotrapezoid (DFN3-50a, b; see Supporting Information, Appendix S1); left scaphoid, semilunate, triquetrum, and pisiform (DFN3-353a–d; see Supporting Information, Appendix S1); left scaphoid (DFN3-352; see Supporting Information, Appendix S1); left metacarpal (DFN3- 1); right proximal metacarpal (DFN3-345; Fig. 5 View Figure 5 G-I); distal metacarpal (DFN3-349); right distal tibia (DFN3-355; see Supporting Information, Appendix S1); right astragalus (DFN3- 2; Fig. 6 View Figure 6 A-D; and DFN3-64); right cubonavicular (DFN3-354; see Supporting Information, Appendix S1); right metatarsal (DFN3-346; Fig. 6 View Figure 6 E-G); right proximal metatarsal (DFN3-91); and distal metapodial (DFN 3-348).

Cranium: The ossicones of the partial cranium DFN3-351a ( Fig. 4 View Figure 4 ) are almost straight, only slightly curved backwards. They are inserted right above the orbits and widely apart on the frontals, having a very lateral position. It is difficult to measure the distance between the ossicones owing to the deformation of the skull; however, they are probably 85–90 mm apart.

The ossicones form a slightly obtuse angle with the postorbital part of the cranium, possibly between 105° and 110°. The ossicones are thicker at the base and become narrower more distally, with their cross-section changing from sub-triangular near the base to oval upwards. The ossicones seem to be dense internally and, judging by the better-preserved left ossicone, their surfaces are smooth. A shallow suture is observed at the point of their attachment on the cranium. Both apices are broken. The minimal basal diameter of the better-preserved left ossicone is ~ 55 mm.

The rostral part of the cranium (premaxillae to lacrimals) is completely missing, and the cranium is compressed mediolaterally in such a way that its right part has moved more rostrally than the left. The frontals are pneumatized with small sinuses, now filled with sediment. In lateral view, the postorbital part of the cranium is highly elongated and becomes narrower caudally. The maximum cranial width is likely to be located at the orbits, although we cannot be sure about the preorbital part owing to the poor preservation. In lateral view, the postorbital part is straight and seems to be somewhat descending caudally. The squamous part of the temporal is preserved caudally. The zygomatic process of the temporal is slender dorsoventrally; it seems to direct dorsally, although this might be attributable to taphonomic dislocation. The temporal fossa is highly elongated, in accordance with the elongation of the postorbital part of the cranium, and sub-rectangular in lateral view. The temporal line is very dorsally located and strong. The nuchal crest is not preserved. The auditory bulla is distinct on the right side; it lies 10 mm anterior to the caudal margin of the cranium, and it is relatively large and almost spherical, with a maximum diameter of 9–10 mm.

In dorsal view, the postorbital part of the cranium is trapezoidal in shape, being wider towards the orbits and narrower caudally. The frontoparietal suture is not traceable. Sigmoid grooves are observed ~ 30–40 mm posterior to the ossicone bases, possibly indicating the frontoparietal border. Although the area of the frontals between the ossicones is deformed, it seems that it was concave. The frontals are inflated in the peri-ossicone area. Those inflations might also be rugose posteriorly. The parietal is mostly flat dorsally but seems to become concave anteriorly and laterally. The caudal-most margin of the skull is straight in dorsal view.

In ventral view, the rostrally protruding muscular processes of the temporal bone are clear but mostly broken. The basioccipital and basisphenoid bones create an almost straight angle. In caudal view, the articulation with the atlas does not allow for detailed observation of the occipital bone. However, the occipital condyles seem to be very enlarged. The jugular process is more ventrally protruding than the occipital condyles.

Radius: Three radii are preserved. The radius DFN3-347 b is much thicker transversely than the other two. It also demonstrates a deep dorsal trough on the distalmost half of the shaft. It cannot be excluded that these features are a result of dorsopalmar compression of the bone. In contrast, DFN3-350 ( Fig. 5C–F View Figure 5 ) is transversely compressed, giving the impression of a more slender bone. The only available complete proximal epiphysis is that of DFN3-350 . The proximal articular surface of DFN3-350 is less broad and deeper than that from DFN. The lateral glenoid cavity for the humerus has the shape of a quadrant, whereas the medial cavity is rectangular and subdivided by a blunt lateral crest. The proximal articular surface of DFN3-347 b is much broader and less deep. The lateral glenoid cavity of that bone is rectangular and very broad. The smooth crest that divides that cavity in two is pointed dorsally. The coronoid process is pointed in DFN3- 350 . All the radii are slightly curved cranially; DFN3-347 b is more curved than DFN3-350 . The radial tuberosity is very broad and textured. The furrow for the extensor carpi radialis seems broad, almost parallel to the distoproximal axis of the bone, and textured; in DFN3-347 b it is also very deep. The furrow for the extensor digitorum lateralis is very deep in DFN3-344 and more subtle in the other two specimens, possibly owing to some taphonomic deformation. Distally, the radial styloid process is blunt in medial view, while the ulnar process is pointed. The two articular facets for the scaphoid and the semilunate are almost equal in size; the semilunate facet is more dorsally located. The scaphoid facet is almost vertical to the dorsopalmar axis of the bone and slightly tilts laterally. The furrow for the triquetrum is oblique to the dorsopalmar axis of the bone, and it is distinguished from the semilunate facet by a pronounced crest. The ulnae of both DFN3-350 and DFN3-347 b are aligned to the lateral board of the radius. In DFN3-350 it tilts medially, whereas it is more straight in DFN3-347 b. The caudal border of the olecranon forms a very sharp crest. The olecranon tuberosity is missing from all the specimens. The proximal interosseus space in DFN3-350 is filled with sediment but seems to be fairly large; the same probably applies for the poorly preserved DFN3-347 b. The anconeus process is sharp. Laterally, the surfaces for the insertion of extensor digitorum communis and extensor digitorum lateralis are distinguished as shallow furrows.

Metacarpal: The complete metacarpal DFN3-1 has an RI of 9.92. Its proximal part and articular surfaces are seriously damaged and dorsopalmarly compressed, such that no anatomical details can be observed. The proximal articular surface of the metacarpal DFN3-345 ( Fig. 5G–I View Figure 5 ) is not on the same plane. The articular facet for the hamate and the respective facet for the capitato-trapezoid are distinguished by a laterally tilting ridge, which forms an angle of almost 70° to the mediolateral axis of the metacarpal. The seemingly enlarged synovial fossa is located for the most part medially. The articular facet for the capitato-trapezoid has the shape of half an arch. The angle between the dorsal and medial surfaces of the facet for the capitato-trapezoid is ~100°. The facet for the hamate is quadrant shaped. On the medial epicondyle there is a wide and thick dorsal tuberosity that it becomes narrower distally. That tuberosity consists of the medial ridge of a shallow dorsal trough. The lateral and medial surfaces of the shaft are almost flat and vertical to the dorsopalmar axis, resembling a metatarsal. However, the dorsal surface of the shaft is convex. Palmarly, the medial and lateral ridges are very thick proximally, thinner but very pronounced at the middle of the shaft, and they disappear distally. The central trough is of intermediate to great depth, resembling that of a Late Miocene Bohlinia ( Ríos et al. 2016) . There is no observable pyramidal rise. In the distal metacarpal DFN3-349, the palmar ridges reach the distal epicondyles, and the pyramidal rise is very short and faint, limited to an area very close to the distal epicondyles. The shaft also flattens towards the distal end.

Astragali: The DFN3 astragali ( Fig. 6A–D View Figure 6 ) are rather square for a giraffid (after Ríos 2018). The groove of the trochlea is very deep. The central fossa is triangular and very wide. The collum tali is shallow. The medial bulge at collum tali is smooth and located immediately above the medial groove. The medial side of the head is narrow and oblique. The lateral side of the head is of similar thickness, and it is almost parallel to the proximodistal axis. Ventrally, the articular surface is sub-rectangular, slightly tilting laterally. The proximal triangular fossa is deeper in DFN3- 64 than in DFN3-2. The interarticular groove is barely observed. The medial ridge is prominent. The medial scala seems like a very narrow groove. The distal intracephalic fossa is extremely weak in DFN3-2 ( Fig. 6B View Figure 6 ) and not observable in DFN3-64. Medially, the proximal groove for the tibia is very shallow. In lateral view a proximal and a distal pit are observed, both of which are very deep. These pits are separated by a crest, which is ventrally faint and wide but becomes narrower and thicker dorsally. At the point where that crest touches the lateral ridge there is a pointy bulge. The two bulges of the head are unequal, with the medial being larger than the lateral. They seem to tilt laterally in distal view.

