One of the most unique features of auditory perception in mammals is the ability to hear sounds above 12 kHz (Fay,1988). Snakes and turtles are relatively insensitive to airborne sounds and can rarely hear sounds above 2 kHz (Fig. 1). Lizards and crocodilians show slightly better sensitivity but are still limited to frequencies below 4–8 kHz. Most birds can hear sounds up to around 10 kHz although some predatory birds such as barn owls can hear sounds as high as 12 kHz (Konishi,1973). In contrast, most mammals have an upper frequency limit of hearing that ranges from 30 to 60 kHz and some bats and aquatic mammals are able to detect acoustic signals above 100 kHz (Fig. 1). Humans (upper limit ∼ 18 kHz), elephants (Elephas maximus—upper limit ∼ 11 kHz), and naked mole rats (Hetercephalus glaber—upper limit ∼ 12 kHz) present a few exceptions to the general mammalian pattern of good high-frequency hearing (Heffner and Heffner,1982,1993; Jackson et al.,1999).
Although it has been argued that the development of a three-bone ossicular system and coiling (and elongation) of the cochlea were key adaptations that led to good high-frequency sensitivity (Masterson et al.,1969; Manley,1972; Fleischer,1978; Echteler et al.,1994; Frost and Masterton,1994; Fox and Meng,1997), determining the timing of key events in the evolution of hearing abilities of mammals has been a topic of debate. One leading hypothesis on the subject proposes that primitive mammals shifted to a primarily high-frequency condition soon after acquiring the three-bone ossicular system (Jurassic) and then gradually (re)developed good low-frequency sensitivity starting in the Cretaceous and extending through the early part of the Cenozoic (Masterson et al.,1969; Jerison,1973; Frost and Masterton,1994).
Comparative studies in living mammals have generally supported this hypothesis by showing that “primitive” mammals like opossums and hedgehogs are characterized by good high-frequency hearing and relatively poor low-frequency sensitivity (Ravizza et al.,1969a, b; Frost and Masterton,1994). These animals also have what is generally considered to be “ancestral” characteristics in ear morphology such as the “microtype” middle ear bone configuration, which seems particularly well adapted for transmitting high-frequency sounds (Fleischer,1978; Rosowski,1992). In recent years, it has become increasingly possible to use fossils to investigate evolutionary hearing patterns, although there are still a few temporal gaps, which have prevented a comprehensive evaluation of proposed evolutionary sequences.
The incipient development of the three-bone ossicular system during the Late Triassic (roughly 200 Ma), such as witnessed in fossil mammaliaformes like Morganucodon, may have resulted in a slight shift toward higher frequencies although these gains were probably modest since these animals had relatively large ossicles and short, uncoiled cochleae (Kermack and Mussett,1983; Rosowski and Graybeal,1991; Rosowski,1992). Comparative studies have shown that high-frequency limits are correlated with ossicular mass (Hemilä et al.,1995; Coleman and Colbert,2010), so the relatively large ossicles in Morganucodon likely placed inertial limitations on how quickly the ear bones could vibrate, limiting high-frequency transmission through the middle ear.
The hearing range in early mammals such as multituberculates (late Jurassic – late Eocene) was probably not that much different from that of living monotremes. Multituberculates such as Lambdopsalis share numerous similarities to monotremes in middle ear structure such as the orientation of the ectotympanic bone, the position of the malleus relative to the incus and the simple morphology of the incus (Meng and Wyss,1995). Mechanical analyses of ear function in echidnas (Tachyglossus aculeatus) suggest that their middle ear bones are relatively inefficient at transmitting sounds, particular at higher frequencies (Aitkin and Johnston,1972). In addition, multituberculates were similar to extant monotremes in having a relatively short cochlea that curves laterally but does not complete one full turn (Fox and Meng,1997). Auditory brainstem response estimates of hearing in echidnas (Tachyglossus aculeatus) suggest that these prototherians, and possibly multituberulates as well, had a relatively limited hearing range from 1.6 to 13.9 kHz (Mills and Shepherd,2001).
