Postcranial elements of small mammals as indicators of locomotion and habitat

Principal Components Analysis

The first component (explaining 24.5% of the variance) distinguishes proximal humeri with a more ovoid humeral head, a relatively larger greater tuberosity and a relatively smaller lesser tuberosity (positive values), from humeri with a more rounded humeral head, a relatively small greater tuberosity and a relatively large lesser tuberosity (negative values). The second component (explaining 17% of the variance) distinguishes taxa primarily on the shape of the greater tuberosity: humeri with the cranial portion of the tuberosity projecting medially (and a slightly larger lesser tuberosity) have positive values, while humeri where the cranial portion of the tuberosity projects anteriorly (and a slightly smaller lesser tuberosity) have negative values. The majority of the terrestrial taxa have positive scores on the first component. Arboreal taxa tend to have positive scores on the second component and terrestrial taxa tend to have negative scores (see Figs. 6A and 6B).

Canonical Variates Analysis

The differences in humeral morphology along the first axis are similar to those along PC1, although there is less difference in the size of the lesser tuberosity between positive and negative values. The morphological changes along the second axis differ from those in PC2 (although the differences in the size of the lesser tuberosity are similar): here humeri with larger greater tuberosities have more positive values although, as with the second component of the PCA, the cranial portion projects medially. The first axis (explaining 71.1% of the variance) largely separates terrestrial taxa (positive scores) from arboreal ones (negative scores), with scansorial taxa lying in between these two groupings. The second axis (explaining 28.9% of the variance) does not appear to contain a strong locomotor signal. Marsupials group with the similar locomotor group of placentals (see Figs. 6C and 6D).

Using Mahalanobis distances between groups, the CVA distinguishes all three locomotor categories from each other; using Procrustes distances arboreal taxa are distinguished from scansorial and terrestrial ones with a high degree of significance, but scansorial and terrestrial taxa cannot be distinguished from one another. Out of the original grouped taxa, 70.5% were correctly classified (50% for cross-validated groups) (see Table 2).

Principal Components Analysis

The first component (explaining 38.8% of the variance) distinguishes humeri with a distal articular surface that is long and tubular, with a rounded capitulum and an elongated and narrow trochlea (positive values), from humeri with a distal articular surface that is short and square, with an especially short and broad trochlea with a pronounced median keel (negative values). The second component (explaining 16.7% of the variance) distinguishes humeri with a short and square capitulum with a pronounced capitular tail and a fairly short and broad trochlea (positive values), from humeri with a more rectangular capitulum and a more elongated trochlea (negative values). Almost all of the arboreal taxa have positive scores on the second component, and most have positive scores on the first component: terrestrial taxa tend to have the converse pattern of scores (negative on both components) (see Figs. 7A and 7B).

Canonical Variates Analysis

The differences in humeral morphology along the two axes are similar to those seen in the PCA. The first axis (explaining 88.3% of the variance) separates arboreal taxa (positive scores) from terrestrial taxa (mainly negative scores): scansorial taxa lie in between these two groups on the first axis, but tend to have more negative scores on the second axis (which explains 11.7% of the variance). Marsupials group with the similar locomotor group of placentals (see Figs. 7C and 7D).

Using both Mahalanobis and Procrustes distances among groups, arboreal taxa can be distinguished from scansorial and terrestrial ones with a high degree of significance, but scansorial and terrestrial taxa cannot be distinguished from one another (even with the inclusion of the medial epicondyle). Out of the original grouped taxa, 67.5% were correctly classified (47.4% for cross-validated groups) (see Table 2).

