In terms of shape, the anatomical descriptions are restricted to the features that represent roughly the general morphological spectra of the astragali across different localities. The PC analyses of the selected characters showed an overall slight distinction between bison originating from different time frames and geographical regions, including both early and later forms. This separation is characterized by an intergroup overlapping; resulting significantly in the compactness of the clusters and therefore it was necessary to average observations for each group (Figs. 3b and 4b). The two first principal components (PC1 and PC2) convey a significant amount of the variance expressed by the 50 variables generated from the examined Cartesian coordinates, together explaining over 90% of the total variance for the proximal and distal articulation margins, respectively (Fig. 3 and Fig. 4). Given the considerable amount of variation accounted by the first axis in both analyses, in terms of landmarks positions, it is reasonable to consider that shape differences among the studied taxa are primarily explained by this component.

In the analysis of the proximal astragalar facet in the upper ankle joint, the first component, summarizing approximately 84% of the variation in the sample (Fig. 3a), is defined by the movement of landmarks positioned on the most proximo-anterior and the most distoposterior extremities, respectively and also those outlining the central area of this radius (Fig. 3b). A significant range of variation is evident along this axis regarding the mean shape differences of this aspect (Fig. 3a, b), as shown by the D _Pr values (Table 1). The astragalus of the Apollonia bison and Bison schoetensacki population from Süssenborn (high positive values along PC1) are characterized by a proximodistal (superior-inferior) compression and an anteroposterior deepening of the articular surface for the distal tibia. This morphological feature, also describes, to a lesser degree, the astragalus of the other Epivillafranchian form, Bison menneri from Untermassfeld. The mean configuration found in the astragali of the British Bison priscus along that of Bison bonasus seem to segregate them from the aforementioned general pattern, exhibiting a wider proximal joint surface (low PC1 negative values). A rather intermediate morphological condition could possibly been depicted by the astragali of the Taubach priscoid bison, Bison schoetensacki from Isernia La Pineta and, to some extent, the Venta Micena bison. PC2 represents only minor variation (accounting for 6.5%, Fig. 3a), reflected primarily in the extremities of the proximal articular surface (Fig. 3b). The astragali of the Apollonia bison fall mostly within the morphospace defined by the positive values, whereas the British Bison priscus and the Bison schoetensacki from Süssenborn tend towards the opposite direction of this component.

A different pattern is apparent in the analysis of the distal trochlear margin on the medial side of the astragalus (Fig. 4a, b, Table 1). Regarding the morphological traits exposed, it appears that sample overlapping is again detectable rather than distinctive morphotypes, as visualized along the PC1 axis (summarizing ∼85% of total variation) and indicated by the mean shape differences expressed in D _Pr values. The position of the modern wisent corresponds to forms that score mostly positively on the first component. The relative displacement of landmarks (Fig. 4b) reflects a more obvious anteroposterior compression and the proximodistal expansion of the trochlear margin along the transverse tarsal joint. Furthermore, one group, consisting of B. menneri (Untermassfeld) and the middle Pleistocene B. schotensacki (Isernia La pineta, Süssenborn) is distributed towards the positive end of PC1. The priscoid bison from Taubach and the Apollonia bison tend to be placed on the other edge of this morphological spectrum (PC1 negative scores), displaying a relatively anteroposteriorly broader and shortened distal articular radii. The early Pleistocene astragali of Venta Micena bison and the British priscoid form seem to resemble, in a lesser extent, the above anatomy in their distal astragalar facet. Similarly, the second component does not seem particularly diagnostic as the majority of the studied populations falls close to the origin of PC2, possibly with the exception of the Apollonia bison (positive scores), the British priscoid bison and the European wisent (negative scores). The slight morphological variation along PC2 (explaining ∼7.4%, Fig. 4a) reflects the anteroposterior relative displacement of landmarks, mainly on the posterior edge of the distal astragalar facet (Fig. 4b).