Metatarsal: The metatarsal DFN3-346 ( Fig. 6E–G View Figure 6 ) possibly belongs to a juvenile individual. The proximal articular surface is dorsoplantarly compressed. The medial epicondyle is slightly wider dorsoplantarly than the lateral one. The ventral head of the lateral epicondyle lies more distally than the other heads of the epicondyles. The articular facet for the intermediolateral cuneiform is semicircular. The articular facet for the medial cuneiform is very small, round, and lies medially of the dorsoplantar plane; it almost touches the articular surface for the intermediolateral cuneiform. The articular facet of cubonavicular resembles the articular facet for the intermediolateral cuneiform but is wider mediolaterally, especially towards its plantar side, being more kidney shaped. The articular facets for the cubonavicular and the intermediolateral cuneiform touch each other dorsally and flare outwards plantarly.The synovial fossa is very large, as in the metacarpals. The pygmaios is half-broken but observable. Medially, a faint trace of metatarsal II is observed. The plantar ridges are very close to each other and thicker proximally. They flare medially and laterally, respectively, towards the middle of the shaft. They disappear on the one-quarter distalmost part of the bone. The lateral ridge seems to be more pronounced and projects more plantarly than the medial one. The central trough is very narrow; its depth is intermediate to deep. A pyramidal rise-like structure is observed, although it could be a misinterpretation owing to some taphonomic distortion. There is a very prominent dorsal trough, which is delimited by a lateral and a medial dorsal ridge, the latter being more pronounced than former. The dorsal trough continues throughout the shaft, and it reaches the distal condyles. The distal condyles are similar in size. Their keels are very prominent dorsally, distally, and plantarly.

The rather adult metatarsal DFN3-91 seems to be as slender as the metatarsal from Dafnero-1 (DFN-68). Its exact RI cannot be calculated, because its distal part is missing. The medial epicondyle is larger than the lateral one. The ventral head of the lateral epicondyle is more distally located than the other three heads of the epicondyles, as in the previously described metatarsals. The articular surface for the cubonavicular is very large, possibly larger than the articular surface for the intermediolateral cuneiform. They are both semicircular, rather than being kidney shaped. The articular facet for the medial cuneiform is very small and circular to oval. It almost touches the articular facet for the intermediolateral cuneiform, and it tilts dorsally and laterally. The synovial fossa is dorsoplantarly elongated. The pygmaios is half-broken, but it seems to have been weak. The dorsal trough is prominent, being wide proximally and narrow for the rest of the shaft; the medial dorsal ridge is more prominent. The central trough is of intermediate to large depth. The two plantar ridges are both sharp; the medial ridge seems to be thicker than the lateral one and more developed plantarly. In medial view the medial ridge is curved, whereas the lateral ridge is straight in lateral view. The two ridges seem to disappear towards the point where the bone is broken, indicating that the central trough vanishes towards the distalmost part.

Volax fossil site

Material: A left third upper molar (M3) (VOL-224c; Fig. 7 View Figure 7 A-C); a right distal radius (VOL-77; Fig. 7D, E View Figure 7 ).

Upper molar 3 (M3): The isolated M3 VOL-224c ( Fig. 7A–C View Figure 7 ) is a worn tooth, hence belonging to an old individual. The enamel is finely wrinkled. The distal lobe is ~20% narrower than the mesial one. The mesostyle is very prominent, followed by the parastyle, and the metastyle is fairly weak. The paracone is very strong, whereas the metacone is extremely weak.On the buccal side of the tooth there is a cingulum that creates a small stylid at the height of the metacone. The fossettes appear very narrow, probably because of the advanced wear of the tooth. The praehypocrista is more developed than the neocrista, but it does not protrude in the mesial fossette. The neocrista slightly tilts mesially. The praeprotocrista and praeparacrista fuse together, enclosing the mesial fossette mesially. A very wide entostyle is observed lingually, between the two lobes. It is either broken or very worn, hence its occlusal surface is flat. The hypocone is narrower than the protocone.

Radius: The distal radius VOL -7 ( Fig. 7D, E View Figure 7 ) is crescent shaped in cross-section. Its palmar surface is slightly concave. Its dorsal surface is distinguished by a lateral ridge, which is very strong distally but fades towards the middle of the shaft. Lateral to that ridge there is a very wide, finely rugose surface for the insertion of the extensor carpi obliquus. That surface is distally concave and proximally flat. There is also a medial dorsal ridge, which is observable only in the distal part of the bone. More proximally, the transition from the dorsal to the medial surface is very smooth. The ulna is missing, except for the ulnar styloid process, which is somewhat blunt (although it might be partly broken). The medial radial styloid process is also blunt and not very protruding. In dorsal view, near the distal border of the bone, there are two textured surfaces: one located centrally for the insertion of the extensor carpi radialis and the other located laterally for the insertion of the extensor digitorum. The surface for the extensor carpi radialis is wide and slightly tilts medially. The distal articular facet for the scaphoid is concave dorsally, convex palmarly, and slightly tilts laterally. The articular facet for the semilunate is kidney shaped, medially tilting, concave dorsally, and with a convex surface palmarly. Laterally, the articular facet for the triquetrum is triangular in shape and wide. It greatly tilts medially, and its dorsopalmar diameter is greater than the dorsopalmar diameter. The articular facet for the semilunate lies more dorsally than the other articular facets.

Tsiotra Vryssi fossil site

Material: Complete ossicone (TSR F7-13; Fig. 8 View Figure 8 ); ossicone broken near the base (TSR F7-5); right deciduous upper toothrow, bearing DP4–M2 (TSR E19-17; Fig. 9 View Figure 9 A-C); three isolated upper premolars, P2–P4 (TSR F8-7; Fig. 9 View Figure 9 D-I); left deciduous lower toothrow, bearing d2–m1 (TSR-151; Fig. 10 View Figure 10 A-D); complete left adult lower toothrow (TSR-150; Fig. 10 View Figure 10 E-G); left proximal radius (TSR G18-12); right proximal metacarpal (TSR E19-9); fragment of the pelvis (TSR D18-119; see Supporting Information, Appendix S1); right femur (TSR F8-46; see Supporting Information, Appendix S1, Fig. 11 View Figure 11 ); left distal tibia (TSR E21-14; see Supporting Information, Appendix S1); three right astragali (TSR 2, TSR E8-11, and TSR D22-6); left astragalus (TSR G20-9b; Fig. 12 View Figure 12 A-D) and; Fig. 12 View Figure 12 E-H); and left calcaneus (TSR G20-9a; see Supporting Information, Appendix S1); right calcaneus (TSR F8-34; see Supporting Information, Appendix S1); right metatarsal (TSR H22-6; Fig. 12 View Figure 12 I-K); left proximal phalanx (TSR 4; see Supporting Information, Appendix S1); and distal part of proximal phalanx (TSR D20-4; see Supporting Information, Appendix S1).

Ossicones: The complete ossicone TSR F7-13 ( Fig. 8 View Figure 8 ) has a length of 310 mm ( Table 1 View Table 1 ). Both ossicones are fairly straight and end in a pointed, polished apex. The cross-section of the ossicones is oval and, possibly, becomes more circular near the apex. From the fragmented ossicone, we can conclude that it is pneumatized towards its base. Their surface is slightly rugose over the proximal half; smoother distally. There are no bumps on the surface of the ossicones.

Upper deciduous toothrow: The mesial lobe of the DP4 is damaged buccally ( Fig. 9A–C View Figure 9 ). The mesial fossette is very narrow, and the protocone creates a protoconal fold distally. The protocone is very wide. The distal lobe is slightly larger than the mesial one, and the fossette is U-shaped and narrow. The praehypocrista does not penetrate the mesial fossette. The hypocone is as wide as the protocone. Buccally, the pillars of the mesostyle and the metastyle are stronger than that of the metacone. A moderately developed cingulum is observed buccally on the distal lobe.

The molars are very alike. Buccally, the parastyle, the rib of paracone, and the mesostyle are strong, followed by the metastyle. The rib of the metacone is the weakest. In both molars, the metacone is narrower than the paracone. There is a very weak buccal cingulum. Buccally, at the height of the metastyle, there is also a stylid, similar to that of DFN-28, although it does not tilt mesially.The protocone is wider than the hypocone. Both molars bear a bifid entostyle, with that of M1 being much stronger. The fossettes are very wide, because the teeth are not very worn. In both teeth, the praehypocrista penetrates the mesial fossette. In the M2, a very slight hypoconal fold is observed.