Principal Component Analysis

The first component (explaining 38.5% of the variance) distinguishes ulnae with a more prominent anconeal process, a less prominent and less sloping coronoid process bordering a more constrained semilunar notch, and a straighter posterior border to the ulna (reflecting a longer and straighter olecranon process) (positive values), from ulnae with an articular surface with a less prominent anconeal process and a more projecting, more sloping coronoid process bordering a more open semilunar notch, and a more curved posterior border to the ulna (reflecting a short, anteriorly-curved olecranon process) (negative values). The second component (explaining 19.2% of the variance) further reflects the morphology of the anconeal process, distinguishing ulnae with a smaller anconeal process and a more curved posterior ulna border (positive values), from ulnae with the opposite morphology (negative values). Terrestrial taxa tend to have more positive scores on the first component and arboreal taxa more negative scores, although there is much overlap (see Figs. 8A and 8B).

Canonical Variates Analysis

Anatomical changes along the first axis (explaining 93.2% of the variance) are essentially reversed from negative to positive values from those seen in the PCA. The anatomical differences along the second axis (explaining 6.8% of the variance) are more subtle than in the PCA, although ulnae with a more prominent anconeal process and a deeper semilunar notch have positive values. The first axis separates arboreal taxa (positive scores) from terrestrial ones (negative scores); scansorial taxa lie in between these two groups on the first axis, but tend to have more negative scores on the second axis. Marsupials group with the similar locomotor group of placentals (see Figs. 8C and 8D).

Using Mahalanobis distances among groups, the CVA distinguishes all three locomotor categories from each other high degree of significance, but using Procrustes distances only arboreal taxa can be distinguished from terrestrial ones, and neither can be distinguished from scansorial taxa. Out of the original grouped taxa, 78.4% were correctly classified (58.1% for cross-validated groups) (see Table 2).

Principal Component Analysis

The first component (explaining 47.6% of the variance) distinguishes radii where the proximal articular surface is ovoid in shape (positive values), from radii where the shape is round (negative values). Terrestrial taxa tend to have positive scores and arboreal ones negative scores, although there is much overlap. The second component does not explain a significant amount of the variance (see Figs. 9A and 9B).

Canonical Variates Analysis

The differences in proximal radial anatomy along the first axis (explaining 59% of the variance) are similar to those along the first component in the PCA. Differences along the second axis (explaining 41% of the variance) appear to relate as to whether the posterior border is rounded (positive values) or straight (negative values). The first axis largely separates terrestrial taxa (positive scores) from arboreal ones (negative scores). The scansorial taxa lie in between these two groups on the first axis, but mainly have negative scores on the second axis (see Figs. 9C and 9D).

Using Mahalanobis distances among groups, the CVA distinguishes all three locomotor categories from each other, but the distinction between scansorial taxa and the other two groups is only weakly significant: using Procrustes distances arboreal taxa are distinguished from terrestrial ones with a high degree of significance, and from scansorial ones with a weaker degree of significance: scansorial taxa cannot be distinguished from terrestrial ones. Out of the original grouped taxa, 67.8% were correctly classified (42.4% for cross-validated groups) (see Table 2).

Principal Components Analysis

The first component (explaining 42.6% of the variance) distinguishes femora with a long femoral neck and a posteriorly (and distally) placed lesser trochanter (positive values), from femora with a short femoral neck and a medially (and proximally) placed lesser trochanter (negative values). The second component (explaining 16.9% of the variance) distinguishes femora with a short femoral neck, a somewhat medially (and proximally) placed lesser trochanter, and a greater trochanter significantly lower than the femoral head (positive values), from femora with a long femoral neck, a slightly more medially (and more distally) placed lesser trochanter, and an elevated greater trochanter that is at the same level as the femoral head. Terrestrial taxa tend to have positive scores on the first component, and arboreal taxa negative scores (see Figs. 10A and 10B).

