General anatomy of the earliest mammals

While the physiology and locomotory behavior of dinosaurs are well-studied and widely discussed (e.g., Currie and Padian 1997; Farlow and Brett-Surman 1997; Klein et al. 2011), knowledge of mammals from the Mesozoic era is still limited, although increasing in recent years (Martin 2018). The most intriguing questions are how the earliest Mesozoic mammals looked like (Fig. 1A), how their bodies worked, and how they differed from non-mammalian synapsids and modern mammals. The Mammalia (Fig. 1B) comprises the last common ancestor of living mammals (Monotremata, Marsupialia, and Placentalia) plus all of its descendants (which includes also extinct clades). The more inclusive taxon Mammaliaformes includes the stem lineage of Mammalia; mammaliaforms have at least an incipient secondary jaw joint (between dentary and squamosal) besides the primary jaw joint constituted by articular and quadrate (Luo 2007).

Fig. 1
figure 1

A Reconstruction of the early mammaliaform Morganucodon (modified from Rowe 2023; with permission of the author). Characters as listed in the text. B Phylogeny and geological abundance of Mesozoic mammals in relation to (in bold:) the major extant mammalian groups (modified from Luo 2007; Wiliamson 2014; Martin 2018). Node a, the Late Triassic-Early Jurassic diversification of mammaliaform stem clades; node b, diversification of docodontans and splits of several extinct groups in Mammalia (green and yellow); node c: the Late Jurassic diversification of eutriconodontans, multituberculates, and cladotherians; node d: Early Cretaceous origin of character-based monotremes; node e: origin of stem-based metatherians (including marsupials); node f: origin of stem-based eutherians (including placentals). C Schematic side view on the dentition of a raptor (in this case a bear). The lever arm of the fangs is about two times longer than that of the carnassials, therefore the force on the fangs is ca. ½ of that on the carnassials. D Schematic anterior view on the same dentition. The fangs are close to each other, the external distance is \(\nicefrac{1}{5}\) of the distance between the adducting muscles (not drawn). Therefore, the muscles of both sides can exert nearly their full force when biting, while the bite force on one carnassial is lower in the case of a bear, and still lower in the pear-shaped tooth rows of a dog or cat (three-cornered hatched field in E. Silhouettes from Phylopic: https://www.phylopic.org/permalinks/f546e56edca97bc0c5fede6d01b1daaa9b987cc228fdf06097142791ff6043d8

Full size image

According to this definition, the oldest currently known Mammaliaformes (Fig. 1B) are the Late Triassic (Rhaetian, 206 myr) Morganucodon (Kühne 1949) and Early Jurassic (Sinemurian, 193 myr) Sinoconodon (Kermack et al. 1973, 1981; Crompton and Luo 1993). Adelobasileus (Lucas and Hunt 1990) from the Late Triassic (Carnian; about 227 myr; Gradstein et al. 2020) of Texas is even somewhat older but more fragmentary (Lucas and Hunt 1990; Lucas and Luo 1993; Gradstein et al. 2020).

The oldest mammaliaforms have several traits in common, namely: 1. multi-cusped teeth, often with several dental roots, which increase in size behind the pointed fang teeth (i.e., the canines), of which only two are developed in the upper and lower jaws (Teaford et al. 2000); 2. a simple jaw joint between dental and squamosal (Anthwal and Tucker 2023); 3. a small body size (i.e., body mass estimate for the largest known Mesozoic mammal Repenomanus giganticus from the Early Cretaceous of China is 12–14 kg; Hu et al. 2005; Cavanaugh 2021); 4. the pelvic girdle constantly consisted of a paired ilium, a paired ischium, and a paired pubis - the halves of which were in close contact in a “pseudarthrosis” along the midline (Bishop and Pierce 2024a); 5. an increasing reduction of neck ribs (Arnold 2021) and the presence of a series of rib-less lumbar vertebrae behind the rib-bearing vertebrae of the thorax (Jones et al. 2018); 6. the presence of a long tail, containing little mass; and 7. parasagittal movements of the extremities (Fig. 1A) (Kemp 2005; Luo 2007; Preuschoft et al. 2022). Finally, in therian mammals, 8. the extended ventral bony plates of the shoulder and pelvic girdles are absent (e.g., Brocklehurst et al. 2022; Bishop and Pierce 2024b), and 9. the scapula is the most strongly developed element of the shoulder girdle (e.g., Bishop et al. 2022). The latter two characters do not apply to living monotremes, which are the sister group of therian mammals, and which still bear a plate-shaped interclavicular and large pro- and metacoracoids (Meng et al. 2017).