Upper premolars: The three upper premolars, P2, P3, and P4 ( Fig. 9D–I View Figure 9 ), belong to the same individual. They are severely worn. The P2 ( Fig. 9D, E View Figure 9 ) is the smallest tooth, lingually rounded in occlusal view. It is slightly wider than long. The paracone and metacone are very prominent, with the paracone being wider. The parastyle is weak. The fossette is open lingually, immediately distal to the parastyle. There are two hypoconal folds distally, with the distal one being larger.

The P3 ( Fig. 9F, G View Figure 9 ) is much wider than long (length/width ratio = 74%). The paracone is very well developed and slightly tilts mesially. The parastyle and mesostyle are moderately developed and they are both narrow. The fossette is probably open distally. There is a hypoconal fold distally. In occlusal view, the tooth has an almost trapezoid shape, being slightly rounded lingually.

The P4 ( Fig. 9H, I View Figure 9 ) is exceptionally wide (length/width ratio = 78.5%), possibly exaggerated by taphonomic distortion. The paracone is strong and slightly tilts mesially. The parastyle is once again prominent but occlusally blunt. The metacone is probably very weak or damaged. The praeparacrista exists, enclosing the fossette, which is wider than in P3 and tends to a ‘W’ shape. The distal hypoconal fold is strong and parallel to the labiolingual axis. The hypocone is blunt. In occlusal view, the outline of the tooth is almost trapezoid, being less lingually rounded than the P3.

Deciduous mandible: The juvenile mandible TSR-151 ( Fig. 10A–D View Figure 10 ) is in poor preservation status. However, some morphological features can still be distinguished.

The dp2 is simple and primitive. It lacks an anteparacristid, and the transverse cristid is not followed by a praemetacristid or a postmetacristid. Hence, the parasinusid and mesosinusid fuse to form a large sinusid that is open lingually. The metasinusid is also open lingually. The distal part of the tooth, formed by the hypocristulid and the entocristid, is as long as the mesial part. The hypocristulid and entocristid are parallel to each other and vertical to the tooth axis.

The dp3 is semi-molarized, with its mesial lobe being closed lingually by a continuous wall. The transverse cristid is absent. The metaconid stands out lingually. The distal lobe retains a plesiomorphic condition, because the hypocristulid and the entocristid are parallel to each other, and obliquely oriented to the tooth axis. The entocristid is continuous with the protocristid. The hypoconulid is enlarged.

The dp4 is slightly damaged. The mesial lobe has a rhomboid shape and is the smallest. Only the lingual side of the middle lobe is retained. Finally, the distal lobe is triangular in occlusal view. Its fossette is simple, narrow, and open distally and lingually. The lingual conids and stylids are all extremely weak.

The m1 has narrow and simple fossettes. The mesial fossette is open mesially and distally. The distal fossette is also open mesially and, probably, distally. The two lobes are rounded in occlusal view. There are a lingual cingulid and a buccal cingulid, mesially in the mesial lobe. The tooth probably bears a very weak ectoconid. As in dp4, the lingual conids and stylids are weak. However, the metaconid is somewhat prominent. The entoconid is damaged.

Adult mandible: The lower toothrow TSR-150 ( Fig. 10E–G View Figure 10 ) has a very low premolar to molar length ratio (p/m row index of 55.9%), owing to the extreme lengthening of the molar row. The p2 is very small compared with the other premolars and similar to the dp2 of TSR-151. The paracristid is probably absent; thus, the mesosinusid fuses with the parasinusid to form an enlarged, lingually open valley between the anteparacristid and the transverse cristid. The transverse cristid does not continue to a premetacristid or a postmetacristid. More distally, the hypocristulid and entocristid converge and fuse at the base of the tooth; hence, the telosinusid is closed lingually. The distal part of the tooth, consisting of the hypocristulid and the entocristid, is as long as the mesial one. At the base of the tooth and at the height of the hypoconid there is a sharp bulge, created by the buccal cingulid.

The p3 is moderately molarized, identical to the dp3 of TSR-151. There is no transverse cristid. The paracristid fuses with the premetacristid and postmetacristid, forming a mesial lingual wall. At the height of the paraconid, the lingual wall is abruptly descending and then ascending to reach the metaconid, creating a ‘V’-shaped valley in lingual view. The distal lobe retains the plesiomorphic condition, in which the hypocristulid and the entocristid are parallel to each other and oblique to the tooth longitudinal axis.The entocristid is continuous with the protocristid. The hypocristid is narrow and parallel to the mesiodistal axis of the tooth and continues to the wider hypocristulid. At the base of the tooth and at the height of the hypoconid there is a bulge created by the buccal cingulid.

The p4 is morphologically similar with the p3, although slightly larger. The main distinction is that, buccally, between the two lobes there is a trace of an ectostylid. Also, the bulge at the base of the hypoconid is more subtle than the respective bulge in p3. Finally, the ‘V’-shaped valley of the lingual wall is absent in p4.

In m1, the mesial lobe is somewhat wider than the distal one. The lingual conids and stylids are all faint. Only the entoconid and the metaconid are weakly developed. The metastylid and entoconulid are almost unobservable. The fossettes are simple and narrow. The postprotocristid extends distally in the distal fossette. In occlusal view, the two lobes are rounded buccally. A strong ectostylid is observed between the two lobes. There are a weak buccal cingulid and a lingual cingulid mesially on the mesial lobe.

The m2 is damaged at certain points. It resembles the m1 but differs in that in occlusal view, the mesial lobe is buccally pointed (the distal lobe is damaged buccally; however, its shape is observable). The postprotocristid is more developed than on m1, reaching the metastylid. Between the two lobes there is a very weak ectostylid. There is a faint buccal cingulid mesially, and although the tooth is very damaged lingually, there is probably a lingual cingulid mesially. The entoconid is weak, but it seems to be more developed than it is in m1.

The m3 is badly preserved but resembles the two mesial molars. The mesial lobe is mostly rounded in occlusal view, closer in shape to that of m1. It is much longer than the mesial lobe of m1. The distal lobe is damaged. The postprotocristid reaches the metastylid. The third lobe is also somewhat damaged but seems to be rounded and parallel to the two mesial lobes. The exact position of the hypoconulid is not clear. Between the two mesial lobes there is a basal bulge, probably corresponding to the ectostylid of the m1 and m2. Mesially, on the mesial lobe there are faint lingual and buccal cingulids, similar to those of m1 and m2.

Radius: The shaft of the proximal radius (TSR G18-12) is dorsoventrally compressed. The coronoid process is very sharp. There is a blunt bump that lies medial to the coronoid process. The lateral relief of insertion is oriented laterally, and not cranially as it does in bovids. Between the coronoid process and the lateral relief of insertion there is another process, more prominent than the coronoid. Proximally, the concave lateral articular facet runs through a smooth ridge. The medial articular facet is about half the length of the lateral one, and it is less broad. The medial tuberosity is rather weak, although that might be attributable to the poor preservation status of the specimen.

Metacarpal: The incomplete metacarpal (TSR E19-9) exhibits poor preservation. The articular facet for the capitato-trapezoid is slightly wider than the lateral one and seems to be completely arched. The articular facet for the hamate is quadrant shaped and on the same plane as the articular facet for the capitato-trapezoid; they are distinguished by a laterally tilting crest, which forms an angle of ~60° with the mediolateral axis of the metacarpal. The synovial fossa is not observable owing to poor preservation. The dorsomedial tuberosity is well developed. A dorsal shallow trough starts very proximally; it quickly weakens more distally. The central trough is barely noticeable, owing to the preservation status of the specimen; its depth is intermediate.

Astragalus: The astragali ( Fig. 12A–D View Figure 12 ) are rather symmetrical and square in comparison to other giraffids ( Ríos 2018). The lateral and medial ridges of the trochlea are very well developed, resulting to a deep intratrochlear notch in proximal view. The groove of the trochlea is also very deep. The central fossa of all three specimens is triangular, very wide, and not very deep. The medial bulge of the collum tali is very pronounced and sharp in TSR G20-96 b, extending strongly medially; it is damaged in TSR D22-6 . Laterally, the collum tali is very broad in TSR D22-96 and slightly less developed in TSR G20-96 b. In lateral view, near the proximal end of the articular facet for the calcaneus there is a prominent crest that ends in a sharp apex. That crest divides the articular facet for the calcaneus into two pits, with the distal one being much larger. Immediately proximal to the proximal fossa for the calcaneus lies the articular facet for the malleolus, which is not as deep as the respective facet for the calcaneus. The lateral notch of both specimens is deep. Ventrally, the medial and lateral ridges are very sharp. The medial ridge tilts towards the intratrochlear notch. The medial scala is very shallow. The distal intracephalic fossa is observed only in TSR G20-96 ; it is extremely small and shallow. Distally, the lateral part of the head extends more distally than the medial one. The distal heads are unequal, with the medial one being the largest. From a distal view, they both slightly tilt laterally.

