A critical survey of vestigial structures in the postcranial skeletons of extant mammals

Specimen examination

We examined 578 mammalian skeletal specimens and skins from 293 species (Table 1) representing 98 (70%) of the 139 extant mammal families listed by Nowak (1999). We supplemented our observations with previously published descriptions as needed. We also examined manual and pedal morphology in 74 live members of 23 hoofed species of the Cetartiodactyla and in museum-supplied photos of seven skeletal specimens of rhinoceroses and manatees (Table 1).

Rodentia is the largest mammalian order, with over 1,700 species, of which over 70% are in the superfamily Muroidea (mice and kin) and the family Sciuridae (squirrels and kin) (Adkins, Walton & Honeycutt, 2003; Steppan, Storz & Hoffmann, 2004). Therefore, we were able to examine only a fraction of the diversity within those two taxa. Nevertheless, we achieved sufficient coverage of them to be of use here. Our overall rodent sample includes representatives of 23 (82%) of the 28 extant rodent families listed by Nowak (1999).

Identification of vestigial structures

Three categories of skeletal structures were examined and considered candidates for identification as vestigial structures: (1) individual bones, (2) parts of individual bones (e.g., the shaft of the ulna), and (3) multiple-bone structures (e.g., the pelvic girdle, a limb, or a digit). Previous studies have identified vestigiality in all three anatomical categories (Tague, 1997; Kearney, 2002; Bejder & Hall, 2002; Maxwell & Larsson, 2007; Senter, 2010; Bensimon-Brito et al., 2011; Moch & Senter, 2011).

We began by identifying examples of postcranial skeletal structures that were greatly reduced in comparison to their homologs in related taxa. To determine whether such structures could be considered vestigial we used the phylogenetic bracketing approach from a previous study (Moch & Senter, 2011). According to this approach, a structure is considered vestigial if it satisfies the three criteria listed below, in comparison to its homolog in three successive sister taxa (Table 2). A phylogenetic bracketing approach only requires confirmation of a character state in two successive sister taxa (Witmer, 1995), but we included a third so as to increase the reliability of the inference.

The first criterion for vestigiality is that in comparison to its state in the sister groups the structure exhibits extreme reduction. For this study, we considered this criterion met if the structure was reduced to one-third its size relative to adjacent skeletal structures, in comparison with its state in the sister groups. This fraction is arbitrary and is not necessarily applicable to other studies; we used it here simply to have a consistent standard for extreme reduction. We used fossil taxa as sister groups for comparison in the Cetacea and Perissodactyla. We used extant taxa as sister groups in all other cases (Table 2).

The second criterion is that the structure has lost the specialized morphology that it exhibits in the sister groups. For example, a finger meets this criterion if its distal phalanx is shaped like an ovoid pebble in the taxon in question but has the form of an ungual (a claw-bearing phalanx) in the sister groups.

The third criterion is that the structure has lost a salient ancestral function. Although it may not be completely functionless, biologists consider it vestigial only if it has lost a major function (e.g., Darwin, 1859; Wilson, 1982; Prothero, 2007; Hall & Hallgrimsson, 2008; Bergstrom & Dugatkin, 2012). A limb satisfies this criterion, for example, if it is too reduced to serve as an organ of propulsion, whereas it is an organ of propulsion in the sister groups and therefore arguably in their common ancestor. Likewise, a pelvic girdle satisfies this criterion if it is too reduced to anchor a full limb and the muscles that operate it for propulsion, whereas in the sister groups it anchors a full limb and propulsive muscles. A pelvic girdle further satisfies this criterion if it is not connected to the vertebral column, because such a connection facilitates propulsion with the hindlimb by ensuring that each step propels the entire vertebral column (Kardong, 2008). A digit satisfies this criterion if it is too reduced for the functions of prehension or bodily support.

It is important to confirm that apparent vestigiality is characteristic of a species and not simply due to aberrance in a single specimen. Therefore, when we found reduced structures in taxa for which, to our knowledge, vestigial structures had not been previously documented, we examined more than one specimen per species when possible (Table 1). This also revealed individual variation, which is important because vestigial structures are often highly variable (Darwin, 1859; Omura, 1980; Conrad, 1982; Tague, 1997). For suprageneric taxa in which adequate previous descriptions of vestigial structures existed, we examined fewer specimens so as to spend a greater fraction of available time on previously undocumented or undescribed vestigial structures (Table 1). For the same reason, we also examined fewer specimens per species of suprageneric taxa lacking vestigial structures.

For this study we examined tail skeletons, clavicles, forelimb bones, pelvic girdles, and hindlimb bones. It is possible that vestigial skeletal structures are identifiable in other parts of the mammalian skeleton, e.g., the skull, parts of vertebrae, parts of the scapula, thyroid bones, cardiac bones, and the baculum. Such were not included in this study but may prove fruitful avenues for future research.

Tracing phylogenetic patterns

We used the cladogram in Fig. 1 (see caption for information sources) to trace phylogenetic patterns in vestigiality and loss of skeletal structures. Onto this cladogram we mapped the phylogenetic distribution of vestigiality and loss in postcranial skeletal elements, as shown in the figure. We then used this mapping, plus information from the fossil record as needed, to determine the phylogenetic points at which vestigiality or loss occurred for given skeletal elements. In this determination we used two assumptions. The first assumption is that if all members of a clade share a character state (e.g., vestigiality or loss of a skeletal structure), then that state arose in the clade’s common ancestor. For example, if the second toe is vestigial in all members of a clade, then it was vestigial in the clade’s common ancestor. The second assumption is that vestigialization and loss are not reversible. Therefore, if the second toe is lost in two clades but is present in a clade that is phylogenetically bracketed by the two clades, then the two clades lost the second toe independently instead of having inherited that loss from a common ancestor.