During the middle to late Jurassic, the ancestors of crown therians (marsupials and placentals) began developing inner ear structures that likely signal the onset of heightened high-frequency hearing. The cladotherian mammals Henkelotherium guimarotae and Dryolestes leiriensis from late Jurassic deposits in Portugal (156–150 Ma) had relatively short cochleae (2.7–3.3 mm) that completed about 3/4 of one full turn (∼270°) (Ruf et al.,2009; Luo et al.,2011), similar to multituberculates and monotremes. However, unlike multituberculates and monotremes, Heneklotherium and Dryolestes demonstrate evidence for development of primary and secondary bony laminae along the basal end of the cochlear canal (Ruf et al.,2009; Luo et al.,2011). Possessing both primary and secondary bony laminae generally reduces the width of the basilar membrane and also allows this membrane to be more tense (stiff) along the basal end, both of which help promote the reception of high-frequencies (Fleischer,1976). Although some animals can hear relatively high frequencies without the presence of a secondary lamina, it is considered a basic anatomical requirement for the specialized high-frequency hearing of bats and cetaceans (Bruns,1980; Ketten,1992; Vater et al.,2004).
Although the bony laminae in Heneklotherium and Dryolestes suggest they may have had nascent adaptations for relatively good high-frequency hearing, additional auditory characteristics indicate that the overall range and sensitivity of hearing was probably limited in comparison with extant therians. For one, the presence of a Meckelian groove in Heneklotherium and Dryolestes suggest that these taxa had a “transitional mammalian middle ear” that was not as efficient as the “definitive mammalian middle ear” at transmitting airborne sounds (Meng et al.,2011). In addition, their relatively short cochleae likely placed limitations on the range of perceptible frequencies because there may not have been adequate space on the basilar membrane to have auditory sensory neurons (hair cells) devoted to both high- and low-frequency sensitivity.
Complete coiling of the cochlear canal may have been one strategy for increasing basilar membrane length in a compact space. The hearing in the first mammals that had short but coiled cochleae was presumably like marsupials such as opossums (Fig. 1), which have coiled but relatively short basilar membranes [e.g., 6.4 mm, Monodelphis domestica—(Muller et al.,1993)]. The limited low-frequency hearing in living opossums is not strictly related to small body size since even the medium-sized Didelphis virginiana (4 kg) can hear sounds only down to about 1 kHz [overall range = 1–68 kHz—(Ravizza et al.,1969a)]. Furthermore, “primitive” extant placental mammals such as hedgehogs (Hemiechinus auritus) also have limited low-frequency sensitivity (500 Hz to >60 kHz) and a presumably relatively short cochlea as implied by the fact that their cochleae have only 1 1/2 spiral turns (Lewis et al.,1985).
The oldest known mammal that demonstrates at least one full coil of the cochlea comes from early Cretaceous deposits (125–100 Ma) in Mongolia and has been attributed to Prokennalestes trofimovi (Wible et al.,2001). Younger mammalian specimens from late Cretaceous deposits in Canada (84–77 Ma) had cochlear canals with approximately 1 1/4 turns (Meng and Fox,1995a) and even younger specimens from the Bug Creek Anthills locality in Montana (∼65 Ma) demonstrate cochleae with 1 1/2 turns (Meng and Fox,1995b). The estimated hearing for the placental specimens from the Bug Creek Anthills locality ranged from 1.23 to 87.7 kHz and that of a marsupial from the same site had a range from 1.58 to 73.8 kHz (Meng and Fox,1995b). This evidence suggests that therian mammals living around the time of the K-T boundary were characterized by heightened high-frequency hearing but relatively poor low-frequency sensitivity.
During the Cenozoic, therians apparently continued these trends in auditory evolution and developed essentially modern morphologies (and presumably hearing patterns) by the Miocene. Fleischer (1976) compared the ear structures in extinct and living cetaceans and concluded that the specialized hearing of odontocetes related to echolocation evolved during the Oligocene and was essentially similar to modern patterns by the Miocene. More recently, Coleman et al. (2010) examined the middle and inner ears of Miocene aged fossil New World monkeys and predicted that these ∼20 my old primates had low- (and possibly high-) frequency sensitivity similar to living monkeys from South America. However, there is still a paucity of studies that have focused on reconstructing hearing patterns from the early Cenozoic, a period that figures prominently in arguments like the “Masterson hypothesis” (Masterson et al.,1969).
The primate fossil record offers an opportunity to begin to address this problem due to the abundance of specimens that span the Cenozoic era. In this study, we examined relevant auditory structures in 21 species of extinct primates and closely related taxa that range in geologic age from 60 Ma until near-modern times. We then compared these specimens with a sample of extant euarchontans (Fig. 2) and used predictive equations to estimate the low- and high-frequency sensitivity of the fossils.