Canonical Variates Analysis

This analysis picks out rather different morphological aspects of the femoral head to the PCA. The first axis (explaining 56.6% of the variance) distinguishes femora with a larger femoral head and a more medially and distally placed lesser trochanter (positive values), from femora with a small femoral head and relatively short neck with a slightly posteriorly placed lesser trochanter (negative values). The second axis (explaining 43.4% of the variance) distinguishes femora with a large femoral head, a long femoral neck, a posteriorly placed lesser trochanter, and a prominent greater trochanter (positive values), from femora with a smaller femoral head, a short femoral neck, a strongly medially-placed lesser trochanter, and a small greater trochanter. The first axis separates arboreal taxa (and many terrestrial taxa) (positive scores) from scansorial taxa (and a few terrestrial taxa) (negative scores). The second axis mainly separates terrestrial taxa (positive scores) from arboreal and scansorial ones (negatives scores). With a few exceptions, marsupials group with the similar locomotor group of placentals (see Figs. 10A and 10B).

Using Mahalanobis distances among groups, the CVA distinguishes all three locomotor categories from each other, but using Procrustes distances only arboreal taxa vs terrestrial are distinguished a high degree of significance. Out of the original grouped taxa, 81.1% were correctly classified (58.1% for cross-validated groups) (see Table 2).

Principal Components Anaysis

The first component (explaining 58.4% of the variance) distinguishes distal femora that are short and broad, lacking an obvious patella groove, and a having broad distal end to the lateral condyle (positive values), from femora with articular surfaces that are long and narrow, with a prominent patella groove and a narrow distal end to the lateral condyle (negative values). This component primarily separates taxa on phylogenetic grounds (as previously noted, there is a marked difference between marsupials and placentals in the degree of femoral condyle asymmetry): in general, marsupials have positive scores and placentals have negatives scores. The second component (explaining 13.4% of the variance) does not distinguish the overall height of the articular surface, but femora with a moderately deep patella groove (and a relatively narrow distal portion of the articular surface), plus a relatively narrow lateral condyle and a deep incursion between the two condyles have positive values. In contrast, femora with a shallow patella groove (and a relatively broad distal portion of the articular surface), plus a relatively broad lateral condyle and a shallow incursion between the two condyles have negative values (see Figs. 11A and 11B).

Canonical Variates Analysis

The first axis (explaining 69.1% of the variance) distinguishes femora on similar morphological features to the first principal component, but there is less of a distinction of the symmetry of the widths of the distal condyles. The second axis (explaining 30.9% of the variance) again reflects somewhat similar morphological differences to the second principal component, but with less emphasis on distal femoral asymmetry or the depth of the incursion between the two condyles. Taxa are separated by locomotor category rather than by phylogeny: almost all arboreal taxa have positive scores on the first axis, and almost all terrestrial taxa have negative scores. The scansorial taxa lie in between these two groups, but tend to have more negative scores on the second axis. With a few exceptions, marsupials group with the similar locomotor grouping of placentals (see Figs. 11C and 11D).

Using Mahalanobis distances among groups, the CVA distinguishes all three locomotor categories from each other, but using Procrustes distances only arboreal taxa vs terrestrial are distinguished a high degree of significance. Out of the original grouped taxa, 77.1% were correctly classified (60.2 % for cross-validated groups) (see Table 2).

Principal Components Analysis

Neither component explains a high percentage of the variance, and the differences in shape are subtle, at best. The first component (explaining 26.6% of the variance) distinguishes the tibial articular surfaces primarily on the basis of the relative size and orientation of the tibial condyles, especially the lateral condyle which varies from being constricted posteriorly (positive values) to being broadened posteriorly (negative values). The second component (explaining 17.8% of the variance) distinguishes tibial articular surfaces primarily on form of the tibial tuberosity and the intercondylar notch: tibiae with a more prominent, medially-directed tuberosity and a constricted intercondylar notch (positive values), and tibiae with a less prominent, laterally-directed tuberosity and a less constricted intercondylar notch (negative values). The distribution of taxa appears to have little locomotor pattern, and instead more reflects phylogeny (see Figs. 12A and 12B).