Metatarsal: The metatarsal TSR H22-6 ( Fig. 12I–K View Figure 12 ) has an RI of 9.58. The proximal end is divided into two epicondyles, with the medial one being significantly larger. The ventral head of the lateral epicondyle is more distally located than the rest of the epicondyle heads. The articular facet for the cubonavicular is kidney shaped; from a proximal view, its plantar side has a straight outline. The facet for the intermediolateral cuneiform is semicircular and probably somewhat smaller than the facet for the cubonavicular, extending slightly more plantarly; it ends in a pointed tip dorsally and laterally. Those two articular facets contact dorsally. A narrow and shallow groove separates the facet for the intermediolateral cuneiform from the one for the medial cuneiform; hence, the two medial articular facets are almost in contact. The shape of the articular facet for the medial cuneiform is elliptical. The synovial fossa is very large; as a result, the facets for the cubonavicular and the intermediolateral cuneiform flare laterally and medially, respectively. The pygmaios is of intermediate development and blunt. Beneath the pygmaios there is a sizeable proximal canal, where the central trough starts. The depth of the central trough is intermediate. The medial ridge is thicker than the lateral one and straight in medial view. The lateral ridge is curved in lateral view. The ridges and the trough disappear in the distalmost one-quarter of the shaft, and then a smooth pyramidal rise begins. At that point, the dorsoplantar diameter of the shaft is narrow. The lateral and medial surfaces of the metatarsal are slightly curved, whereas the dorsal surface is rather flattened to slightly convex. The dorsal surface is run through by a prominent dorsal trough that begins almost at the proximal end of the shaft and reaches the wide trochlear incision. The keels of the distal condyles are strong dorsally, distally, and plantarly.

Krimni-3 fossil site

Material: Complete right forelimb ( Fig. 13 View Figure 13 ) including the radius (KMN 6a), articulated carpals (KMN 6b; see Supporting Information, Appendix S1), the metacarpal (KMN 6c), the medial and lateral proximal phalanges (KMN 6d; see Supporting Information, Appendix S1), and the sesamoids (proximal exterior, proximal interdigital, and distal: KMN 6e; see Supporting Information, Appendix S1).

Radius: The radius KMN 6a ( Fig. 13A–D View Figure 13 ) is fairly robust, with an RI of 11.4. The cross-section of the shaft is semicircular. The borders of the palmar with the lateral and medial sides are smooth in the proximal half of the shaft. In the distal half they are distinguished by sharper ridges. Proximally, the coronoid process is strong and almost on the same plane as the dorsolateral border of the square-shaped lateral articular facet for the humerus. The medial articular facet for the humerus is rectangular. The lateral relief for the insertion of the extensor digitorum lateralis is textured and not very protruding. The textured area, on which the same muscle is inserted palmarly, is also not protruding. In lateral view, that surface is triangular. The radial tuberosity is rectangular in dorsal view; it extends over a large part of the dorsoproximal surface, and it is very textured. Distally, the ulnar styloid process is broken. The radial styloid process is weakly protruding, although it is the most distally marked feature of the bone. The insertion for the extensor carpi radialis is almost aligned with the shaft axis. On the distal articular surface, the scaphoid facet is concave dorsally and convex palmarly, while it slightly tilts laterally. The semilunate facet forms an elongated pit, almost parallel to the mediolateral axis, which narrows dorsally; it is followed by a laterally tilting convex surface on the palmar side. Finally, there is a slightly convex facet for the triquetrum. The articular facet for the semilunate lies more dorsally than the other two articular facets.

Metacarpal: The metacarpal KMN 6c ( Fig. 13K–M View Figure 13 ) is long (420 mm), with an RI of 9.56. The lateral and medial epicondyles are almost at the same level. In proximal view, the palmar border of the proximal articular surface forms a straight line. The articular facet for the hamate is rectangular. The outline of the medial aspect of the articular facet for the capitato-trapezoid is only slightly curved dorsally. The medial side of the articular facet for the capitato-trapezoid is straight, and it slightly tilts laterally. The transition from the medial to the dorsal side of the capitato-trapezoid facet forms an angle of almost 100°. The two proximal articular facets are separated by a step, which is inclined laterally at an angle of 70° to the mediolateral axis of the bone. Between the two facets there is a large synovial fossa that communicates with the proximal metacarpal canal. The dorsomedial tuberosity is enlarged and textured. There is a shallow dorsal trough, which starts at the height of the proximal metacarpal canal and quickly fades; it is no longer observable at the middle of the shaft. Palmarly, the central trough begins at the height of the metacarpal canal. It is very deep proximally, accompanied by strong palmar ridges, with the medial ridge being the thicker. It becomes shallower more distally, and the ridges also become weaker. Finally, at the distal one-quarter of the shaft, the central trough disappears completely, and an extremely weak pyramidal rise is observed. The distal condyles are fairly enlarged. The keels are well developed throughout the perimeter of the distal condyles. The trochlear incision is very narrow.

Comparison among Plio-Pleistocene giraffids

Ossicones of Plio-Pleistocene giraffids in Greece have been found in Volax ( Sickenberg 1967), Sésklo ( Athanassiou 2014), Dafnero-3, and Tsiotra Vryssi. Ossicones are also known from Kuruksay, Tajikistan ( Sharapov 1974); Dmanisi, Georgia ( Vekua et al. 2008); Liventsovka, Russia ( Alexejeva and Motuzko 1985); and Denizli, Turkey ( Boulbes et al. 2014). The ossicones from Tsiotra Vryssi (TSR F7-5 and TSR F7-13) are morphologically identical to those on the cranium from Sésklo and the Volax frontlet (holotype of ‘ Macedonitherium martinii ’). The complete specimen TSR F7-13 is the longest known for Plio-Pleistocene giraffids ( Table 1 View Table 1 ). The ossicones from Sésklo (AMPG SE3-31) are slightly shorter, and the Volax ossicones (AMPG 980) are much shorter. The basal minimal diameter of the Tsiotra Vryssi ossicone is very close to the measurements for Sésklo and Volax ossicones ( Table 1 View Table 1 ). The straight to slightly curved left (better preserved) ossicone from the Dafnero-3 cranium with a smooth to slightly rugose surface is also very similar to those from Sesklo, Tsiotra Vryssi, and Volax, and it has a similar basal minimal diameter ( Table 1 View Table 1 ). The orientation of the DFN3-351 ossicones is very similar to that from the cranium of Sésklo (AMPG SE3- 31) and to that from the holotype cranium of ‘ Sogdianotherium kuruksaense ’ from Kuruksay, Tajikistan ( Sharapov 1974); in all specimens, the ossicones are rostrally inclined at an angle of 105°–110° to the parietal roof. In the Volax specimen, the orientation is slightly different, in that the ossicones seem stand more vertically to the parietal roof. The Kuruksay ossicone is shorter than that of Sésklo, but of similar basal diameter. Two ossicones from Dmanisi ( Vekua et al. 2008) are dimensionally very close to those of the Volax frontlet and significantly shorter than the ossicones of SE3-31 and TSR F7-13. The Dmanisi ossicones have a basal transverse diameter very similar to the ossicones from Volax, Sésklo, Dafnero-3, Tsiotra Vryssi, and Kuruksay ( Table 1 View Table 1 ). An incomplete ossicone from Liventsovka, Azov Sea ( Alexejeva and Motuzko 1985: fig. 1; Bajgusheva and Titov 2002: fig. 2) is morphologically very similar to the ossicones described from the Greek sites. It might be shorter, but its basal diameter is slightly larger than those from the Greek localities, Kuruksay, and Dmanisi. Finally, an incomplete ossicone known from Denizli, Turkey, referred to Palaeotragus sp. ( Boulbes et al. 2014), is morphologically similar to all the other known specimens, in having a rather smooth surface, a pointy apex, and a slight caudal curvature.