MATERIALS AND METHODS
The fossil specimens examined in this study are given in Table 1. The actual fossils themselves were not analyzed. Instead, high resolution X-ray computed tomography (HRXCT) was used to construct digital models of auditory structures (see below) that are often preserved in fossils and that have been found to be functionally relevant. The fossil specimens were scanned at various institutions and the voxel dimensions for each specimen are given in Table 1. The geologically oldest assemblage of fossils belong to the group referred to as plesiadapiformes, which are now thought to be stem primates with a sister-group relationship to Euprimates (Bloch et al.,2007). Our sample consisted of six species of small to medium-sized plesiadapiforms from North America and Europe that range in age from 60 to 54 Ma. The next group consisted of three medium-sized fossil primates from North America and Europe called adapoids that range in age from 50 to 35 Ma. We also examined two similarly aged (45–35 Ma) small to medium-sized species of omomyoids from North America and Europe. Our sample included one 45 Ma primate specimen from China that has been suggested to represent a basal anthropoid‡ (MacPhee et al.,1995). In addition, we sampled four species of medium-sized unambiguous fossil anthropoids: one 30 Ma fossil from Africa and three ∼20 Ma fossils from South America. Lastly, we examined five species of extinct large-bodied subfossil lemurs that come from recent geological deposits in Madagascar (late Pleistocene-Holocene).
Table 1. Fossil taxa examined in this study
The units for cochlear length (CL) are in millimeters, those for oval window area (OWA) are in millimeters squared, those for body mass (BM) are in grams and those for voxel dimensions are in microns. CS refers to number of cochlear spirals and SBL refers to the number of spirals of the secondary bony lamina (when present). BM Ref. refers to references used for body mass estimates. Body mass estimates with an asterisk (*) based on the mean of the smallest and largest species in the genus (because the exact species designation is unknown). Body mass estimate for Tremacebus (**) based on the value for Dolichocebus but reduced to 1,500 g because of the slightly shorter skull length (Kay et al.,2008).
The main comparative sample analyzed in this study (Table 2) consisted of a phylogenetically diverse group of 25 species (72 specimens total) of living euarchontans. This sample included seven species of platyrrhines (New World monkeys), three species of catarrhines (Old World monkeys) and four species of Tarsiers, which are collectively referred to as haplorhines (= anthropoids + tarsiers). In addition, we sampled five species of lorisoids and two species of lemuroids that belong to the primate suborder termed strepsirhines. We also examined four species of nonprimate euarchontans: two species of scandentians (treeshrews) and two species of dermopterans (flying lemurs).
Table 2. CL estimates for extant taxa examined in this study
CL units are in millimeters, N represents number of specimens examined and S.D. represents one standard deviation. CS and SBL abbreviations the same as in Table 1. Body mass (BM) estimates are in grams and are taken from Smith and Jungers (1997) for primates, from Askay (2000) for tree shrews, and from Myers (2000) for colugos except where noted by asterisks. Body mass estimates for Nycticebus javanicus (*) based on Nekaris et al. (2008) and those for Tarsius pelengensis based on mean values for T. bancanus and T. syrichta which have similarly sized skulls as T. pelengensis (unpublished data).
2 3/4 – 3
2 1/2 – 2 7/8
3 – 3 1/8
2 1/4 – 2 5/8
2 7/8 – 3 1/8
2 1/2 – 2 3/4
2 5/8 – 2 7/8
2 3/8 – 2 1/2
2 7/8 – 3 1/8
2 3/8 – 2 1/2
2 – 2 3/8
2 1/4 – 2 1/2
2 7/8 – 3 1/8
2 7/8 – 3
2 7/8 – 3
2 7/8 – 3
3 1/2 – 3 5/8
3 5/8 – 3 7/8
3 3/4 – 3 7/8
2 3/4 – 3
As with the fossil specimens, HRXCT was used to create digital models of relevant auditory structures. Most extant specimens were scanned at the University of Texas High-Resolution X-ray CT facility (Table 2). These scans had voxel dimensions measuring 62.5 μm × 62.5 μm × 68.0 μm and the final images were 16 bit TIFF files. However, a subset of the extant specimens was scanned at higher resolutions. These included one specimen of Tarsius syrichta and two specimens of Tarsius tarsier (15.0 μm × 15.0 μm × 15.0 μm), one specimen of Cynocephalus volans (15.0 μm × 15.0 μm × 15.0 μm), and four specimens of Ptilocercus lowii (24.0 μm × 24.0 μm × 26.0 μm). Besides the extant specimens that were analyzed using computed tomography, additional data on oval window area (or stapedial footplate area) for 71 primate species were taken from published reports (Coleman et al.,2010) augmented with previously unpublished data for several species (Appendix 1).