Canonical Variates Analysis

The first axis (explaining 75.8% of the variance) distinguishes tibiae on the basis of slight differences in shapes of the condyles: additionally, between tibia with a shallow intercondylar notch (positive values) and tibiae with a deep intercondylar notch and a more prominent, laterally-directed tuberosity (negative values). The second axis (explaining 24.2% of the variance) distinguishes tibiae where both condyles are somewhat contracted posteriorly, and the tibial tuberosity is slightly directed laterally (positive values), from tibiae where both condyles (especially the lateral one) are expanded posteriorly (the lateral more than the medial), and the intercondylar groove is very shallow. The first axis separates the arboreal and scansorial taxa (positive scores) from most of the terrestrial taxa (with mostly negative scores). The second axis separates the arboreal taxa (positive scores) from almost all of the scansorial taxa (negative scores), and the scansorial taxa have the most negative scores on this axis. Marsupials group with the similar locomotor grouping of placentals (see Figs. 12A and 12B).

Using Mahalanobis distances among groups, arboreal forms can be distinguished from terrestrial forms with a high degree of significance, as can terrestrial forms from scansorial forms, but using Procrustes distances none of the locomotor groups can be distinguished from each other. Out of the original grouped taxa, 77.1% were correctly classified (33.3 % for cross-validated groups) (see Table 2).

Principal Component Analysis

There is an overall locomotor signal, especially within the different phylogenetic groupings, relating primarily to the degree to which the humerus is stabilized on the scapula by the rotator cuff muscles. In general, taxa with positive scores on PC1 and negative scores on PC2 are terrestrial forms with the humeroscapular motion more confined to the parasagittal plane, as indicated by the larger greater tuberosity. The larger lesser tuberosity seen in taxa with negative scores on PC1 and positive scores on PC2 may relate to the medial rotation of the humerus during use of the forelimb during active climbing (see discussion in Janis et al. (2020)); this may be the reason why it is the arboreal taxa, but not the scansorial ones, that have this combination of scores along the components.

Carnivorans in general have positive scores on PC1, but they sort out by locomotor grouping along PC2 (arboreal forms with positive scores, scansorial and terrestrial forms with mostly negative ones). Almost all of the marsupials have negative sores on PC2, but sort out by locomotor mode along PC1 (the arboreal forms having the most negative scores). Most of the terrestrial caviomorph rodents have negative scores on PC2 (Chinchilla (= Chin) and Dasyprocta (= Dasypro) being exceptions), while the arboreal and scansorial rodents mostly have positive ones. Among the “insectivores”, the scansorial scandentians have negative scores on PC2, but the terrestrial afrosoricidans and lipotyphlans have positive scores on this component (with the exception of the giant shrew Crocidura (= Crocid)). Primates have negative scores on PC1 and positive scores on PC2 (with the exception of the saki, Chiropotes (= Chirop), with a positive score on PC1) (see Figs. 6A and 6B).

Canonical Variates Analysis

The first axis has a strong locomotor signal, but arboreal and terrestrial taxa are both widely scattered on the second axis, with no apparent correlation with phylogeny. The preponderance of arboreal taxa with negative scores on the first axis resembles their distribution in the PCA, reflecting the possession of a larger lesser tuberosity and a smaller greater tuberosity. The majority of the scansorial taxa have negative scores on the second axis, but this does not distinguish them from the other two locomotor groups. The arboreal carnivorans have the among the least negative scores out of the arboreal forms, clustering with a few of the terrestrial carnivorans along this axis. Arboreal forms with positives scores on the first axis include the possum Phalanger (= Phalan) and the caviomorph rodent Coendu. A couple of scansorial taxa have highly positive scores on the first axis, clustering among the terrestrial taxa: the carnivoran Genetta and the quoll Dasyurus maculatus (= Dasyum) (see Figs. 6C and 6D).