In this essay, we discuss the selective value of these characteristics from a strictly functional perspective (in the sense of Bock and Wahlert 1965). Through Cenozoic mammalian evolution, several traits have been reduced or largely modified, including tooth, limb, and tail anatomy (Kermack and Kermack 1984). These traits are not discussed in this article. However, we concentrate on the earliest mammals only by contrasting them to non-mammalian synapsids.

Toward an akinetic skull

To begin at the most anterior part of the body: A rarely discussed question, which must be raised here, is whether the mouth cavity of early mammals respectively mammaliaforms was confined by lips and cheeks. Together with the muscular (and therefore deformable) tongue and the secondary palate, these are typical traits of mammals today (Starck 1995). All members of Theria have in common that food is comminuted inside the mouth cavity, not outside, as in the majority of other vertebrates (Schwenk 2000). Neither lips, nor cheeks, nor the tongue have a good chance of being preserved as fossil documents. Direct proof in the form of body fossils is therefore unlikely (Martin et al. 2015). Despite this lack of information, the existence of lips in early mammals is generally assumed (Maier 1999).

Maier (1999) concluded that the development of lips, cheek, and tongue is connected with the most typical trait that characterizes all mammals: the sucking of milk in the early stages of life (Werneburg and Spiekman 2018). This trait is closely related to the formation of a secondary hard and soft palate along synapsid evolution (Maier et al. 1996). Related to the evolution of a robust palate, there is also an increasing development of akinetic skulls toward Mammalia (Kemp 2005; Kammerer et al. 2014). In particular, this concerns the formation of a secondary lateral braincase wall by expansion of the alisphenoid, a homolog to the mobile epipterygoid in other amniotes (Maier 1987, 1989; Werneburg and Maier 2019).

The presence of an akinetic skull permits hard biting because the skull bones form a stable construction to transmit related stresses (discussed by Ferreira et al. 2020). Also, the reduction of the lower jaw to just one massive bone, the dentary, appears to be favorable for that purpose (Lautenschlager et al. 2018). If compared to reptiles (Evans 2008; Werneburg 2019), synapsids are adapted to strong bite anterior in the snout (Weishampel 1997; Sues and Reisz 1998; Schwenk 2000; Sues 2000; Kemp 2005). Bite force on the large canines results in compressive stress on the skull roof. The base of the skull, however, has to withstand tensile stresses when biting anteriorly, which results in strong ossification in the respective skull regions, including a secondary palate and the jugal arch (Werneburg and Preuschoft 2024). When handling resisting food items, transversal components result in comprehensive forces that are transmitted beyond the skull toward the neck, shoulder, and limbs. Strong transversal forces are then associated with broad shoulder plates and sprawling limbs (Fig. 2B) (Preuschoft 2022; Preuschoft et al. 2022; Werneburg and Preuschoft 2024).

Fig. 2
figure 2

Locomotion is inevitably connected with torsion of the trunk, between fore- and hind limbs (A) because the lifting of at least one extremity is essential. Catching prey or rupturing tough fibers of plants also leads to torsion (B), which now, however, includes the neck and reaches rearward to the pelvis. The resultant of transverse and weight-forces must be inside the limits of area of support, and therefore requires sprawling forelimbs

Full size image

Tooth anatomy and food processing

The shape of the jaw joint in Mesozoic mammaliaforms, which might be relevant in this discussion, has not yet been investigated in detail (but see Lautenschlager et al. 2018). It can be assumed, however, that a simple secondary joint between dentary and squamosal (Fig. 1A: No. 2) can sustain higher forces than a combined joint between articular, quadrate, and squamosal, as seen in the Triassic mammaliamorph Probainognathus (Romer 1970) for example.

The existence of only one large and pointed tooth in each of the jaws is sufficient for catching prey (Bels and Whishaw 2019; Schwenk 2000). This tooth must be located in the most anterior part of the snout, far from the temporomandibular joint. This is despite the long lever arms, which reduce the bite force (Fig. 1C). However, this disadvantage is partly compensated by the position of the canines close to the midline of the jaws, thus using the muscle power of both sides (Fig. 1D-E; see Werneburg and Preuschoft 2024, for a summary of the basic biomechanics).