Missing data and parallel evolution cause challenges when character states (traits) are mapped onto phylogenies, making it difficult to distinguish convergences (in which two or more lineages gain the same character state) from reversals (in which members of a taxon revert to a previous character state). Software for phylogenetic analysis often resolves the problem by presenting two alternate solutions: one that interprets the phylogenetic pattern according to the principle of accelerated transformation (in which reversals are considered more likely than convergences) and one that interprets the phylogenetic pattern according to the principle of delayed transformation (in which convergences are considered more likely than reversals) (Maddison & Maddison, 1992). The two principles yield identical results when patterns of changes in character states are unambiguous. Our data set is sufficiently simple that changes in character states are unambiguous in most cases, obviating the need for software. However, the reader should note that because we assumed non-reversibility of loss, we used the principle of delayed transformation in the few cases that did exhibit ambiguity. For example, in the case of vestigialization of the clavicle, we used the principle of delayed transformation because of a problem with missing data, i.e., tiny clavicles that are present in the live animal are often missing or overlooked in disarticulated museum specimens (see the Clavicle section of the Results). For other bones, there were no problems with missing data. Similarly, disagreement in the literature regarding rodent phylogeny engendered ambiguity in the interpretation of the evolution of vestigialization of the hallux (the first toe) of rodents, and we used the principle of delayed transformation to interpret the results according to multiple possible phylogenies (see the Foot and toe section of the Results). For other mammalian taxa, there were no problems with phylogenetic disagreement.

Our employment of the assumption that losses are irreversible deserves further comment. There are exceptions to the general rule that vestigiality and loss are irreversible. In iguanodontian dinosaurs the ancestrally-vestigial fifth finger became elongated and useful for grasping (Senter, 2010), and atavistic limbs in aberrant cetacean and sirenian individuals show that loss is not completely irreversible (Abel, 1908; Andrews, 1929; Ogawa & Kamiya, 1957; Ohsumi, 1965). Lost body parts can be regained in the case of meristic traits in which the number of repeated parts is determined late in ontogeny (Galis, Arntzen & Lande, 2010). However, regaining of lost structures is rare (Goldberg & Igić, 2008; Galis, Arntzen & Lande, 2010), and there is no evidence that it has occurred within the Mammalia except in such atavistic cases.

Mammalian taxa without vestigial skeletal structures

There are a few major mammalian clades among whose extant representatives we did not find vestigial structures in the postcranial skeleton. These are the Monotremata (platypus and echidna); marsupial orders other than Dasyuromorphia and Peramelemorphia; and the eutherian orders Tubulidentata (aardvark), Proboscidea (elephants), Pholidota (pangolins), Eulipotyphla (shrews, moles, hedgehogs, and kin), and Scandentia (tree shrews).

Sternum

Ancestrally, the mammalian sternum consists of several segments called sternebrae, the first of which is called the manubrium (Fig. 2). The ribs articulate with the sternum via cartilaginous extensions called costal cartilages.

In most Odontoceti (toothed whales) most of the ribs maintain their connection with the sternum, and the sternum is unreduced (Fig. 3). An exception is Physeter catodon (sperm whale), in which the post-manubrial sternum has lost all but two sternebrae. They are reduced in relative size, but not enough to satisfy the first criterion for vestigiality (Fig. 3C).

In the Mysticeti (baleen whales) most of the ribs have lost the costal cartilages and are therefore no longer connected to the sternum. This enables the ribcage to collapse more than is possible in other mammals, so that a greater amount of air can be expelled from the lungs for deep diving. The sternum—which would prevent such ribcage collapse if it were fully expressed—is vestigial. Only the manubrium is retained (Figs. 3D–3F) (Howell, 1930).

In the Sirenia (manatees and dugongs) the number of segments in the sternum is also reduced. Reduced rib mobility in sirenians prevents ribcage collapse (Howell, 1930), so there is less selection pressure to maintain a large sternum. The sternum retains the manubrium and a second ossification that appears to be homologous to the rest of the sternum but is reduced in size and undivided into sternebrae (Howell, 1930). However, the size of the sirenian sternum is not reduced enough to satisfy the first criterion for vestigiality (Fig. 3G), and because it maintains its cartilaginous connection to several ribs it also fails to satisfy the third criterion.

Tail

Mammalian tails vary widely in length. Even short tails with a small number of vertebrae, such as those of many ungulates, perform important functions such as fly swatting and social signaling and therefore fail to satisfy the third criterion for vestigiality. It is therefore difficult to find examples of unambiguously vestigial tails in mammals.

The coccyx of apes and humans, a fused series of three to six vertebrae (Fig. 4), satisfies all three criteria for vestigiality. It does not protrude from the body’s surface and therefore cannot be used for social signaling, fly swatting, etc.

Clavicle

Ancestrally, the mammalian clavicle is a robust bone that articulates with the sternum and the acromion process of the scapula, bracing the forelimb against the axial skeleton (Fig. 2A). Most mammals retain this condition, but some have lost the clavicle altogether. This loss facilitates forward motion of the scapula, which increases stride length during running (Ewer, 1973; Hildebrand & Goslow, 2001). The clavicle is lost in the orders Perissodactyla (odd-toed hoofed mammals), Cetartiodactyla (even-toed hoofed mammals and whales), Sirenia (manatees and dugongs), Hyracoidea (hyraxes), and Proboscidea (elephants); most members of Carnivora (carnivores); and some rodents (Flower, 1870).

A vestigial clavicle is retained in two carnivoran families: Canidae (the dog family) and Felidae (the cat family). In both, only a short sliver ossifies (Figs. 5A–5C) within a ligament that connects the sternum to the acromion and represents the degenerate remainder of the ancestral clavicle, and in the Canidae the clavicle is often absent (Ewer, 1973). A similar situation is present in rabbits (Figs. 5D–5E) (Flower, 1870).