Apart from the newly described DFN3-351, more or less fragmented cranial giraffid specimens are known from Sésklo and Volax in Greece ( Sickenberg 1967, Athanassiou 2014) and from Kuruksay and Liventsovka ( Sharapov 1974, Bajgusheva and Titov 2002, Titov 2008). Although greatly damaged, the partial cranium DFN3-351 resembles that from Sésklo described by Athanassiou (2014) in: the degree of frontal pneumatization; the concave frontals between the ossicones; the supraorbital position of the ossicones; the long and inclined postorbital part of the cranium in lateral view; the position, elongation, and sharpness of the temporal lines; the fairly slender zygomatic process of the temporal; the concave and wide temporal fossa (the maximum width is ~ 58 mm in DFN3-351 and 73 mm in the Sésklo cranium); the sharp descending part of the nuchal crest (although in DFN3-351 it is strongly damaged); the frontoparietal fusion on the dorsal surface of the cranium; and the more ventrally protruding jugular process compared with the occipital condyles. The rugose inflation of the frontals, which is observed caudal to the ossicones in DFN3-351, is rather uncertain in the Sésklo cranium. The Volax frontlet does not allow extended cranial comparisons. However, the concavity of the frontals between the ossicones is clear, and the frontals show rugose inflations immediately caudal to the ossicones, as in DFN3- 351. On all three Greek cranial specimens the ossicones have a minimal basal distance of ~ 55 mm, and the bases of the ossicones are ~ 85–90 mm apart. As far we can judge from the poor illustrations ( Sharapov 1974, Athanassiou 2014), the Kuruksay cranium is very similar morphologically to the Greek cranial sample. Although Sharapov (1974) does not provide a detailed description of the Kuruksay cranium, it demonstrates strong parietal crests, as is the case for SE3-31 and DFN3-351. The Liventsovka partial cranium ( Titov 2008) is very similar to DFN3-351 and SE 3-31 in having a broad and flat cranial roof and very strong parietal crests. The flattened occipital region is another similarity between the Liventsovka cranium and SE3- 31; that feature is uncertain in DFN3-351 owing to the articulated atlas and the general preservation status of the specimen.

The only known upper dentitions of Greek Plio-Pleistocene giraffids come from Dafnero-1 (DFN-28) and Sésklo (cranium SE3-31), plus the isolated premolars (TSR F8-7) and upper deciduous toothrow (TSR E19-17) from Tsiotra Vryssi and the upper third molar from Volax (VOL-224c) described herein. There are also known upper giraffid teeth from: Fântâna lui Mitilan, Romania; Kuruksay, Tajikistan; and Huélago, Spain ( Samson and Radulesco 1966, Sharapov 1974, Van der Made and Morales 2011). The P2s of SE3-31 and TSR F8-7 are very similar in the circular occlusal outline and the fossette being open linguodistally.In both specimens the P3 is profoundly larger than the P2 (~ 6–8 mm wider). The P3s of SE3-31 and TSR F8-7 are similar in the centrally located paracone. They differ in that the hypocone area is narrower in TSR F8-7 and in that the fossette is open linguomesially and linguodistally in SE3-31. However, the TSR F8-7 P3 is extremely worn, which it might affect the morphology of the fossette. Finally, the P4s of DFN 28 and SE3- 31 are very alike, being sub-rectangular to trapezoidal in occlusal view. The TSR F8-7 P4 is taphonomically distorted, such that its original shape is rather uncertain. The Huélago P4s ( Van der Made and Morales 2011) and the P4 of the Kuruksay cranium also have a trapezoidal occlusal outline. All TSR adult premolars plus the DP4 of TSR E19-17 possess a distal hypoconal fold. That fold is also present in DFN-28 P4 and in the P3 from the SE3-31 cranium. The hypoconal fold is also very prominent in the P4 from the Kuruksay cranium. In the few P4s from Huélago, the hypoconal fold is absent. The upper molars from Dafnero-1, Volax, and Sésklo are very similar to each other in having a very prominent mesostyle and paracone, a less prominent parastyle and metacone, and a very weak metastyle. The TSR M1 and M2 differ in that the parastyles and metastyles are better developed and that the lobes of all the molars are rounded in occlusal view. Finally, the praehypocrista protrudes in the mesial fossette in all the molars. The M3 VOL-224c and the molars from Sésklo cranium and TSR E19-17 bear entostyles and a weak buccal cingulum. On the M3 VOL-224c the buccal cingulum creates a prominent sharp bulge in the height of the metacone. In the M3 of DFN-28 and the M2 of TSR E19-17 there is also a prominent buccal cingulum; it creates a pillar-like structure that reaches more than half of the crown of the tooth in the former specimen, but it is much shorter in the latter specimen. The distal lobe in all Sésklo, Volax, and Dafnero M3s is ~20%–25% narrower than the mesial lobe. The molars from Huélago also have rounded lobes in occlusal view and a mesially protruding praehypocrista. The distal lobe of the Huélago M3 is ~20% narrower than the mesial lobe, like those from the Greek record. The lobes of the molars from the Kuruksay cranium are also rounded in occlusal view. The mesially protruding praehypocrista is not observed in the M1 and M2 of the provided sketch ( Sharapov 1974: fig. 2b), possibly owing to the seemingly advanced wear stage of these teeth. However, it is present on the M3, which is less worn. The distal lobe of the Kuruksay M3 is also ~20% narrower than the mesial one.

The lower dentitions of Plio-Pleistocene giraffids are scarce. Apart from the TSR specimens described herein, no other adult lower dentition is known from Greece. Some more specimens are known from Liventsovka and Huélago ( Godina and Bajgusheva 1985, Titov 2008, Van der Made and Morales 2011). Juvenile lower dentitions are known from Sésklo (AMPG Σ 184) and from Karnezeika ( Sianis 2023: KZ 815). The most exceptional feature of the p2 of TSR-150 is the absence of a paracristid. The distal part of the tooth, consisting of the hypocristulid and entocristid, is also very elongated. The same applies to the dp2 of TSR-151, but not to the dp2 of AMPG Σ 184 from Sesklo, in which the distal part is reduced and there is a paracristid. In TSR-150, the p3 and p4 are morphologically similar to each other and differ only in size. There are only a p3 or p4 from Huélago and a p4 from Liventsovka for comparison. They are similar to those of TSR-150: a buccal groove separates the teeth in two lobes in all specimens; the mesial lobe is similar to the lobe of a molar with a trigonid lingual wall; the distal lobe retains a more plesiomorphic condition, in which the entocristid and hypocristulid are parallel to each other and oblique to the mesiodistal axis of the tooth; the entocristid is continuous with the protocristid; and the hypocristulid is connected to the protocristid through a very week hypocristid. The dp3s of TSR-151, AMPG Σ 184 from Sésklo, and KZ 815 from Karnezeika ( Sianis 2023) are morphologically similar to the adult p3s and p4s, suggesting that the semi-molarized lower premolars were a rather fixed feature in both adult and juvenile individuals. The dp4 of TSR-151 is damaged, meaning that it cannot be compared in detail with the respective dp4 of AMPG Σ 184 and KZ 815. However, the mesial lobe in all specimens is rhomboid.

The m 1 in TSR-150, TSR-151 from Tsiotra Vryssi, and AMPG Σ 184 from Sésklo bear fossettes that are narrow and V-shaped. In TSR-150 and in AMPG Σ 184, the praentocristid penetrates the mesial fossette. That penetration is very subtle in TSR-150 and very pronounced in AMPG Σ 184. In TSR-151, the praentocristid does not penetrate the mesial fossette. In TSR-150, TSR-151, and AMPG Σ 14 the mesial fossette is mesially open and the distal fossette is distally open. In the sketch from the Azov region provided by Titov (2008: fig. 67) the fossettes of m1 seem to be closed. The lobes of all known m1s are rounded in occlusal view. The presence of the ectostylid in m1 varies; it is present in TSR-150 and in the specimen described by Titov (2008) from Liventovska but absent in TSR-151 and AMPG Σ 184 from Sesklo. The m2 sketch provided by Titov (2008) seems to have a more rounded mesial lobe than the respective lobe from TSR-150. The m3s of TSR-150 and the specimen from Liventsovka are similar in the orientation of the distal lobe, which is parallel to the mesiodistal axis of the tooth. The lower molars and premolars from all the aforementioned localities are very close biometrically.

Distal humeri of Plio-Pleistocene giraffids are known only from Sésklo, Dafnero-3 (see Supporting Information, Appendix S1), Vatera ( de Vos et al. 2002), and Liventsovka (holotype of ‘ Palaeotragus priasovicus ’) ( Godina and Bajgusheva 1985, Titov 2008). In dorsal view, the epicondylar crest of the humerus ends in a very thick blunt bulge and does not protrude more laterally than the trochlea in the Sésklo specimen. In DFN3-346a, the morphology of this structure is not accessible, although it is probably similar to that of Sésklo. In the Liventsovka humerus ( Godina and Bajgusheva 1985, Titov 2008), the epicondylar crest also ends in a thick blunt bulge that does not protrude more laterally than the trochlea.