Basic information about how teeth work has been provided by Lucas and Luke (1984), and their principles are used in the present paper. The highest cusps in mammaliaform morganucodontans and crown mammalian eutriconodontans were arranged in a row, one behind the other, forming long cutting blades that comprise several teeth (Fig. 1A: No. 1; Martin 2018; Jäger et al. 2019; Martin et al. 2020). The serial arrangement of cutting blades is suitable for slicing animal tissue and eventually cutting off those parts of the seized prey, which extend laterally from the mouth. So, they can be reduced more easily to palatable portions (Bels and Whishaw 2019; Schwenk 2000).

Among Mammaliaformes, some docodontans (e.g., Simpsonodon, Kermack et al. 1987) and Dsungarodon (Pfretzschner et al. 2005) (Fig. 1B) possess molars with a crushing-grinding function, adapted to an omnivorous diet. This function has evolved independently from and prior to the tribosphenic molars of therian mammals (Kermack et al. 1987; Pfretzschner et al. 2005; Martin et al. 2020). The crown mammalian multituberculates (for alternative phylogenetic placement see e.g., Mao et al. 2024) developed multicusped molars with grinding surfaces, which resemble those of rodents and probably have been used to comminute tough, fiber-rich food (e.g., Lazzari et al. 2010). The tribosphenic molars evolved in the therian stem lineage, and Juramaia from the Late Jurassic of northeastern China is the oldest known mammal with tribosphenic dentition (Luo et al. 2011). Tribosphenic molars combine a puncture-cutting with a squeezing-grinding function, which together permit the exploitation of new food sources (Maier 1978; Schultz and Martin 2014; Martin et al. 2020).

Chewing is a key character of Mammalia which allows crushing of food items to small pieces for better digestion starting in the mouth (Maier 1978; Lucas and Luke 1984; Schwenk 2000; Fuentes et al. 2021). By chewing in the back of the mouth, local compression and shearing forces are applied that help shaping the jugal arch as a bow (Fig. 1A); the strong transversal forces, that still characterized early synapsid feeding behavior, however, are largely reduced (Werneburg and Preuschoft 2024). Consequently, bite forces are less transmitted to the postcranium, supporting the transformation of sprawling to upright posture of the earliest mammals (further discussed below).

Body size and locomotor speed

The shape of the vertebrate body is highly linked to the mode of the animal’s locomotion (Preuschoft 2022). Two values characterize the locomotion: acceleration of the body and the locomotory speed (Preuschoft 2023). When the animal is slowing down, speeding up, or changing its direction, the small body size of the early mammals does not evoke high resistance against adopting the new speed. In terms of mechanics (e.g., Schreyer et al. 1964; Grote and Feldhusen 2007; Preuschoft 2022), an acceleration of the body takes place against the mass inertia. The essential factor is Mi = m * a2, where the mass (m) is accelerated and (a) is the acceleration the body is exposed to. The ability of the animal to start quickly is part of the ability to move rapidly.

The locomotor speed of the animal (V), on the other hand, depends on the length of its stride cycle (s) and stride cycle frequency (F): V = s * F. The length of stride cycles depends on the excursion angles of the limbs and the distance between the ground and the joint closest to the trunk; it is the cosine of limb length at a given angle. The angle is usually chosen in a way that ground reaction forces (GRF = resultants of a share of body weight and braking or re-accelerating forces) pass close to the limb joints (Fig. 1F). So, they evoke the lowest possible torque at the joints. The frequency of steps depends upon the laws of the suspended pendulum during the foreswing. The pendulum period (T) is calculated as T = 2 * π * √ (l/g). The values for π (3.14) and g (earth acceleration, i.e. 0.981) are constant and the length (l) is variable. The shorter the cord of the pendulum (l), the faster it swings and the higher its frequency (Fig. 1G). To sum up: long and light-weight extremities allow faster locomotion than short and heavy limbs. The earliest mammals possessed slender and rather long, light-weight extremities that permitted relatively fast locomotion and extended posture (Fig. 3).

Fig. 3
figure 3

A Hypothetical animal with long (drawn) and short (schematic) limbs. If excursion angles (black) are the same, long limbs yield longer steps (s = 2 * sinus of excursion angles * length; indicated by horizontal lines at the lower margin) than short limbs (horizontal limbs above those mentioned before). B Period of a mathematical pendulum. It follows a root function. The speed of the mass becomes higher with a shorter length of the chord (data from Demes and Preuschoft 1984)

Full size image

The transition from sprawling to an extended posture

The sprawling limbs as present in the mammalian synapsid ancestors (Preuschoft et al. 2022; Bishop and Pierce 2024a,b) were short, providing the selective advantage of keeping the torques in the joints within limits (Fig. 4A-B). This is most evident for the stylopodia—humerus and femur (Christian and Garland 1996). The zygopodia—ulna/radius and fibula/tibia—could not be much longer, too. This is because the rotating moments in the elbow, knee, carpal, and tarsal joints would have grown large—a fact that would have entailed stronger and heavier musculature on the forearms and lower legs.