A vestigial clavicle is present in some rodents, such as the guinea pig (Cavia porcellus), in which it is embedded in muscle, has no connection with the sternum, and has a loose attachment of fibrous tissue to the capsule of the shoulder joint (Cooper & Schiller, 1975). We did not confirm the presence of a vestigial clavicle in other rodents, because a vestigial clavicle is difficult to identify in osteological specimens. Due to its tiny size, a vestigial clavicle is easily missed in a box of disarticulated bones, and it is often missing on articulated skeletons because it does not articulate with other bones. Radiography of rodents, to determine the taxonomic distribution of vestigial clavicles, was prevented by logistical constraints during the course of this study but would make an interesting study for future researchers.

The phylogenetic distribution of character states (Fig. 1) indicates that the clavicle was independently lost at least four times in the Mammalia: once in the common ancestor of the Hyracoidea, Sirenia, and Proboscidea; once in the Cetartiodactyla; once in the Hyaenidae; and once in the common ancestor of the Ursidae, Mustelidae, and Procyonidae. The clavicle became vestigial independently in at least three groups: Lagomorpha, Felidae, and Canidae.

Forearm

Ancestrally, the mammalian forearm contains two bones, the radius and ulna. Each articulates with the humerus proximally and the carpus distally and functions as a strut between the carpus and humerus (Figs. 2B and 2D). The joint between the humerus and ulna is the hinge of the elbow; the proximal end of the ulna is therefore never lost in mammals. The shaft of the ulna, however, is reduced to a vestigial state in some mammals. In such cases the radius is the only strut between the carpus and humerus. Such is the case in the Chiroptera (bats); most Macroscelidea (elephant shrews); and Equus (horses). In bats the ulnar shaft is reduced to a threadlike sliver (Figs. 6A–6C). Among elephant shrews, the ulna is unreduced in the genus Rhynchocyon, but its shaft is vestigial in other genera, tapering to a point about halfway down the length of the radius (Evans, 1942) (Fig. 6D). In Equus the shaft of the ulna tapers to a point and does not reach the carpus (Nickel et al., 1986) (Figs. 6E–6G).

The ulnar shaft is reduced in the Camelidae (camels and kin) and ruminants. This reduction is extreme in the Camelidae and the ruminant families Giraffidae (giraffe and okapi), Cervidae (deer), Antilocapridae (pronghorn), and Bovidae (cattle, sheep, goats, and antelope). In these families the ulnar shaft is present for its full length but is reduced in transverse diameter to one-fourth or less the transverse diameter of the radial shaft, and the shafts of the two bones are co-ossified (Figs. 6I–6L). Because the ulnar shaft is present for its full length, it retains its ancestral function as a strut between the carpus and humerus and therefore does not satisfy the third criterion for vestigiality.

The phylogenetic distribution of character states (Fig. 1) indicates that the ulnar shaft became vestigial independently in elephant shrews, bats, and horses.

Hand and fingers

Ancestrally, the mammalian hand has five digits with two phalanges in the thumb and three phalanges in each other finger, and a metacarpus in which all five metacarpals are of similar diameter (Fig. 2B). The metacarpals function as struts between the phalanges and the carpus. In numerous mammalian taxa one or more fingers have become vestigial (Fig. 7). In numerous others, one or more fingers are lost and the associated metacarpals are reduced to a vestigial state. We did not find examples of identifiably vestigial carpal bones.

In the marsupial Chaeropus ecaudatus (the recently-extinct pig-footed bandicoot) the first and fifth fingers and their metacarpals are lost, and the fourth finger and its metacarpal are vestigial. Together they are less than one-third the full length of the metacarpus (Flower, 1870).

In the Hyracoidea (hyraxes) the thumb is vestigial and is not externally visible. Its metacarpal is highly reduced in both length and diameter. It bears a single, miniscule phalanx (Flower, 1870) (Figs. 7A and 7B).

In the Sirenia the thumb is vestigial and usually retains only one phalanx. In Dugong (dugongs) the phalanx is reduced to a pebble-like nub (Fig. 7C). In Trichechus (manatees) the phalanx is relatively larger than in dugongs but is very reduced in length and diameter in comparison to the proximal phalanges of the other digits (Fig. 7D).

Vestigial fingers are abundant in the Xenarthra (armadillos, anteaters, and sloths). The fifth finger is vestigial in Priodontes maximus (giant armadillo). Its metacarpal is tiny, and it retains only one phalanx, which is reduced to a nub (Humphry, 1870) (Fig. 7E). In Tolypeutes matacus (southern three-banded armadillo) the first and fifth fingers are lost. Their metacarpals are vestigial; each is reduced to a tiny, pebble-like, transversely flattened bone (Fig. 7F). In Dasypus novemcinctus (nine-banded armadillo) the fifth finger is vestigial; it is present only as a single, miniscule, grain-shaped phalanx. In Myrmecophaga tridactyla (giant anteater) the fifth finger retains two phalanges but has lost the third, and the second is reduced to a nub and can therefore be considered vestigial (Fig. 7G). In Tamandua (lesser anteaters) the fifth finger is vestigial. It is reduced to a single phalanx that is but a nub (Fig. 7H). In Cyclopes didactylus (silky anteater), the phalanges of the thumb and fifth finger are lost, as is the fifth metacarpal, and the first metacarpal is vestigial. It is reduced to a short, subquadrangular, transversely flattened bone. In Bradypus (three-toed sloths) the first and fifth fingers are lost. The corresponding metacarpals are reduced to a vestigial state, remaining only as small, hook-shaped bones (Humphry, 1870) that may be coossified with the neighboring metacarpals (Fig. 7I). In Choloepus (two-toed sloths) the first, fourth, and fifth fingers are lost, as is the fifth metacarpal. The first and fourth metacarpals are vestigial; each is approximately half the length of the neighboring metacarpal and is very reduced in diameter (Humphry, 1870) (Fig. 7J).