Complete radii of Plio-Pleistocene giraffids are known from Dafnero-1, Dafnero-3, Krimni-3, and Libakos ( Steensma 1988) from Greece and from Valea Grăunceanului, Romania ( Bolomey 1965, Samson and Radulesco 1966). Proximal radii are known from Tsiotra Vryssi, Sésklo, and Vatera ( de Vos et al. 2002). Distal radii are known from Volax, Dafnero-3, and Libakos ( Steensma 1988). All the Greek specimens share several common features: proximally, the lateral articular facet for the humerus is rectangular to semicircular; the shape of the medial articular surface for the humerus is sub-rectangular or square; the medial tuberosity, in all cases, is sub-rectangular to square, rugose, and barely reaching to the medial side of the shaft; in distal view, the articular facet for the semilunate is, in all specimens, more dorsally located than the articular facet for the scaphoid; the length of the surface for the extensor carpi radialis represents ~31%–35% of the maximum distal breadth of the radius. There is some variation concerning the size of the radii. Those from Dafnero-1 and Dafnero-3 have a length ranging between 515 and 530 mm, whereas that of Krimni-3 has a length of only 471 mm. The Libakos complete radius is also as short, with a length of 453 mm ( Steensma 1988). The robustness of the radii also varies. The Dafnero-1 and Dafnero-3 fossil record shows that there was variation of the dimensions of the radii in the same population, in terms of robustness, with a range between 9.7 and 12.1. The KMN 6a and Libakos radii fall within this range, with an RI of 11.4 and 11.9, respectively. The incomplete radius AMPG Σ 670 from Sésklo is the broadest proximally. The radius of Valea Grăunceanului ( Bolomey 1965: fig. 5a) is proximally similar to the Greek ones, also having a medial tuberosity, although its total length is intermediate between the longer Dafnero and the shorter Krimni-3 and Libakos radii. The RI of the Valea Grăunceanului radius is 10.2, within the range of the Greek sample.

Apart from the metacarpals described here from Dafnero-3, Tsiotra Vryssi, and Krimni-3, other metacarpals are known from: Volax; Fonelas, Spain ( Garrido and Arribas 2008); Gülyazı, Turkey ( Mayda et al. 2016); and Valea Grăunceanului, Romania ( Bolomey 1965, Croitor et al. 2024). There is also a distal metacarpal from Libakos and a proximal metacarpal from Haliakmon Q-Profil, Greece ( Steensma 1988) and Dmanisi, Georgia ( Vekua et al. 2008). Although some of the metacarpals studied here are poorly preserved, some key points can be extracted. The total length of complete specimens is limited to a narrow range of 410–421 mm (N = 3). The robusticity indices are also very similar (9.56 for KMN-6c and 9.75 from DFN3-1). The AMPG (V)-981 metacarpal from Volax is as long as DFN3-1 and with a similar robusticity index (9.78) ( Fig. 14A View Figure 14 ). Morphologically, there are also consistent similarities. Firstly, in all the available metacarpals there is a strong dorsomedial tuberosity that continues to a dorsal medial ridge and becomes thinner more distally in the shaft; it forms the medial border of a short and shallow dorsal trough. Secondly, the medial border of the articular facet for the capitato-trapezoid is straight and tilts laterally. Thirdly, the transition from the medial side to the dorsal side of the articular facet for the capitato-trapezoid is a slightly pointed obtuse angle. Finally, the dorsal side of the capitato-trapezoid articular facet is arched in all the specimens. In these features the metacarpals studied herein agree with those from Volax AMPG (V)-981. Only the specimen KNM-6c demonstrates the synovial fossa in detail, because the other specimens are more or less damaged; it is palmarly enclosed.DFN3-345 and TSR E19-9 should also have had a closed synovial fossa. The articular facet for the hamate is consistently rectangular, and in proximal view its palmar outline is flat; the hamate articular facet of AMPG (V)-981 from Volax differs, probably owing to taphonomic deformation. The hamate articular facet of the Libakos proximal metacarpal is also rectangular in proximal view ( Steensma 1988: pl. X, fig. 4). The palmar medial ridge is thicker than the lateral one in all studied specimens. In Dafnero-3 metacarpals, the medial ridge is curved in medial view, whereas the lateral one is straight. In this feature they differ from KMN-6c, in which both ridges are straight. A general distinction is that the shaft is slightly less broad in the DFN3 metacarpals than it is in the specimens from KMN and TSR. The complete metacarpals from Valea Grăunceanului and Fonelas are slightly shorter than the metacarpals of the Greek record but of similar robusticity indices. The lateral proximal articular surface of the Valea Grăunceanului metacarpals have a shape of one-quarter of a circle, and the medial articular facet has the shape of half an arch, as in the Greek metacarpals. In the Valea Grăunceanului sample, the synovial fossa varies. In some specimens ( Bolomey 1965) it is palmarly closed, whereas in others (VGr.0972, VGr.0973, and, possibly, VGr.0401) it is open. The only known metacarpal from Fonelas belonged to a juvenile individual ( Garrido and Arribas 2008). However, the general morphology of the proximal articular surface agrees with that of the Greek and Romanian samples. There is also a well-developed dorsomedial tuberosity, similar to the Greek specimens. The dimensions of the incomplete metacarpals from Libakos, Haliakmon Q-Profil, and Dmanisi fit within the size range of the remaining Eurasian Plio-Pleistocene giraffids. Although never described in detail, a metacarpal from Gülyazı, Turkey appears significantly longer than those of the Greek sites (446 mm), but its RI and general morphology are similar to the Greek sample (S. Mayda, personal communication with D.S. Kostopoulos, 2023).

Unfortunately, the femurs are very scarce in the giraffid fossil record; therefore, a detailed comparison is difficult. The TSR F8-46 femur is the first known for the Eurasian Plio-Pleistocene. It is shorter than the femurs of taxa such as Deccenatherium Crusafont, 1952 ( Ríos et al. 2017), Samotherium Forsyth-Major, 1888 ( Bohlin 1926, Kostopoulos 2009), Helladotherium Gaudry, 1860 , Giraffa Brisson, 1762 ( Geraads 1974), and Birgerbohlinia Crusafont and Villalta, 1951 ( Crusafont 1952) and longer than the femurs of the small-sized P. rouenii and Palaeotragus microdon (Koken, 1885) ( Bohlin 1926, Geraads 1974). The robusticity of the TSR femur is comparable to that of Deccenatherium reo Ríos et al., 2017 ( Ríos et al. 2017). It is more slender than the femurs of Samotherium and Helladotherium and slightly more robust than those of P. rouenii .

Tibiae are known from Dafnero-1, Dafnero-3, Tsiotra Vryssi, and Sésklo ( Kostopoulos and Athanassiou 2005; AMPG Σ 50) from Greece and from Valea Grăunceanului, Romania ( Croitor et al. 2024). The tibia DFN-150 is much smaller than TSR E21- 14 and AMPG Σ 50; DFN3-355 is by far the largest specimen. The cochleae have the same morphology in the DFN, DFN3, TSR, and Sésklo specimens, with the medial surface being longer and narrower than the lateral by ~20%. In DFN-150, the caudal border of the medial pit of the cochlea is located completely medially. In DFN3-355 and TSR E21-14 it is more laterally located; from a caudal view, the malleolus medialis is observable. The same is true for AMPG Σ 50. In DFN3-355, medial and proximal to the malleolus medialis there is a bony tuberosity, which is much more pronounced than it is in the DFN and TSR specimens. The lateral border of the tibia and the caudal ridges have the same morphology in DFN and DFN 3 specimens; their morphology is uncertain in TSR E21-14 and AMPG Σ 50. In DFN3-355, the point where the two caudal ridges converge is also uncertain. A difference between the Dafnero tibiae is that in DFN3-356, the bifurcation of the lateral border occurs more distally than in DFN-150, although it is a much longer bone. Once again, that feature is uncertain in TSR E21-14 and AMPG Σ 50. The Valea Grăunceanului distal tibia (VGr.0961) is dimensionally similar to those from Dafnero-1 and Sésklo. The caudal ridges of that specimen conform to the morphology of the DFN and DFN3 tibiae. The cochlea is similar to the Greek specimens in having a narrower and longer medial surface.