Fig. 4
figure 4

Schematic presentation of sprawled extremities (A-B) and parasagittal limbs (C-D). A and C seen from behind, B and D seen from the left side. Torques are high in A and B, but small in C and D. This holds especially true on the right side of C, while on the left side two of the joints are placed underneath weight, so reducing torques to nearly zero. In D, two positions of the upper arms and the lower leg are shown (see text). The coat of muscles which covers the skeletal rod (E) can be concentrated without a loss of volume to its proximal part (F). While its center of volume moves more and more away from the pivot, so that its pendulum cord becomes linearly longer with one half of leg length (G), it lags behind when muscle mass is concentrated on the proximal part, depending upon the chosen ratio length/thickness of the muscles

Full size image

A means, permitting the (plesiomorphic) sprawling limbs (Fig. 4A-B) to become longer and at the same time to reduce their weight, is the following: If—in frontal view—limb segments are placed one on top of each other, these “stilt legs” could become long, without increasing the moment arms of body weight. This is because the segments of the limbs are close to and parallel to the resultant ground reaction force (Fig. 4C-D); hence, little or no bending occurs in the limbs. For running at high speeds, it is no disadvantage, if all excursions of the joints take place in the direction of movement.

In addition, the heavy musculature can be concentrated in the proximal part of the limb (Fig. 4E-F). The distal part then consists only of the load-carrying bones of the skeleton, the tendons to control the joints, and the skin to wrap up all this. These morphological changes lead to the well-known “clubs” of bird extremities (“chicken drumstick”) and the “calfs” of humans. In terms of mechanics, the pendulum length (l) is shortened, leading to an increased pendulum frequency (F) (see section above; Fig. 4G).

Construction of the pelvis

As mentioned above, in the mammalian pelvic girdle, the ventral plate is not developed (Fig. 5A), so no origin for a strong adductor exists (Starck 1995). The tail of mammals may be missing, short, or lightweight and does not require a sprawling of the hind limbs (Fig. 5B-C), as seen in their synapsid ancestors which have massive, though short tails when compared to the original amniote, reptilian-like condition (Fig. 2A) (Preuschoft et al. 2022). Therefore, their hind limbs can be moved in parasagittal planes. The experimentally documented flow of high forces from the hip joint to the vertebral column is directed dorsally and slightly cranially (Preuschoft 2022)—exactly like the more or less rod-shaped ilium.

Fig. 5
figure 5

A Schematic drawing of a mammalian pelvis, with ilium cranial to the hip joint, ischium caudal to the hip joint, and pubis bracing the strongly bent ischium like a console. B-C Model animal bounding with the vertebral column (black), scapula, and pelvis (white) and moving the tail to equilibrate foreward swinging of the hind limbs

Full size image

Behind the hip joint, all Mesozoic mammals—so far as known—possess a rearward-directed ischium (Martin 2018). It provides the origin for the important retractors of the thigh. If these muscles contract, the bone is exposed to very high bending moments, which are sustained by a “console”, formed by the pubes of both sides, which brace the ischia and support each other in the middle.

This arrangement is realized in nearly all living mammals (Starck 1995). In non-mammalian synapsids, however, Preuschoft et al. (2022) reported different post-coxal outgrows of the ilium, which replace the attachment site for the retractors of the upper thigh and the entire hind limb.

A simplified shoulder and mechanically enhanced girdle elements

In therian mammals, the shoulder girdle does not contain a ventral plate (Bendel et al. 2022). The scapulocoracoid in amphibians, reptiles, non-mammalian synapsids, and monotremes provides an extended origin of the powerful musculus (m.) pectoralis, necessary to keep the upper arm in its sprawled position (Fig. 4A-B; see Preuschoft 2022; Preuschoft et al. 2022). The limbs of Mesozoic stem therian mammals were no longer in the sprawling posture but instead were moved in parasagittal planes in cranial and caudal directions. In evolution, the pelvic girdle and hind limbs preceded the shoulder girdle and forelimbs in the derived parasagittal posture as evident from the Early Cretaceous eutriconodontan Jeholodens jenkinsi (Ji et al. 1999) from China. However, in therian mammals, the heavy body stem is still suspended by muscles from the dorsally and slightly caudally inclined scapula (Fig. 4C-D). This element, therefore, is preserved as the only remnant of the shoulder girdle—because it carries the body stem via the m. serratus (Preuschoft 2022; Preuschoft et al. 2022).