In Ateles (spider monkeys), Brachyteles (woolly monkeys), and Colobus (colobus) the thumb is vestigial. Its metacarpal is reduced in diameter and length (Fig. 7K). Some specimens retain a single phalanx that is reduced to a nub, while in others the thumb lacks phalanges (Tague, 1997; Tague, 2002).

In Perodicticus potto (potto) and Arctocebus (angwantibos), the African members of the primate family Lorisidae, the second finger is vestigial, although the thumb is fully expressed. The second metacarpal is reduced in length, its proximal phalanx is reduced in length and diameter, its middle phalanx is reduced to a nub, and its distal phalanx is lost (Fig. 7L). In the fleshed-out animal, the second finger is reduced to a short stump in Arctocebus. In P. potto it is further reduced and exists as a mere lump at the edge of the palm.

Reduction of the thumb (Fig. 8) is common in rodents. In many cases it is miniscule in comparison to the other fingers (Figs. 8A and 8B). This is common in the Muroidea (rats, mice, and kin), Dipodidae (jerboas, jumping mice, and kin), Gliridae (dormice), Heteromyidae (kangaroo rats and kin), Octodontidae (degus and kin), and Sciuridae (squirrels) (Kingdon, 1974; Kingdon, 1997; Garbutt, 1999; and P Senter, pers. obs., 2012–2014). It is also the case in Chinchilla (chinchillas). In many such cases, the claw on the thumb is not pointed at the tip but is flattened into a shape that resembles a primate’s nail (Figs. 8A–8C). Because the terminal phalanx is therefore an ungual, the digit does not satisfy the second criterion for vestigiality. Also, in such cases the tip of the thumb is used in opposition to the second finger, to grasp objects, as P Senter has personally observed (pers. obs. 2014, 2015) in the chinchilla (Chinchilla lanigera) and the degu (Octodon degus); the thumb therefore does not satisfy the third criterion for vestigiality.

In several other rodent families are a plethora of cases in which the thumb is lost or is so reduced that it does not protrude externally and satisfies all three criteria for vestigiality. In the squirrel genus Sciurus is an ambiguous case: the thumb is tipped with a claw, but only the claw protrudes externally. Future studies will be necessary to determine what function, if any, this thumb claw serves.

The thumb is lost in the Caviidae (cavies, capybaras, and kin), and the metacarpal is vestigial. It remains only as a tiny, ovoid bone no larger than a distal carpal and usually much smaller (Figs. 7M–7O). In the Chinchillidae the thumb is lost in Lagidium (mountain viscachas) and Lagostomus trichodactylus (plains viscacha). In the latter two the first metacarpal is vestigial; it is a tiny, transversely flattened ovoid (Fig. 7P). In Erethizon dorsatum (North American porcupine) and Coendou (prehensile-tailed porcupines) the thumb is vestigial. Its metacarpal is highly reduced, and it retains only one highly reduced phalanx with a variable shape (Fig. 7Q).

The genus Hystrix, a member of Hystricidae (Old World porcupines) is unusual in that different species of one genus exhibit different degrees of thumb reduction (Fig. 9). In H. indica (the Indian porcupine) the thumb is fully expressed and robust. In H. sumatrae (the Sumatran porcupine), H. javanica (the Sunda porcupine) and H. cristata (the crested porcupine) the thumb is reduced relative to its state in H. indica, but it is not vestigial. In H. crassispinis (the thick-skinned porcupine) the thumb is vestigial. It retains only one phalanx, which is no larger than the highly reduced metacarpal that is typical for the genus. In H. brachyura (the Malayan porcupine) the thumb is lost, leaving only its reduced metacarpal.

In the Canidae and Felidae the thumb (Figs. 8D and 8E) is called the dewclaw. It is shorter than the other digits and does not contact the ground. Some authors consider it vestigial (e.g., Eldredge, 2007), but it is not reduced enough to satisfy the first criterion for vestigiality. Also, it retains a claw and therefore does not satisfy the second criterion. Nor does it satisfy the third criterion, because it retains the typical function of a finger with a sharp, curved ungual and claw: prehension. Its use is an important part of prey capture in felids (Londei, 2000), and we have personally observed that domestic dogs use the dewclaw to snag and maintain a grip on objects. Its shortening is therefore not the reduction of an unused organ. We suggest that the functional advantage of the shortening is to keep the claw sharp by preventing wear that would result from contact with the ground. Indeed, according to P Senter (pers. obs., 2005–2006), the canid dewclaw has a sharper, less worn tip than the other claws and is more effective at puncturing and maintaining prehension. An analogy can therefore be made between the dewclaw and the second toe of dromaeosaurid and troodontid dinosaurs, which was also held clear of the ground (Senter, 2009), was used to puncture (Fowler et al., 2011), and had a function that differed from those of the other digits, as shown by its difference in length and in claw curvature.

A vestigial dewclaw is present in Hyaena (striped hyena and brown hyena) and Crocuta (spotted hyena). In both, the metacarpal is reduced to a small block, the distal phalanx is lost, and the proximal phalanx is reduced. The proximal phalanx is a small, shapeless lump in Hyaena and a tiny spike in Crocuta (Figs. 7R–7T). The hyaenid Proteles cristata (aardwolf) has an unreduced dewclaw (Fig. 8F).

In the Perissodactyla the first finger and its metacarpal are lost. In Equus (horses) the second, fourth, and fifth fingers are also lost, as are the first and fifth metacarpals. The second and fourth metacarpals are vestigial. They remain as thin splints that taper to a point without reaching the distal end of the metacarpus (Fig. 7U).

In the Cetartidoactyla the first finger is lost, and in ruminant cetartiodactyls the second and fifth fingers are reduced (Tragulidae [chevrotains] and Cervidae [deer]) or lost (Antilocapridae [pronghorn], and Giraffidae [giraffes and okapi]). The reduced second and fifth fingers of deer are called dewclaws (McBride, 2001; Elbroch, 2003), and some authors consider them vestigial (e.g., McBride, 2001). However, they bear hooves and therefore do not meet the second criterion for vestigiality. Nor do they meet the third criterion for vestigiality, because during fast locomotion they make sufficient contact with the ground (Elbroch, 2003) to exhibit a major function of digits: bodily support.