Principal Components Analysis

There is distinct locomotor signal along PC1, similar to that found along the first component in the PCA of Figueirido, Martín-Serra & Janis (2016), distinguishing between humeral anatomy allowing for a high degree of rotation at the elbow joint in arboreal taxa (positive scores), from that restricting the amount of rotation and providing a greater degree of forelimb stability with the forelimb locked into a prone position in terrestrial ones (negative scores). An extremely elongated and narrow trochlea is especially prominent in arboreal marsupials. Our results here reflect this combination of functional and phylogenetic distinction: taxa with high positive scores on PC1 are almost exclusively arboreal and scansorial marsupials (plus a couple of terrestrial marsupials), such as the dasyurid Antechinus (=Antes) and the didelphid Metachirus (= Metach), and also lemuriform primates). Other primates and, with a few exceptions, arboreal carnivores and rodents have lower positive scores, as do the peramelid marsupials (all terrestrial). Taxa with negative scores on PC1 are primarily terrestrial and scansorial carnivorans and rodents. Some anomalies exist: some terrestrial carnivorans (Spilogale (= Spilo)) and rodents (Microtus and Thomomys (= Thomo) have positive scores on PC1, clustering with the arboreal taxa. Conversely, some arboreal forms have negative scores, clustering with the terrestrial taxa (e.g., the carnivoran Ailurus, the caviomorph rodent Erithizion (= Erith), and the dasyurid marsupial Phascogale (= Phasco)). The terrestrial “insectivores” are scattered along PC1 (but all with negative scores on PC2) (see Figs. 7A and 7B).

Canonical Variates Analysis

The taxa that fall in the exclusively arboreal portion of the first axis (positive scores) are, as with the proximal humerus, primarily primates and arboreal marsupials, and some of the arboreal rodents. With the exception of Potos, the arboreal carnivorans have less positive scores, falling close to the overlap with the terrestrial taxa. Terrestrial taxa falling within the confines of the arboreal area along PC1 are mostly carnivorans (the highest positive scores belonging to Spilogale and Galidictis (= Galidf)), plus the rodent Microtus and the dasyurid marsupial Antechinus (= Antes). Both Microtus and the terrestrial caviomorph rodent Dasyprocta are notable outliers on the positive side of the second axis (see Figs. 7C and 7D).

Principal Components Analysis

There is a considerable scatter of locomotor modes along PC1, but in general arboreal taxa have negative scores, with an ulna morphology reflecting considerable rotational movement of the forearm around the humeroulnar joint. The exceptions here are the arboreal caviomorph rodents and the dasyurid marsupial Phascogale, with positive scores. With a few exceptions (e.g., the scansorial procyonid carnivoran Nasua (although note that this taxon is not an outlier on the CVA)), the taxa with the highest negative scores are arboreal and scansorial marsupials, and terrestrial marsupials also tend to have negative scores (e.g., the bandicoot Isoodon (= Isood)), demonstrating an effect of phylogeny as well as locomotion. In contrast, most scansorial and terrestrial placentals have positive scores on PC1 (a notable exception being the caviomorph rodent Cavia) (see Figs. 8A and 8B).

Canonical Variates Analysis

Although the overlap between scansorial forms and the other locomotor groups is considerable, there is little overlap between terrestrial and arboreal forms. The arboreal forms with the highest positive scores along the first axis are mainly the marsupials (especially Petaurus breviceps (= Petaurb)), although the striped possum Dactylopsila (= Dacty) has low positive scores, as do a couple of primates (e.g., Microcebus (= Microc) and Callithrix (= Calli)). Terrestrial taxa with very high negative scores include the afrosoricidan Tenrec (with a high positive score on the second axis), and the didelphid marsupial Metachirus and the caviomorph rodent Dasyprocta (both with high negative scores on the second axis) (Figs. 8C and 8D).