Preuschoft (2022) observed the inclination of the elements between the joints of the girdle skeletons and the vertebral column in a quasi-experimental approach by FES-analysis. These girdle elements (scapula in the anterior and ilium in the posterior) carry the highest stresses and guide them into the vertebral column. While the scapula is inclined caudally, the ilium is inclined cranially (cf. Figure 5A-C)—both according to the flow of stresses.

The question, why the support of the body stem differs between shoulder and pelvis, has been illustrated by the “ladder experiment” (Preuschoft 2022; Preuschoft et al. 2022). If two persons carry a ladder by walking between two rungs and pressing one rung tightly against their pelvis, they are compelled to walk in lockstep. If they do not, their balance is seriously disturbed. To avoid this disturbance and select step length according to size, or ground structure, one of them has to de-couple the connection to the ladder, by loosening the contact with the rung and carrying it on the arms alone (Fig. 6A-D). For small animals, an “explosive” takeoff, a change of direction, or a rapid attack to catch agile prey or to avoid predators, are vital. Therefore, the fixed bony connection between the pelvis and vertebral column, which permits fast acceleration, is more important than the slowing down in the braking phase, in which damping by muscles even offers advantages.

Fig. 6
figure 6

A–D Ladder experiment. The ladder represents the not compressible vertebral column; the two hind limbs of the carriers represent the fore- and hind limbs of a quadruped. C is decoupled and D is coupled to the hip. See text for further details. E, F Vertebral column. Even if the flexion between segments is strictly limited, a greater number of segments permits wider excursions. Joints = circles between the black segments

Full size image

Body torsion and rib reduction in the lumbar region

Locomotion on limbs requires the lifting of at least one of them. Any lifting of a limb leads to torsion of the trunk in a quadruped (Fischer 1994). This torsion is particularly marked if the limbs are kept in abducted or sprawling postures (Fig. 2A).

Preuschoft et al. (2022) demonstrated that the neck of non-mammalian synapsids was often exposed to very high torsional moments by movements of seized prey or by the resistance of tough fibers of plants against being disrupted (Fig. 2B). To make sure that the anterior body was not turned over, a sprawling posture with widely abducted anterior limbs is advantageous. The maximal torsional moment (M) that can be sustained on a forelimb is determined by the ground reaction force (GRF) multiplied by the horizontal distance (d) between the loaded forepaw and the long axis of the animal: M = GRF * d. The first value does, according to experience, not exceed two times body weight, and the second cannot be larger than the length of the upper arm plus the forearm. Despite these limitations, the torsional moments can become impressively large. Resistance against torsion is provided by the ribs and the intercostal muscles. The points of attack of most transverse forces are the jaws, and therefore the neck needs these structures, especially the bony neck ribs (see Arnold 2021; Galis et al. 2021). Since the torsional moments are very large, they are also transmitted to the trunk, with the consequence of a rib cage reaching rearward back to the pelvis.

Because the limbs of the earliest mammals were not abducted, and probably also because their teeth allowed cutting of food items at the sides of the mouth (see above), torsional moments were much smaller than in the non-mammalian synapsids. Since the structures that resist torsion are the ribs, their existence in the trunk and their number can be largely reduced in caudal direction. This may have led them to increase the number of “lumbar” vertebrae losing their ribs and the lumbar region becomes longer.

By forming a lumbar region, the ancestral thorax with a caudally extended rib cage became considerably shortened (Kort 2023). The extended rib cage offered a resistance against flexing the trunk before, while the rib-less lumbar vertebrae could now be flexed and extended in relation to the thorax. Together with the flexible lumbar vertebrae, the pelvis and the hip joint could be moved.

The more numerous and the longer the lumbar vertebrae are, the greater is the distance over which an animal moves forward by an extension (Fig. 6E-F). This extension and flexing of the trunk are used by nearly all living “small” mammals (in the sense of Fischer 1994) when moving in one of the “springing” gaits: bounding or galloping. In hyraxes, for example, the movements of the trunk contribute 40% of the distance covered by the hind limbs (Fischer 1994).