In the Cervidae (deer) the second and fifth fingers are not vestigial according to our criteria. However, the second and fifth metacarpals of Cervinae (Old World deer) are vestigial; they are reduced to proximal splints that resemble the vestigial metacarpals of horses (Geist, 1998). In the second and fifth metacarpals of Capreolinae (New World deer) the proximal end is lost, leaving only the distal end, which articulates with the proximal phalanx. The shaft of each of these metacarpals is vestigial and is reduced to a small splint (Geist, 1998) (Fig. 7V).

In the Bovidae (cattle, antelope, sheep, and goats) the second and fifth fingers have only one or two phalanges apiece, and these do not articulate with the rest of the skeleton. However, they bear hooves and therefore do not meet the second criterion for vestigiality. When present, the fifth metacarpal is vestigial. It remains only as a tiny, proximal splint (Nickel et al., 1986).

Vestigial fingers are common in the Odontoceti (toothed whales). In odontocetes other than Physeteroidea (sperm whales and kin) the thumb is vestigial (Figs. 7W–7Y). It typically either retains only a single, pebble-like phalanx, or just a metacarpal (Van Beneden & Gervais, 1879; Cooper et al., 2007). In a few species there are some individuals that have two thumb phalanges, but their conspecifics have only one phalanx or none (Cooper et al., 2007). In several odontocete clades, the fifth finger is also vestigial, retaining one or two pebble-like phalanges or just a reduced metacarpal. Such is the case in Inia (New World river dolphins), Phocoena (porpoises), and the delphinid subfamilies Delphininae (dolphins) and Globicephalinae (pilot whales and kin) (Van Beneden & Gervais, 1879; Cooper et al., 2007). In the latter two clades the fourth finger is also reduced enough to consider vestigial. It usually retains only two phalanges, and only the proximal phalanx has the typical form of an odontocete phalanx instead of being reduced to a tiny, pebble-shaped bone (Van Beneden & Gervais, 1879) (Figs. 7X and 7Y).

In addition to the first finger and its metacarpal, the fifth finger and its metacarpal are also lost in the Rhinocerotidae (rhinoceroses). Flower (1870) identified a small bone in the wrist of Dicerorhinus sumatrensis (Sumatran rhinoceros) as a vestigial fifth metacarpal, but it is more likely a sesamoid. It does not articulate with the lateral surface of the fourth metacarpal or the lateral surface of the hamate carpal, as would be expected of a fifth metacarpal. Rather, it is on the palmar surface of the hamate. We found the homologous bone in the wrist of a specimen of Rhinoceros sondaicus (Javan rhinoceros) and in photos, supplied by the Museum of Comparative Zoology, of articulated hands of two specimens of R. unicornis (Indian rhinoceros). The bone is small and rounded, is on the palmar side of the hamate, and does not articulate with the lateral surface of the hamate or the fourth metacarpal. These are characteristics that are consistent with a sesamoid but not with a vestigial fifth metacarpal.

The phylogenetic distribution of character states (Fig. 1) indicates that the thumb independently became vestigial at least seven times in the Mammalia: in the Sirenia, Hyracoidea, Lagidium + Lagostomus, Erethizontidae, Hystrix crassipes, Crocuta + Hyaena, and Odontoceti. In the preceding sentence and below, the phrase “at least” expresses uncertainty as to whether a skeletal structure became vestigial before its loss in the taxa that have lost it. The thumb was independently lost eight times: in Chaeropus, Tubulidentata, Ateles + Brachyteles, Colobus, Caviidae, Perissodactyla, Camelidae, and Ruminantia. The second finger became vestigial once: in Arctocebus + Perodicticus. It was lost independently four times: in the Antilocapridae, Giraffidae, Camelidae, and Equidae. The fourth finger became vestigial in Chaeropus, and is not vestigial in any other Recent mammal. It was independently lost twice: in Choloepus and Equus. The fifth finger became vestigial at least seven times: in Priodontes, Dasypus, Tamandua, Globicephalinae, Delphininae, Phocoena, and Inia. It was independently lost ten times: in Chaeropus, Cyclopes, Tolypeutes, Equus, Rhinocerotidae, Camelidae, Antilocapridae, Giraffidae, Bradypus, and Choloepus. Although Choloepus is the closest living relative to Bradypus, it must have lost its fifth finger independently, because the two genera are in different families, and some extinct members of the Megalonychidae (which includes Choloepus) retained a vestige of the fifth finger (P Senter, pers. obs., 2008).

The first metacarpal became vestigial independently at least eight times in the Mammalia: in Ateles + Brachyteles, Colobus, Tolypeutes, Cyclopes, Caviidae, Equus, Bradypus, and Choloepus (some extinct megalonychids retained a fully expressed first metacarpal). The first metacarpal was independently lost four times: in Tubulidentata, Perissodactyla, Camelidae, and Ruminantia. The second metacarpal became vestigial independently at least twice: in Equus and Cervidae. It was lost independently lost four times: in Camelidae, Antilocapridae, Giraffidae, and Bovidae. The fourth metacarpal became vestigial independently three times: in Chaeropus, Choloepus, and Equus. The fifth metacarpal became vestigial independently at least three times: in Tolypeutues, Bradypus, and Bovidae + Cervidae. It was independently lost eight times: in Chaeropus, Choloepus, Cyclopes, Equus, Rhinocerotidae, Camelidae, Antilocapridae, and Giraffidae.