Astragali are known from Dafnero-3, Sésklo ( Kostopoulos and Athanassiou 2005), and Vatera ( de Vos et al. 2002) from Greece and from Valea Grăunceanului, Romania ( Bolomey 1965, Croitor et al. 2024). The Dafnero-3, Tsiotra Vryssi, and Sésklo astragali are enlarged, with the exception of DFN3-64. Morphologically, the astragali are very similar to each other. The intratrochlear notch is U-shaped in all the specimens. The medial bulge of the collum tali is very prominent and sharp, and it protrudes dorsally. That protrusion is much greater in TSR and Sésklo specimens than in DFN 3 specimens. The lateral surface of the astragalus demonstrates very deep concave areas. In ventral view, the medial ridge is tilted. In all specimens, the distal intracephalic fossa does exist, but it is faint, and the medial scala is narrow. The incomplete astragalus from Vatera is dimensionally closer to DFN3-64. The astragali from Valea Grăunceanului, Romania (N = 5) are enlarged, like most astragali from Dafnero-3 and those from Tsiotra Vryssi and Sésklo. Additionally, they resemble the Greek astragali in: the wide and triangular central fossa; the deep intratrochlear notch; the protruding medial bulge at the collum tali; and the tilting medial ridge plantarly. Only index 3 (sensu Ríos 2018) could be estimated for the Valea Grăunceanului astragali and demonstrates a wide range of 89–95.7. Index 3 of the Greek sample demonstrates a narrower range (88.6–93.5) and fits within the range of the Romanian specimens. The high index 3 indicates a rather symmetrical astragalus. Moreover, the relatively high index 1 (sensu Ríos 2018) calculated for the Greek sample (61.5–67.7) indicates a rather square astragalus for the Plio-Pleistocene giraffid.

Metatarsals of Plio-Pleistocene European giraffids are known from Dafnero-1, Dafnero-3, Tsiotra Vryssi, and Sésklo in Greece and from Valea Grăunceanului, Romania ( Bolomey 1965, Croitor et al. 2024); Fonelas, Spain ( Garrido and Arribas 2008); Gülyazı, Turkey ( Mayda et al. 2016); and Dmanisi, Georgia ( Vekua et al. 2008). All studied Greek metatarsals have a deep dorsal trough that reaches the trochlea. In all specimens, the ventral head of the lateral epicondyle is more distally located than the dorsal head. As a result, the whole proximal lateral epiphysis of the metatarsal is occupied by the articular facet for the cubonavicular. In all available specimens, the articular facets for the cubonavicular and the intermediolateral cuneiform contact dorsally. The articular surface for the cubonavicular has roughly the same size as that for the intermediolateral cuneiform. The shape of the facet for the intermediolateral cuneiform is consistently semicircular. The shape of the articular facet for the cubonavicular shows some variation; in the specimens DFN3- 91 and AMPG Σ 58 from Sesklo it is semicircular, whereas in the other specimens from DFN, DFN3, and TSR is mostly kidney shaped. The cubonavicular articular facet is also aligned with the dorsopalmar axis of the epiphysis. Medially and in between the two medial articular facets there is a metatarsal II rudiment located slightly more distally than the proximal epiphysis. The lateral ridge is significantly curved, but less so in TSR H22-6 than in specimens from Dafnero-1, Dafnero-3, and Sésklo. The pyramidal rise is prominent in all specimens, as a smooth ridge. The overall dimensions of the metatarsals of the Greek sample are similar; however, TSR H22-6 is somewhat more robust than the other specimens ( Fig. 14B View Figure 14 ). DFN3-346 is significantly shorter than the other specimens, although it probably belonged to an immature individual. The Dmanisi and Valea Grăunceanului metatarsals are very close dimensionally to the Greek sample. The Fonelas P1 metatarsal is exceptionally long and slender, hence distinct from the other specimens ( Fig. 14B View Figure 14 ). The morphology of the proximal epiphyses of the metatarsals from Valea Grăunceanului, Fonelas P1, and Dmanisi is very similar to the morphology of the Greek sample in: the semicircular-shaped facet for the intermediolateral cuneiform; the kidney-shaped articular facet for the cubonavicular; and the more distally located ventral head of the lateral epicondyle than the dorsal head. The proximal articular facets of the Fonelas P1 metatarsal seem to be slightly more remote from each other than is the case for all the other studied metatarsals. However, the Fonelas P1 metatarsal bears a well-developed dorsal trough, similar to that of the Greek sample. A metatarsal from Gülyazı, Turkey appears identical to the Greek/Balkan sample in both morphology and proportions (S. Mayda, personal communication with D.S. Kostopoulos, 2023)

Comparisons with Late Miocene palaeotragines

There are several morphological features that unify the Eurasian Plio-Pleistocene giraffid ensemble with the Late Miocene palaeotragines (including Palaeotragus and Samotherium ; ( Hamilton 1978; Geraads 1986). The most important are: the overall cranial and dental biometrics; the size and proportions of the metapodials; and the morphology of the ossicones, crania, and lower premolars.

The ossicones of the Plio-Pleistocene giraffids, which are rather slender, slightly curved, smooth, and ending in a pointy apex, are morphologically closer to the Late Miocene Palaeotragus and Samotherium species than to the Giraffinae or Sivatheriinae taxa (see Geraads 1991, Ríos et al. 2017). The supraorbital and lateral location of the ossicones, as evidenced by the Volax, Sesko, Dafnero, and Kuruksay record, is also shared with the Late Miocene Palaeotraginae. However , in the Late Miocene Palaeotragus and Samotherium , the ossicones are located above the caudal margin of the orbits, whereas in the Plio-Pleistocene specimens they seem to be more rostrally located, right above the orbits. The Plio-Pleistocene ossicones are also longer and more robust than in the Late Miocene species of Palaeotragus ( Godina 1979, Athanassiou 2014, Ríos et al. 2017), and they are also tilting rostrally, whereas in the Late Miocene Palaeotragus they tilt caudally (except for one P. microdon specimen; Bohlin 1926). The ossicones of Samotherium are longer than the ossicones of the Plio-Pleistocene giraffids and with greater basal diameters ( Kostopoulos 2009; Ríos et al., 2017; Supporting Information, Appendix S1).

Some similarities among the crania of the Plio-Pleistocene and Late Miocene Palaeotragus and Samotherium are the elongated postorbital part and the concave frontals between the ossicones. In all these taxa, the postorbital part has a trapezoid shape, being narrower towards the caudal border of the cranium. The temporal lines are much stronger in the Plio-Pleistocene specimens than they are in the Late Miocene Palaeotragus and Samotherium . The same is probably true for the nuchal crest, which is well observed only on the SE3-31 cranium from Sésklo.

The dimensions of the upper cheek teeth of the Plio-Pleistocene sample perfectly match those of large-sized Palaeotragus , such as P. coelophrys and P. berislavicus . The dimensions of the lower premolars are also indistinguishable from the Late Miocene large-sized Palaeotragus ( Laskos and Kostopoulos 2022) . The lower molars of the Plio-Pleistocene specimens exceed the size of those of large-sized Palaeotragus . This results in a very low premolar to molar length ratio (p/m: 55%) in comparison to the p/m range of the Late Miocene P. rouenii , P. coelophrys , and P. berislavicus (64%–72%). In this feature, the Plio-Pleistocene taxon (but judging only from the TSR sample) is closer to P. microdon (57%–62%), although P. microdon teeth are smaller ( Fig. 15 View Figure 15 ). The overall dimensions of the teeth of the Plio-Pleistocene taxon are much smaller than those of the Late Miocene Samotherium . Samotherium boissieri Forsyth-Major, 1888 and Samotherium neumayri (Rodler and Weithofer, 1890) also demonstrate low p/m ratios (57%–60% and 58%–67%, respectively; Merceron et al. 2018), but still higher than TSR-150.