Principal Components Analysis

A round radial head, permitting rotation of the forelimb around the elbow joint is well known to distinguish arboreal taxa from terrestrial ones (which have an ovoid radial head, restricting forelimb rotation) (MacLeod & Rose, 1993). Here, perhaps because we have added marsupials to the mix of taxa, this PCA reveals a strong phylogenetic component, and there is little functional signal. All of the marsupials have negative scores on the first axis, perhaps reflecting the need for rotation of the forelimb in early ontogeny (or alternatively simply reflecting the retention of a more conservative type of forelimb anatomy). Within the array of marsupials in the morphospace there appears to be no correlation with locomotor mode: the highest negative scores along PC1 belong to terrestrial australidelphids, the bandicoot Isoodon and the dasyurid Sarcophilus (= Sarco). The placentals with high negative scores on this component are primarily primates (and, strangely, the terrestrial sciuromorph rodent Marmota (= Marmot)). Taxa with positive scores on PC1 are primarily terrestrial and scansorial rodents and carnivorans, and the arboreal members of these groups tend to have more negative scores (with the notable exception of the caviomorph rodent Echimys) (see Figs. 9A and 9B).

Canonical Variates Analysis

Although there is a large overlap between the arboreal and terrestrial taxa, a locomotor signal is apparent. Those arboreal taxa that have highly negative scores on the first axis are primates and the arboreal phalangeriform marsupials, and the terrestrial taxa with highly positive scores on this axis are primarily lipotyphlan and afrosoricid “insectivores.” The peramelid marsupials, which usually cluster near to one other in our analyses, show widely divergent anatomies of the radial head. Macrotis (the bilby, = Macrot) has one of the most positive scores on the first axis, clustering with the terrestrial placentals: but one of the bandicoots (Isoodon) has high negative scores, while the other (Perameles (= Peram)) falls within the overlap of the three locomotor groups. The second axis potentially distinguishes scansorial taxa from the other locomotor groups. Taxa with highly negative scores are primarily squirrels (sciuromorph rodents), but also the carnivoran Genetta, and the scansorial carnivorans Nasua and Bassariscus (= Bassas) (see Figs. 9C and 9D).

Principal Components Analysis

The first component distinguishes between taxa where the hindlimb motion is more restricted to the parasagittal plane (positive scores) from those where there is a greater degree of rotation of the hip joint (negative scores). Although PC1 does largely reflect locomotor type, almost all of the marsupials have negative scores on this component, the bandicoot Perameles being the exception. However, within the marsupials, the arboreal and scansorial taxa have the most negative scores on PC1 (with the exception of the diminutive terrestrial dasyurid Antechinus (= Antef)), sharing this area of the first component with (largely) scansorial sciuromorph rodents and scandentians, but also with the terrestrial “insectivores” (the latter also having positive scores on PC2). Interestingly, the primates do not have highly negative scores on PC1, with some of the anthropoids (e.g., Saimiri (= Saimi)) even having weakly positive scores. In the positive portion PC1 the carnivorans and caviomorph rodents largely sort out along an arboreal plus scansorial (lower scores) to a terrestrial (higher scores) axis. PC2 does not appear to distinguish taxa on either locomotor mode or phylogeny (although the leaping primates do have low scores) (see Figs. 10A and 10B).

Canonical Variates Analysis

There is quite a large overlap between terrestrial and arboreal taxa along the first axis, although the arboreal rodents have the highest positive sores. The highest negatives scores belong to either terrestrial forms (rodents (Cavia and Tamias), the carnivoran Viverricula (= Viverr), and the bandicoot Perameles), which also have positive scores on the second axis, and scansorial sciuromorph rodents, which also have negatives scores on the second axis. Arboreal primates and carnivorans (plus some arboreal marsupials) tend to fall in the overlap between the terrestrial and arboreal taxa on the first axis. The second axis shows a somewhat clearer separation between terrestrial taxa (mostly positive scores) and scansorial plus arboreal taxa (mostly negative scores). The main terrestrial outliers on the second axis (with high negative scores) are the rodent Chinchilla, the dasyurid Antechinus and the “insectivore” Setifer and the arboreal outlier (with a high positive score) is the rodent Toromys (= Torom) (see Figs. 10C and 10D).