Pelvic girdle

Ancestrally, the mammalian pelvic girdle consists of three bones: the ilium, ischium, and pubis (Figs. 2E–2G). All three bones contribute to the acetabulum (hip socket). The ilium is attached to the vertebral column, and via this attachment the hindlimb propels the entire vertebral column during locomotion. The vertebrae that contact the ilium are fused to a few more vertebrae posterior to them. Together this series of fused vertebrae is called the sacrum (Figs. 2E and 2F). The left and right pubes are ventral in location and meet in the midline at a symphysis. The left and right ischia extend posteriorly and do not meet each other. The pubis and ischium surround an opening called the obturator foramen. In most mammals, by adulthood the three bones of the pelvic girdle have fused together to form a single bone called the coxal bone or innominate.

The pelvic girdle is vestigial in Sirenia. It is extremely reduced in size and has lost contact with the vertebral column (Figs. 10A and 10B). Abel’s (1908) comparison of the pelvic girdles of extant and fossil Sirenia shows that the pelvic girdle of the dugong (Dugong dugon) retains the acetabular region, which is where the ilium, ischium, and pubis converge. Therefore, none of the three bones is lost. In the dugong the pubis is highly reduced, and the pelvic girdle consists mainly of the ilium and ischium (Abel, 1908). According to Abel (1908), in Trichechus (manatees), the pubis and ilium are both lost or reduced almost unto loss, leaving only the ischium, which retains its original shape.

The pelvic girdle is vestigial in Cetacea (whales). It is extremely reduced in size, has lost contact with the vertebral column, and has lost a distinct acetabulum (Figs. 10C–10F). In Mysticeti the pelvic girdle is usually shaped like a very shallow “V,” with an anterior process meeting a posterior process at an obtuse angle, and with a third, much shorter process projecting from the point of union between the two main processes, slightly extending the point of the “V” (Howell, 1930; Lönneberg, 1910; Arvy, 1976) (Figs. 10E and 10F). In Odontoceti the pelvic girdle is typically not V-shaped but is a straight or slightly curved rod (Howell, 1930; Lönneberg, 1910; Arvy, 1976) (Figs. 10C and 10D).

Previous authors have disagreed as to whether the cetacean pelvic girdle retains the ischium alone (Struthers, 1881; Simões-Lopes & Gutstein, 2004), the ischium and ilium (Abel, 1908; Lönneberg, 1910; Schulte, 1916; Howell, 1930), or the ischium, ilium, and pubis (Gol’din, 2014). Its connections with soft anatomy indicate that much of the girdle is the ischium (Struthers, 1881; Simões-Lopes & Gutstein, 2004). However, the presence of the acetabulum in mysticetes (Struthers, 1881; Fordyce et al., 2000; Gol’din, 2014) suggests that all three bones are present, at least in reduced form, in the mysticete pelvic girdle. No part of the odontocete pelvic girdle bears evidence that it represents the pubis or ilium (Simões-Lopes & Gutstein, 2004), and its simple rodlike shape suggests that one tine (containing the vestigial pubis and ilium) of the mysticete “V” has been lost, leaving only the ischium. In the reduced pelvic girdles of Basilosaurus isis and Chrysocetus healyorum, members of the extinct whale family Basilosauridae from the Eocene Epoch, the ischium and ilium are more highly reduced than the pubis, which meets its counterpart at a midline symphysis (Gingerich, Smith & Simons, 1990; Uhen & Gingerich, 2001). Apparently, then, in the evolution of the cetacean pelvis, reduction of the ilium and ischium occurred first, followed by the reduction of the pubis and ilium and subsequently their loss in odontocetes. The cetacean pelvic girdle is certainly not an abdominal bone such as marsupials have, as one author has suggested (Arvy, 1976; Arvy, 1979), because it retains a muscular or ligamentous connection to the femur in specimens that retain vestigial hindlimbs (Struthers, 1881; Hosokawa, 1951; Ogawa & Kamiya, 1957), and because the soft tissues that attach to it are those that typically attach to an ischium (Struthers, 1881; Simões-Lopes & Gutstein, 2004).

The phylogenetic distribution of character states (Fig. 1) shows that the pelvis became vestigial twice independently: in the Sirenia and Cetacea.

Femur, tibia and fibula

Ancestrally, the mammalian hindlimb includes a single bone in the thigh (the femur) and two in the shank or crus (the tibia and fibula, with the fibula the more lateral of the two). The proximal end of the tibia articulates with the femur, and the proximal end of the fibula articulates with a lateral shelf of the tibia. Both the tibia and the fibula articulate distally with the tarsus (Figs. 2C and 2D). The fibula functions as a strut between the tarsus and the proximal tibia.

The shaft of the fibula is fused to the tibia in many small mammals, including Macroscelidea (elephant shrews), Tarsiidae (tarsiers), Eulipotyphla (shrews, moles, hedgehogs, and kin), Lagomorpha (rabbits and pikas), and many rodents. In some cases only the distal half of the fibular shaft is fused to the tibia, but in others only a small, proximal portion of the fibula is free of the tibia (Figs. 8G–8I). This yields the illusion that most of the fibula has been lost, in which case the remaining portion could be considered vestigial. However, close inspection shows that even in cases with extreme amounts of fusion, the fibula is present for its full length and is a strut between the tarsus and the proximal tibia. It therefore does not satisfy the first or third criterion for vestigiality.

An unambiguously vestigial fibula is present in three extant ungulate taxa: Camelidae (camels and kin), Pecora (ruminants other than chevrotains), and Equus (horses). In Camelidae the shaft and proximal end of the fibula are lost. All that remains is a distal vestige: a block of bone called the malleolar bone or os malleolare, which fits into a cleft in the tibia and articulates with the two proximal tarsal bones (Flower, 1870). In Pecora the shaft of the fibula is replaced by a ligament, and its proximal and distal extremities remain as vestiges. The proximal vestige, all that remains of the head of the fibula, is a small spike that is fused to the lateral condyle of the tibia. The distal vestige is a malleolar bone resembling that of camelids (Nickel et al., 1986) (Fig. 11). In Equus the distal half of the fibular shaft is lost. The proximal vestige of the tibia includes the head and a thin rod that represents the remainder of the fibular shaft. The distal vestige is similar to that of camelids and pecorans but is fused to the tibia (Nickel et al., 1986).