Size-shape relationship is given in the linear regressions of PC1 with centroid size for both margins (Fig. 5, plots of the rest PCs not shown here). In the studied material, shape was not found to be a function of size given that there is no significant correlation between centroid size and the first component and that the comparison of the studied populations revealed clear size differences (Table 2). A set of analysis of variance (ANOVA, pairwise comparison tests among the samples of different localities) was conducted on the first component scores derived by the PCA on the Procrustes coordinates (Table 3). The observed shape variation of the two astragalar facets along PC1 is supported by significant ANOVA results (P < 0.05). Regarding the proximal margin, significant intergroup differences are not evident based on the Tukey-HSD pairwise comparison, however due to the conservative character of the test; differences could be suggestively detected between the Apollonia bison in relation to B. bonasus (Table 3). On the other hand, multiple comparisons succeed to distinguish the distal margin shape of several populations. More particularly, the Tukey-HSD test identifies differences between B. menneri and the other Early Pleistocene bison from Venta Micena and Apollonia and the B. priscus from Taubach as well. Furthermore, significant differences are observed between the three aforementioned bison populations and the European wisent (Table 3).

Regarding the geometric patterns revealed on the medial side of the astragalus, several shape trends are mainly recognized based on the combined morphological traits of each examined anatomical region (Fig. 6). Furthermore, a significant overlapping of the majority of the examined bison populations is apparent; however mean shape changes along PC1 do not seem to follow size variability and possibly are associated with an open-close habitat gradient. A moderate anteroposterior shortening and proximodistal deepening of the proximal and distal trochlear radii, respectively, is roughly represented by the astragali of the Epivillafranchian bison form B. menneri and the B. schoetensacki representative of the middle Pleistocene from Süssenborn (Fig. 6). Part of the studied extinct and extant Bison material was assigned exclusively to the same part of the exposed morphological spectra. More specifically, the early Pleistocene bison from Apollonia seems to be placed alongside the former group, according to the shape of the proximal trochlear margin, but resembles the priscoid bison from Taubach based on the extensive broadening of the distal margin (Fig. 6). On the other hand, the reverse configuration, in relation to the one described for the Apollonia bison, is uncovered in the modern wisent, the relatively wide and shallow proximal astragalar trochlea of which also corresponds to the morphology of the astragalus of the British Last Interglacial B. priscus population (Fig. 6). Concerning the shape of the two astragalar margins on the rest of the material, a rather intermediate morphological pattern can be distinguished for the B. schoetensacki population from Isernia La Pineta. The astragali of the Venta Micena primitive form are characterized by a fairly wide and shallow proximal trochlea, whereas a more prominent anteroposterior broadening can be exposed in the distal part of the bone (Fig. 6).

It is questionable how the spatial and temporal phenotypic diversity could be ascribed to locomotor adaptations and linked to discrete habitats. In the study of DeGusta and Vrba (2003), astragali of forest and heavy cover taxa appeared to be anteroposteriorly compressed, whereas light and open cover taxa show anteroposteriorly expanded astragali. Barr, 2014 and Barr, 2015 claimed that open-country bovids have relatively short astragali in comparison to closed-country bovids, contributing in an increased range of angular excursion. Furthermore, according to these and other authors (e.g., Hildebrand and Goslow, 2001, Plummer et al., 2008 and Plummer et al., 2015) the measured distal trochlear radius appears to distinguish successfully the open and closed habitat dwellers, as it tends to be greater in open-country bovids in comparison to closed-country forms. Their work has shown that, apart from distal limb relative elongation, increased stride cycles can be achieved via this morphological arrangement, as it hypothetically provides a wider rotational arc and, thus, a correspondingly increased motion range and overall limb velocity.

These assumptions could be also applied on the current findings. Consequently, even within the general restricted locomotor repertoire of bison, it is possible to segregate more and less cursorial forms. In this sense, the shape of the distal astragalar trochlea of Venta Micena and Apollonia primitive bison and the two priscoid bison of the Last Interglacial could be related with an increased medial depth of the tarsal articulation (Fig. 4 and Fig. 6), an adaption associated with rapid locomotion and operating in relatively open environments. This grouping is fairly consistent with that reported in a previous study by Maniakas and Kostopoulos (2017), where according mainly to the dental mesowear signature and the relative metapodial shape, the late Villafranchian/Epivillafranchian bison from Apollonia has been found to be suited to overall open and dry habitats with well-developed grassy cover. The rather increased anteroposterior distal astragalar mobility of the Venta Micena bison could be even compatible with the presence of a partially wooded open savanna, as inferred by biogeochemical and ecomorphological data (see Mendoza et al., 2005, Palmqvist et al., 2008).