Tail reduction

The long and slender tails of the earliest mammals may well have played a role in the characteristic locomotion as counter-force against the inertia of the hind limbs during the fore-swing. The latter are swung forward while no other force can be applied. The tail is lifted dorsally in that phase and lowered while the animal pushes its body forward with the rearward moving hind limbs (see Nieschalk 1991, for Cheirogaleus, or Peters and Preuschoft 1984, for Tarsius). An exact calculation is not possible, because the relative masses in fossils can be determined only tentatively. We assume, however, that the body and tail weight ratio of the majority of Mesozoic mammals were comparable to that of the extant Tarsius. For this animal, Peters and Preuschoft (1984) have conducted a precise calculation. The authors found that the mass and length of the tail provide a mass moment of inertia large enough to rotate the strong and heavy hind limbs from their rear position during take-off to an anterior position during landing.

Because tails are not so much moved laterally in most mammals, and because their mass moment of inertia is restricted, there is no reason to keep the hind limbs in an abducted, sprawled position. This was still necessary for non-mammalian synapsids with their massive tails that were probably moved in an undulating manner.

Locomotion in small and large mammals

In the course of the evolution of the oldest mammals or their premammalian synapsid ancestors, greater locomotor speed became a considerable evolutionary advantage despite the disadvantage of the higher metabolic rate (Newham et al. 2020). The ability to start explosively is part of the ability to move rapidly. Many derived traits of the locomotor apparatus can be seen as means to achieve higher speed.

It should be noted, however, that any increase in speed must be seen in relation to absolute size. The absolute speed of small animals always lags behind that of large animals. On the other hand, rapid acceleration and swift change of direction are much easier in small than in large mammals, simply because of the large mass moment of inertia in the latter.

Extension and flexing of the trunk, the “newly invented” gaits, are used by most living “small” mammals (in the sense of Fischer 1994). These gaits can be symmetrical (with both fore- and hind limbs moving at the same time and close to each other: bound) or asymmetrical (with forelimbs moving one after the other and hind limbs synchronously: half-bound).

In large animals (perissodactyls, bovids, cervids, giraffes, carnivorans) and in tree-living primates, including the small callithrichids (Arms 1998), bounding is often replaced by the gallop, in which longer phases of ground support by at least one leg yield better ground contacts and therefore more safety. Flexing and extending the trunk is replaced in galloping large mammals by rapid lifting of the hip region. Thereby the fore-swing of the hind limbs takes place in a stronger gravity field (in horses: 1.7 g; see Witte et al. 1995) and, therefore, in a shorter time interval than without this effect.

In very large animals, the angles between segments are wide and therefore the balancing does not require much muscle force (Schreyer et al. 1964; Grote and Feldhusen 2007; Preuschoft 2022). Small animals always can produce greater forces (proportional to cross-sectional areas of segments and therefore of muscles) in relation to body mass (proportional to volume, that is mass). Therefore, small animals can afford narrower joint angles in their limbs, which allow for the production of rapid accelerations during rather short time-intervals.

Conclusions

  1. 1.

    A certain independency of the jaw apparatus from the locomotor system is already exemplified by the grinding teeth in gomphodont cynodonts (e.g., Diademodon) and multituberculates. However, sprawling forelimbs of gomphodont cynodonts illustrate a strong argument for mosaic evolution. The same applies to their overlapping ribs, which relieve the vertebral column from compressive force. In contrast, mammaliaform morganucodontans and crown mammalian eutriconodontans as earliest mammals appear to be derived from these progressive synapsids.

  2. 2.

    Only two fang teeth exist in the lower and upper jaw. Teeth are arranged in a continuous row and have several cusps and roots. Lips, cheeks, and a muscular tongue are essential for sucking milk from mammae in very young individuals, and permitting fully grown animals the comminution of food inside the mouth cavity, not outside, as in other animals.

  3. 3.

    The construction of the limb girdles differs widely from that in synapsid forerunners, although some of the traits, especially of the pelvic girdle existed in some non-mammalian synapsids as well.

  4. 4.

    The existence of a rib-less lumbar vertebral column permits the invention of a new, springing gait.

  5. 5.

    The rather small oldest mammals were characterized by a great locomotor speed, which implies a considerable evolutionary advantage. Small body size makes it necessary and permits rapid acceleration forward, change of direction, and braking in rapid locomotion.

  6. 6.

    The rather long, but light-weight tails may have contributed to the newly invented rapid gait.