Hindlimb buds appear in the embryos of both mysticete and odontocete whales (Ogawa, 1953; Bejder & Hall, 2002). In odontocetes the hindlimb buds regress and limbs are not formed, except in occasional atavistic cases (Ogawa & Kamiya, 1957; Ohsumi, 1965). In mysticetes, vestigial hindlimbs are often present, with much individual variation in morphology (Struthers, 1881; Hosokawa, 1951; Gol’din, 2014). Typically, the bowhead whale (Balaena mysticetus) retains the femur and tibia; the humpback (Megaptera novaeangliae) and fin whale (Balaenoptera physalus) retain only the femur; the minke whale (Balaenoptera bonaerensis) retains the femur in about one-third of individuals; and the hindlimb is absent in the sei whale (Balaenoptera borealis) (Struthers, 1881; Hosokawa, 1951; Omura, 1980). Occasional atavistic specimens retain the more distal elements. For example, Andrews (1929) described a humpback whale with an ossified tibia and metatarsal and a cartilaginous femur and tarsus.

In the extant Sirenia there is usually no hindlimb. However, an example of an atavistic, diminutive femur has been described in an example of Trichechus manatus (West Indian manatee) (Abel, 1908). It is tiny enough to consider vestigial.

The phylogenetic distribution of character states (Fig. 1) shows that the femur was independently lost three times: in the Sirenia, Odontoceti, and Balaenoptera borealis. The crus was independently lost three times, in the Sirenia, Odontoceti, and Balaenopteridae. The fibula became vestigial independently at least three times: in the Camelidae, Pecora, and Equus.

Foot and toes

Ancestrally, the mammalian foot has five digits with two phalanges in the first toe and three phalanges in each other toe, and a metatarsus in which all five metatarsals are of similar diameter (Fig. 2C). The metatarsals function as struts between the phalanges and the tarsus. In numerous mammalian taxa one or more toes have become vestigial (Fig. 12). In numerous others, one or more toes are lost and the associated metatarsals are reduced to a vestigial state. We did not find examples of identifiably vestigial tarsal bones.

In marsupials of the orders Peramelemorphia (bandicoots and bilbies) and Diprotodontia (wombats, possums, kangaroos, and kin), the second and third toes are syndactylous. The skeletons of the two digits share a common sheath of soft tissue, so that the two toes are separate only at the last phalanx, which bears the claw. The two toes are therefore functionally a single toe with two claws. The metatarsals and phalanges of the two toes are usually half or less the diameter of those of the fourth toe, and their reduction in size makes them appear vestigial (Figs. 8K and 8L). However, the two digits do not meet the second criterion for vestigiality, because they bear claws. They also fail to meet the third criterion for vestigiality, because together they functionally constitute a single toe that is used as a toe.

Chaeropus ecaudatus (pig-footed bandicoot) is an exception to the above rule. Its second, third, and fifth toes fit all three criteria for vestigiality (Flower, 1870). The foot of C. ecaudatus is functionally monodactyl and uses only the fourth toe for support.

The first toe is reduced in some members of the marsupial family Dasyuridae. It is reduced enough to fit all three criteria for vestigiality in Antechinomys laniger (the kultarr) (Szalay, 1994), Dasyuroides byrnei (the kowari), and Sarcophilus (Tasmanian devils). In all three cases it is represented only by a very short metatarsal and a single phalanx shaped like a small spike (Fig. 12A).

In Thylacinus cynocephalus, the recently-extinct thylacine or Tasmanian wolf, the first toe is lost, and the first metatarsal is vestigial. It is a flattened oval, not much longer than the transverse width of one of the other metatarsals (Fig. 12B).

In the Hyracoidea the first and fifth toes are lost, as is the first metatarsal. The fifth metatarsal is vestigial. It is tiny and transversely flattened (Fig. 12C).

In Bradypus (three-toed sloths) and Choloepus (two-toed sloths) the first and fifth toes are lost, and the first and fifth metatarsals are reduced to a vestigial state. In Bradypus these two metatarsals remain only as small, hook-shaped bones (Humphry, 1870) that are coossified with the neighboring metatarsals (Fig. 12D). In Choloepus the first and fifth metatarsal are each little more than half the length of the neighboring metatarsal (Flower, 1870); these vestigial metatarsals are transversely flattened and lack a distal articulating surface (Fig. 12E).

In the Lagomorpha the first toe is lost. Its metatarsal is vestigial in Leporidae (rabbits) (Fig. 12F) and absent in Ochotonidae (pikas).

In the Caviidae the first and fifth toes are lost, and their metatarsals are vestigial (Figs. 12G–12I). Previous authors (e.g., Mivart & Murie, 1866; Cooper & Schiller, 1975) have identified each of these two vestiges as a sesamoid, which is a bone that ossifies inside a tendon or ligament. However, several lines of evidence show that these two bones are not sesamoids but are the first and fifth metatarsals. Most importantly, as shown in Cooper & Schiller’s (1975) illustrations, neither of the two bones is within a tendon or ligament. Secondly, the two bones are in the locations of the proximal ends of the first and fifth metatarsals of other rodents, and they exhibit the articulations with neighboring bones that the first and fifth metatarsals of other rodents do. Furthermore, the vestigial first metatarsal strongly resembles its counterpart in other rodents, and the proximal end of the vestigial fifth metatarsal serves as the insertion of the peroneus brevis muscle (Cooper & Schiller, 1975), which inserts on the proximal end of the fifth metatarsal in other mammals (Parsons, 1898; Meehan, 1992; Marieb, Mallatt & Wilhelm, 2005; Budras et al., 2007), including other rodents (Parsons, 1898; Hebel & Stromberg, 1986). In all caviid genera, the vestigial first and fifth metatarsals are transversely flattened. The first metatarsal is a proximodistally elongate ovoid, and the fifth is near-circular in shape in lateral view.

