Mammalian cochleae vary absolutely in length more than those seen in any other group of vertebrates. In the immediate ancestors of mammals, among the synapsid “reptiles”, the cochleae were less than 2 mm long, including the lagenar macula, meaning that the basilar membrane was probably about 1.5 mm long. The shortest known in true mammals were also about 2 mm long, seen in fossils of different lineages (e.g., Hadrocodium, Dryolestes, Henkelotherium; Luo et al. 2001, 2010; Ruf et al. 2009) and not coiled. The longest known cochleae in modern therians are those from some modern baleen whales and are more than 50 mm long, a length difference to the earliest species of more than 30-fold. All of these cochleae were supplied by three-ossicle middle ears of various stages of development (Fig. 1.1). In two mammalian lineages, the cochlea never evolved to great lengths. One of these lineages led to the modern egg-laying monotremes, and the other was a long-lived and diverse lineage that died out about 30 Ma ago and is known as the multituberculates (Vater et al. 2004). Indeed, in multituberculates, after 150 Ma of the possession of a three-ossicle middle ear, the cochlea was still 2–7 mm long (Hurum 1998; Luo and Ketten 1991). This clearly demonstrates the mosaic nature of the evolution of the auditory periphery and that different components can be in quite different stages of their evolution. Even in modern monotremes, cochlear length does not exceed 8 mm (e.g., Ladhams and Pickles 1996). It is thus fair to observe that the possession of a three-ossicle middle ear does not in any way automatically lead to dramatic evolution of the cochlea. The question thus becomes more interesting: Why did the cochlea evolve in a spectacularly different way in one mammalian lineage (therian ancestors)?

A comparison of all three modern groups of mammals (monotremes, marsupials and placentals) shows that all have a hearing organ that can be termed an organ of Corti: There are two groups of hair cells that are on the inner and outer sides of a tunnel formed by pillar cells (e.g., Ladhams and Pickles 1996). This strongly suggests that this structural configuration was already present in the ancestors of mammals and that multituberculate mammals also possessed an organ of Corti. Of course, if we can assume that the monotreme cochlea represents the primitive state, during the evolution of the therian cochlea, the number of cells across the organ has been greatly reduced. Comparative evidence points to hearing being restricted to low frequencies in ancestral mammals. This indicates that the possession of an organ of Corti as such does not automatically lead to high-frequency hearing, even after more than 100 Ma of evolution; monotremes have an upper frequency limit of roughly 15 kHz, and they are relatively insensitive (Gates et al. 1974; Mills and Shepherd 2001). So what is unique about the therians that led to them evolving a long, coiled cochlea, the DMME and, later, high-frequency hearing above 20 kHz?

The fossil evidence indicates that after the origin of mammals, it took more than 50 Ma for the therian mammal lineage to achieve full coiling of the cochlea (Wible et al. 2001), which occurred shortly before placentals and marsupials split. At that time, the length of the hearing organ was 4–5 mm, which is shorter than the cochlea of any modern mammal. Examining all that is known about cochlear structure of the various lineages leads to the conclusion that therians developed one feature that is unique and perhaps decisive for further evolution. During the middle Jurassic era, the soft tissues of the therian cochlea became integrated with the bony canal that surrounded them. This led to the evolution of primary and secondary bony laminae that support the basilar membrane and a bony enclosure for the afferent ganglion (Luo et al. 2001, 2010; Kemp 2005; Ruf et al. 2009). This bony integration presumably improved the impedance match between the organ of Corti and the stiff middle ears of the period.

After the development of a stiffer support for the basilar membrane, the middle ear and the cochlea of therian ancestors evolved until, shortly before the marsupial and placental lineages split, both a DMME and a fully coiled cochlea (about 5 mm long) had been achieved. This became the basic structural configuration of therians. About 80 Ma ago, a marsupial cochlea existed with 1.25 turns in its coil (Meng and Fox, 1995). In modern therians, the minimum number of turns is 1.5 (e.g., in mice). Of course marsupials and especially placentals later radiated diversely into a huge number of groups that vary in the characteristics of their middle and inner ears. Some hugely elongated the cochlea, some evolved echolocation and very high-frequency hearing, and others became much larger and more low-frequency efficient. Even some very small mammals specialized for low-frequency hearing by, e.g., greatly enlarging their middle-ear bullae. Although it has been speculated that the early presence of primary and secondary laminae indicate that even the earliest therians were specialized for high-frequency hearing (Luo 2011), this is very unlikely. The distribution of secondary laminae is sporadic in modern therians; although it is always present where ultrasonic hearing is found, some placentals lack a secondary lamina, e.g., primates (some of which have quite high-frequency hearing). It can thus be surmised that the laminae were an important component of the impedance matching between early middle and inner ears, no matter what the frequency range being processed at that time. More likely is a gradual rise in the upper frequency limit. Clear fossil evidence of ultrasonic sound processing is not found in the earliest bats (Simmons et al. 2008) but probably evolved soon after in the early Cenozoic, 50–55 Ma ago.