Furthermore, the extensively massive late Pleistocene priscoid forms have been also linked with the dwelling of environments with increased soil moisture as was further indicated in the relative trochlear verticilli splaying (Maniakas and Kostopoulos, 2017). Kahlke (1999) claimed that the bison of the Eemian Interglacial have to be adapted to soft woodland forage in continentally influenced conditions; however, the revealed astragalar morphology of the two B. priscus populations links these two forms with a rather advanced cursoriality and open biomes. We can assume that this ecomorphological feature does not necessarily contrast with the vegetational properties suggested for the early and late/latest Pleistocene palaeoenvironments or the need for an increased stability during cursorial locomotion in (constantly or periodically) less solid substrates (see Maniakas and Kostopoulos, 2017). Furthermore, plain dwelling should not be surprising for these two forms given their heavy-built structure. Nevertheless, B. priscus was a taxon that originally immigrated the European subcontinent from Asian steppe-like environments following the landscape opening in the later part of early middle Pleistocene (see Kahlke et al., 2011).

In accordance with the aforementioned interpretations, the relatively narrower distal trochlear condyle seems to be functionally related to decreased anterosterior mobility and powerful plantar flexion, i.e., a more closed habitat (Barr, 2014, Barr, 2015, Hildebrand and Goslow, 2001 and Kappelman, 1988). Under this scope, it is reasonable to attribute the astragalar morphology that characterizes the group consisting of the Epivillafranchian B. menneri and the two middle Pleistocene B. schoetensacki representatives (Fig. 5), to the generation of roughly shorter stride cycles, despite their relatively long-legged distal limbs (see Maniakas and Kostopoulos, 2017 and Sher, 1997). This would come in some agreement to the reported sparsely wooded grasslands of Süssenborn and Isernia La Pineta (Kahlke et al., 2011 and Sala, 1986), as well as to the mixture of open and gallery forested landscapes proposed for Untermassfeld (Kahlke, 2006 and Kahlke et al., 2011), especially in comparison to the affiliated southern geographic regions of Apollonia and Venta Micena. This distinction seems to correspond with the assumption of Sher (1997) and Sala (1986) that bison of light construction, like B. menneri and B. schoetensacki, had to be forest dwellers. On the other hand, their habitat should be expected to differ significantly in vegetational composition in relation to that of the forest-dwelling modern European bison. In order to fit further with the “railed” trochlear condyle configuration (decreased splaying) reported by Maniakas and Kostopoulos (2017) for these forms, solid substrate conditions could be associated with relatively more arid climate conditions and the possible development of mixed sclerophyll woodlands and shrublands.

According to previous studies (Barr, 2014, Plummer et al., 2008 and Plummer et al., 2015), an analogous tightly interlocking formation mechanism was not supported for the proximal trochlea. Barr (2014) reported no significant differences regarding the proximal radii between open-country bovids and forest or heavy cover (a term referring to bushlands, woodlands and swampy habitats) dwellers, which presented surprisingly higher values than those found in low cover (referring to light bushlands or tall grass habitats) forms, an inconsistency that was attributed partly to a body size signal. Similarly, confusing results were commented by Plummer et al., 2008 and Plummer et al., 2015. In the current study, a morphofunctional interpretation for this anatomic region appears also rather problematic, given that the supposed open-country Apollonia bison demonstrates a relatively narrower proximal joint margin in relation to that of the modern wood wisent. These findings perplex the prediction that a larger proximal radius could be associated with an increased range of angular excursion during dorsiflexion (sensu Schaeffer, 1947), indicating advanced cursoriality. Furthermore, the relative magnitude of shape changes between different bison populations was not supported (Table 3b). Following Barr's (2014) allegation, we might also consider the possibility that other factors than habitat have influence on the observed intrageneric shape variation of the trochlear diameter of the tibial articulation, suggesting that further studies are required in order to verify its relation to habitat specific locomotion, particularly at the intra-genus level.