In the Dasyproctidae (agoutis and acouchis) the first and fifth toes are lost, as is the fifth metatarsal. The first metatarsal is vestigial (Fig. 12J) and resembles that in the Caviidae and Chinchillidae.

In the Chinchillidae (chinchillas and viscachas) the first toe is lost and its metatarsal is vestigial. It is a transversely flattened, proximodistally elongate ovoid (Figs. 12K and 12L). Of the three genera in this family, the fifth toe is present and fully expressed in two: Chinchilla (chinchillas) and Lagidium (mountain viscachas). In the remaining species, Lagostomus trichodactylus (the plains viscacha), the fifth toe is absent and its metatarsal is vestigial; it resembles the first metatarsal but is proximodistally shorter (Fig. 12L).

In the Pedetidae (springhares) the first toe is lost, and its metatarsal is vestigial. As in other rodents with a vestigial first metatarsal, it is transversely flattened and proximodistally elongate, and its distal margin is rounded. However, unlike the case in other rodents, it tapers to a point proximally (Fig. 12M).

In the Thryonomyidae (cane rats) the first toe is vestigial. It retains only one phalanx, which is less than half the size of the highly reduced metatarsal and is shaped like a small spike (Fig. 12N).

In Dipodomys (kangaroo rats) and Jaculus (African jerboas), members of the Dipodidae, the first toe is lost, and its metatarsal is vestigial; it is reduced to a tiny, proximal sliver (Howell, 1932). In other members of the Dipodidae all five toes are fully expressed.

Most members of the Carnivora retain all five toes. Exceptions are the families Canidae, Felidae, and Hyaenidae. In the Canidae the first toe is vestigial. Its metatarsal is tiny, and its toe has only one phalanx, which is reduced to a small spike or nubbin (Figs. 12O–12Q). In the Felidae the first toe is lost and its metatarsal is vestigial (Flower, 1870) (Figs. 12R and 12S). In the Hyaenidae, as in the Canidae, the first toe is vestigial. Its metatarsal is tiny, and its toe has only one phalanx, which is reduced to a small spike or nubbin (Figs. 12T–12V).

In peccaries (Tayassuidae) the fifth toe is lost, and its metatarsal is retained but vestigial. It is a transversely flattened, proximal splint that is shorter than the other metatarsals (Figs. 12W and 12X).

The second and fifth toes are reduced in the Tragulidae (chevrotains), Bovidae (cattle, antelope, sheep, and goats), and Cervidae (deer), but they do not meet our criteria for vestigiality, because they bear hooves (Fig. 8J). The second and fifth metatarsals are fully expressed in the Tragulidae. It was once thought that these two metatarsals are absent in other ruminants (Flower, 1870), but they are often present and fused to the rest of the metatarsus. When present, each remains only as a proximal sliver that fits the criteria for vestigiality. Vestigial second metatarsals are common in the Bovidae (cattle and kin), Capreolinae (New World deer), and Giraffidae (giraffe and okapi); they are uncommon in Cervinae (Old World deer) and unknown in pronghorn (Antilocapra americana). Vestigial fifth metatarsals are common in the Bovidae, Capreolinae, and Cervinae; they are absent in giraffe (Giraffa camelopardalis) but present in about 40% of okapi (Okapia johnstoni) and pronghorn specimens (Silvia, Hamilton & Silvia, 2014).

In Equus the first and fifth metatarsals are lost, as are the phalanges of all but the third toe. The second and fourth metatarsals are vestigial (Figs. 12Y and 12Z). They remain as narrow splints that taper to a point distally without reaching the distal end of the third metatarsal (Nickel et al., 1986).

In the Rhinocerotidae (rhinos) and Tapiridae (tapirs) the first toe is lost and the first metatarsal is vestigial. In tapirs the first metatarsal is a small, blocky bone, almost cube-shaped in some specimens, at the tip of the ectocuneiform (the tarsal bone proximal to it). It is medially displaced onto the palmar surface of the foot (Fig. 12Z’). In rhinos the first metatarsal is a tiny nub that is fused to the end of the ectocuneiform (Radinsky, 1963).

The entire foot is lost in the Sirenia and Cetacea. For other extant mammal taxa, the phylogenetic distribution of character states (Fig. 1) indicates that the first toe became vestigial independently at least five times: in the Dasyuridae, Thryonomyidae, Canidae, Hyaenidae, and Rhinocerotidae + Tapiridae. It was independently lost eight times in mammals other than rodents: in Chaeropus, Thylacinus, Hyracoidea, Macroscelidea, sloths, Lagomorpha, Felidae, and Perissodactyla + Cetartiodactyla. Uncertainty in rodent phylogeny makes it difficult to tell how many times the fist toe was lost in rodents. If Dipodomys and Zapus are not sister genera, and if Chinchillidae and Caviidae + Dasyproctidae are not sister clades, then the toe was lost five times. If those two pairs of possible sister taxa are indeed pairs of sister taxa, then the toe was lost independently at least three times in the Rodentia. The second toe became vestigial independently at least twice: in Chaeropus and the Cervidae. It was lost independently four times: in the Camelidae, Antilocapridae, Giraffidae, and Bovidae. The third toe is vestigial in only one Recent mammalian taxon: Chaeropus. Among extant mammals, the fourth toe has been lost only in Equus and is not vestigial in any taxon. The fifth toe became vestigial at least twice: in Chaeropus and the Cervidae. It was independently lost ten times: in the Hyracoidea, sloths, Caviidae + Dasyproctidae, Lagostomus, Perissodactyla, Camelidae, Tayassuidae, Antilocapridae, Giraffidae, and Bovidae.