Principal Component Analysis

Taxa are separated primarily by phylogeny. With the exception of the peramelids (the only cursorial quadrupedal marsupials included here), and the dasyurid Myoictis (= Myoic), all of the marsupials have positive scores on PC1, and all of the taxa with scores of ~0.14 or greater are marsupials (Sarcophilus being the only terrestrial taxon among them). Phascogale, the sole arboreal dasyurid, is the only arboreal marsupial not in this grouping. In the more negative portion of PC1, occupied mainly by placentals, the arboreal taxa have positive (or only slightly negative) scores (reflecting a femoral morphology less restricted to parasagittal motion): the terrestrial taxa in general have the most negative scores, and the scansorial taxa fall in the middle. The exceptions are as follows: the terrestrial “insectivorans” Solenodon (= Soleno), Setifer, Atelerix (= Atelx) and Crocidura have positive scores and cluster with the arboreal placentals, and the lemuriform primates (Microcebus and Cheirogaleus (= Cheiro)) have highly negative scores, clustering with the terrestrial placentals. The case of the lemuriforms is more easily explained: these little primates move from tree to tree by leaping, and their femoral morphology reflects a relatively restricted parasagittal motion of the lower limb on the femur.

Along PC2, the arboreal marsupials have almost entirely negative scores, but among the placentals the negative scores belong mainly to the terrestrial and scansorial taxa. However, this component does separate, to a large extent, carnivorans (with positive scores) from terrestrial and scansorial rodents (especially caviomorphs) (with negative scores) (see Figs. 11A and 11B).

Canonical Variates Analysis

There is a moderate degree of overlap between the arboreal taxa (positive scores) and the terrestrial ones (negative scores) along the first axis: the main taxa with anomalous placements being the arboreal carnivoran Ailurus (with a negative score) and terrestrial rodent Thomomys (with a positive score). The highest scores on the first axis belong to arboreal and scansorial marsupials, and to primates, with the arboreal caviomorph rodents having the lowest positive scores. The bandicoot Isoodon and the carnivoran Viverricula are outliers on the first axis with highly negative scores. On the second axis, all of the scansorial taxa have negative scores, with the exception of the didelphid marsupial Marmosops (= Marmn) and the carnivoran Prionodon (= Priono). However, both arboreal and terrestrial taxa span positive to negative scores on the second axis, and there does not appear to be any influence of phylogeny to the pattern of their distribution.

Principal Components Analysis

The separation of taxa here largely reflects phylogeny. Almost all of the carnivorans all have negative scores on PC1, while the marsupials have mostly positive scores. The caviomorph rodents all have positive scores, while the other rodents mostly have negative ones. The “insectivorans” and the primates all have scores that are around zero or weakly positive. There appears to be little in the way of functional or phylogenetic pattern to the placement of taxa on the second component.

Canonical Variates Analysis

Despite the relatively poor resolution of this analysis in terms of distinguishing locomotor mode, it is nonetheless interesting as this is the only analysis where the scansorial taxa are not positioned in between the arboreal and terrestrial ones on the first axis. Here, the scansorial taxa occupy the same part of the positive side of the first axis as the arboreal ones, and indeed are the ones with the most positive scores, and the arboreal taxa are restricted to the positive side of the second axis. Two arboreal carnivorans (Ailurus and Nandinia (= Nand)) have negative scores on the first axis, falling into the area of overlap between terrestrial and scansorial taxa, while the terrestrial carnivoran Eupleres has a positive score. Ailurus (the red panda) also clustered with terrestrial taxa in the CVAs for the femora and the distal humerus. Unlike the condition in many of the other analyses the terrestrial “insectivorans” all cluster with the other terrestrial taxa and, the carnivoran Viverricula is an outlier with highly negative scores. The scansorial taxa mostly have negative scores on the second axis, and include the taxa with the most negative scores.