The p2 from TSR resembles the p2s of P. rouenii and P. coelophrys , which have the same primitive configuration. The p3 is a tooth with diagnostic significance in Giraffidae . However, in the genus Palaeotragus the morphology of p3 varies ( Laskos 2020: fig. 26). Some specimens of P. rouenii from Pikermi, Samos, and Kerassia ( Greece) retain a plesiomorphic ruminant condition. Other specimens, from Pikermi, Samos, Strumyani ( Bulgaria), Şerefköy-2, and Akkaşdağı ( Turkey), have a more derived morphology with a molarized mesial lobe. In those specimens, a trace of the anteparacristid is probably retained, leaving a very reduced parasinusid, in contrast to TSR-150. The p3 of P. microdon (e.g. AMNH 26360; see also Bohlin 1926) has a very reduced distal lobe, and the lingual wall, which begins at the trigonid, occupies almost the whole length of the tooth, and there is also an anteparacristid. P. coelophrys has, in general, a more primitive p3 morphology. However, a trigonid lingual wall associated with an anteparacristid is certainly observed in specimens of P. cf. coelophrys from China (see Zhang and Hou 2023; also, AMNH 26363), in the paratype of the enigmatic Palaeotragus moldavicus Godina, 1979 (PIN 649/66), and in two specimens of two large-sized Palaeotragus from Kerassia ( Iliopoulos 2003) and Thermopigi ( Xafis et al. 2019), referred to as Palaeotragus sp. The only available p3 of P. berislavicus from Berislav is also molarized, and it probably retains an anteparacristid ( Pidoplichko and Korotkevitch 1970: fig. 22). The p3 of the second more widely distributed palaeotragine genus, Samotherium , resembles the Late Miocene Palaeotragus in that it usually retains an anteparacristid. The p3 of TSR-150 is identical to some specimens of Late Miocene Palaeotragus and Samotherium , except for the absence of an anteparacristid. It is completely different from the p3s of other giraffid taxa, such as Giraffa , Okapia Lankester, 1901 , Bohlinia Matthew, 1929 , Deccenatherium, Palaeogiraffa Bonis and Bouvrain, 2003 , and the large-sized sivatheres ( Ríos et al. 2017, Laskos and Kostopoulos 2024).

The p4 of the Late Miocene Palaeotragus is less diverse than the p3. Many specimens of P. rouenii and P. coelophrys demonstrate a molarized p4, with a more mesiodistally oriented distal lobe and an entocristid more disconnected from the protocristid than in TSR-150. However, those differences might also be attributable to the stage of wear of the tooth. The p 4 specimens of P. rouenii (from Kerassia, Kryopigi), P. microdon and P. cf. coelophrys from China ( Zhang and Hou 2023), and P. coelophrys from Maragheh ( Iran) show a morphology identical to that of the Plio-Pleistocene taxon. The p4s of Samotherium are also identical to the p4s of both the Late Miocene Palaeotragus and the Plio-Pleistocene giraffid.

Although scarcer, the Plio-Pleistocene lower deciduous giraffid teeth show some differences from the Late Miocene Palaeotragus . The dp2 is always primitive in both the Late Miocene and Plio-Pleistocene samples. The dp3 appears to have a more primitive morphology in Late Miocene Palaeotragus than in Plio-Pleistocene known samples, in which there is a trigonid lingual wall. However, in the deciduous mandible PIK 1680 of P. rouenii from Pikermi, the dp3 is slightly molarized, having a trigonid lingual wall, although the anteparacristid and paracristid retain a primitive morphology. The dp3 of PNT-121f, attributed to P. coelophrys , is completely primitive ( Laskos and Kostopoulos 2022). Samotherium also demonstrates some variation in the morphology of the dp3; some specimens retain a primitive morphology (e.g. BMNH M4238), whereas others have a trigonid lingual wall (e.g. AMNH 9863, AMNH 22885, and MTLB 227). The Samotherium dp3s also retain an anteparacristid, as in the Late Miocene Palaeotragus . As is the case for the adult p3, the dp3s of the Plio-Pleistocene specimens TSR-151, AMPG Σ 184, and KZ 815 are similar to some dp3s of Late Miocene Palaeotragus and Samotherium , except for the absence of an anteparacristid. The rhomboid mesial lobe of dp4 is constant in both the Late Miocene Palaeotragus and the Plio-Pleistocene taxon.

The general biometrics of the postcranials of the Plio-Pleistocene ensemble are close to the large-sized Late Miocene Palaeotragus (e.g. P. coelophrys and P. berislavicus ), significantly larger than the small-sized P. rouenii and P. microdon , and, importantly, smaller than in Samotherium (see Athanassiou 2014). The proportions of the Plio-Pleistocene giraffid metacarpals are intermediate between those of P. rouenii and P. coelophrys ( Fig. 14A View Figure 14 ). There are only a few P. rouenii specimens (mostly from Cimislia and Tudorovo, Moldova; see Godina 1979) that have a similar length and robusticity index. Palaeotragus aff. berislavicus metacarpals from Nikiti-1 are shorter and more slender than the Plio-Pleistocene ones. Biometrically, the Plio-Pleistocene metatarsals are longer than those of P. coelophrys and are within the length range of P. rouenii , but generally they are more robust, close to the upper robustness values of that species ( Fig. 14B View Figure 14 ). It should be noted that the only P. rouenii metatarsal with comparable robustness and/or breadth to the Plio-Pleistocene taxon comes from Moldova (see Godina 1979), as it is the case of the metacarpals. In all cases, the breadth of the distal epiphysis of the Plio-Pleistocene giraffid metatarsals clearly exceeds that of P. rouenii , especially when compared with the length of the shaft. The dimensions of the distal epiphyses of the Plio-Pleistocene sample are much closer to those of P. coelophrys and P. berislavicus than to P. rouenii . The metapodials of the Plio-Pleistocene taxon are significantly more slender than the respective ones of Samotherium .

The proximal epiphysis of the Plio-Pleistocene metacarpals is morphologically similar to some Late Miocene Palaeotragus taxa, such as P. coelophrys and P. aff. berislavicus , in the rectangular shape of the lateral epicondyle in proximal view; Samotherium demonstrates the same shape, whereas P. rouenii has a more trapezoid shape. The shape of the articular facet for the capitato-trapezoid in proximal view is different in the Plio-Pleistocene metacarpals from that in the Late Miocene Palaeotragus . However, there is intrageneric diversity concerning that feature inside the Late Miocene Palaeotragus . Samotherium has a similar morphology to the Plio-Pleistocene taxon concerning the proximal view of the articular surface for the capitato-trapezoid; the medial side is slightly tilting and creates an obtuse angle with the dorsal side, which is arched.In Samotherium , the medial aspect of the articular surface of the capitato-trapezoid makes up ~60%– 62.5% of the total proximal depth of the proximal epiphysis. In the Plio-Pleistocene metacarpals, the same proportion is much higher,>75%. In this feature it resembles P. rouenii . The synovial fossa could be open or closed in Samotherium , and probably the same is true for the Plio-Pleistocene metacarpals. It is also closed in the Late Miocene P. coelophrys , but not in other Late Miocene Palaeotragus spp. The central trough is similar to that of the Late Miocene Palaeotragus and Samotherium . The dorsal trough is shallow and short in the Plio-Pleistocene giraffid, as it is in the Late Miocene Palaeotragus , whereas it is very pronounced in Samotherium , where it reaches the trochlea.

The prominent dorsal trough detected in the Plio-Pleistocene metatarsals is significantly shallower in the Late Miocene Palaeotragus but prominent in Samotherium . The proximal epiphysis of the Plio-Pleistocene taxon also resembles that of Samotherium more than that of the Late Miocene Palaeotragus , in that: the ventral head of the lateral epicondyle is more distally located in both the Plio-Pleistocene taxon and in Samotherium ; the articular facets for the cubonavicular and the intermediolateral cuneiform are dorsally contacting, and they have roughly the same size and shape; and the articular surface for the cubonavicular is aligned to the dorsoplantar axis of the bone. The proximity of the articular facets of the intermediolateral and medial cuneiforms seen in the Plio-Pleistocene sample also exists in Samotherium and in some Late Miocene Palaeotragus taxa. In contrast to the Plio-Pleistocene taxon, in the Late Miocene Palaeotragus the ventral head of the lateral epicondyle is not more distally located than the dorsal one, the shapes of the articular surfaces of the cubonavicular and intermediolateral cuneiform vary, and the articular facet for the cubonavicular is more obliquely oriented to the dorsoplantar axis of the bone.

The astragali of the Plio-Pleistocene taxon are very similar morphologically to those of the much larger Samotherium major Bohlin, 1926 ( Solounias and Danowitz 2016). The pronounced medial bulge of the collum tali and the tilting medial ridge of the ventral articular surface unify these taxa and distinguish them from the Late Miocene Palaeotragus . The distal intracephalic fossa is a feature that is probably lacking from the Late Miocene Palaeotragus , whereas it exists in S. major and in the Plio-Pleistocene giraffid, in which it is, however, much weaker. In contrast, the Late Miocene Palaeotragus and the Plio-Pleistocene taxon are similar in the very narrow medial scala, which appears wider in S. major . Distally, the groove of the head is deeper in S. major than it is in the Late Miocene Palaeotragus and in the Plio-Pleistocene taxon. Finally, using the indices provided by Ríos (2018), the Plio-Pleistocene ensemble is intermediate between the Late Miocene Palaeotragus and Samotherium concerning index 1 and shows a higher index 3 than in both Late Miocene genera, indicating a more symmetrical astragalus ( Table 2 View Table 2 ).