Measuring Auditory Structures
Cochlear length (CL) in fossil and living specimens was estimated by creating digital endocasts of the inner ear using HRXCT and measuring the outer circumference of the cochlear canal. Image stacks were imported into ImageJ 1.35f (NIH) where threshold values were determined using the half-maximum height protocol described in Coleman and Colbert (2007). The image stacks were then loaded into 3D Slicer 2.6 open source software, where all measurements were taken on three dimensional digital models. The measurements were taken by placing markers (fiducials) at the smallest possible intervals along the outer circumference of the digital cochlear endocast, starting at the distal edge of the round window and continuing until the approximate location of the helicotrema. The distance between adjacent fiducial points was then measured and all distances summed to derive the total length estimate.
The number of cochlear spirals (CS) was counted by placing a transparent radial grid (divided into one eighths) over two dimensional images of the cochlear endocasts in apical view. The measurement was started from the distal edge of the round window similar to previous studies (West,1985). The digital cochlear endocasts were also used to evaluate the potential presence of secondary bony laminae that support the outer edge of the basilar membrane (all specimens observed appeared to have primary bony laminae). It should be noted that because secondary bony laminae are relatively thin and fragile structures, it may be possible to not detect their presence (in specimens that possess them) if the voxel dimensions are too large or if the specimen is damaged. However, we detected secondary bony laminae in some specimens that were scanned at the lowest resolution (62.5 μm × 62.5 μm × 68.0 μm) of any of the specimens in our dataset, suggesting it should be possible to distinguish its presence when adequately developed. The final structure measured using HRXCT was oval window area (OWA). This structure was estimated by measuring the major and minor axes (length and width) of the oval window (or stapedial footplate if the stapes was preserved). These measurements were then used to calculate the area using the formula for an ellipse.
Predicting Hearing Sensitivity
The predictive equations used to estimate certain parameters of hearing sensitivity in the fossil specimens were based on previous research on the functional morphology of the auditory system in living euarchontans with known hearing abilities (Coleman,2007; Coleman and Colbert,2010). This research found a significant relationship (r2 = 0.922, P < 0.001) between CL and sound pressure level at 250 Hz (SPL@250 Hz) as described by the formula:
where x = CL in millimeters and y = SPL@250 Hz in decibels (relative to 20 μPa). This research also detected a significant, albeit weaker association (r2 = 0.579, P = 0.011) between OWA and sound pressure level at 32 kHz (SPL@32 kHz) described by the formula:
where x = OWA in millimeters2 and y = SPL@32 kHz in decibels. In this study we used SPL@250 Hz as a measure of low-frequency sensitivity and SPL@32 kHz as a measure of high-frequency sensitivity and predicted these variables in fossil taxa using the formulae presented above.
In addition to predicting low- and high-frequency sensitivity in individual fossils, CL and OWA were reconstructed for ancestral nodes and these values were then used to predict low- and high-frequency sensitivity at the basal stems leading to haplorhines, strepsirrhines, and all primates. Ancestral state reconstructions were performed with Mesquite (1.12) modular system for evolutionary analysis (Maddison and Maddison,2006), based on a squared-change parsimony model. Using a relatively well-resolved phylogeny and character values for terminal taxa (species), evolutionary programs such as Mesquite use algorithms to reconstruct values at internal nodes of a phylogenetic tree (i.e., essentially weighted mean values). Incorporating temporal information (branch length data) and fossil taxa can greatly increase the confidence in ancestral reconstructions, particularly toward the base of a phylogenetic tree (Finarelli and Flynn,2006).
Phylogenetic Relationships and Divergence Dates
The data used to construct the phylogenetic relationships and divergence dates for living euarchontans was based primarily on Perelman et al. (2011) that used both molecular evidence and fossil calibration points. In addition, the split within scandentia (Tupaia from Ptilocercus) was based on Bininda-Emonds et al. (2007). The position of fossil New World monkeys (Dolichocebus, Tremacebus, Homunculus) as stem platyrrhines is based on analyses presented in Kay et al. (2008) and the designation of Aegyptopithecus as a stem catarrhine is from Fleagle (1999). The divergence times and relationships of the subfossil lemurs is based on Godfrey and Jungers (2003) and Orlando et al. (2008). The designation of omomyoids as stem haplorhines and adapoids as stem strepsirrhines is based on traditional as well as recent phylogenetic analyses (Szalay and Delson,1979; Martin,1990; Ross et al.,1998; Fleagle,1999; Seiffert et al.,2009). The intragroup relationships and position of plesiadapiforms as sister taxa to Euprimates is based on Bloch et al. (2007). The taxonomic affinity of the Shanghuang petrosal has been argued to resemble either a basal anthropoid (MacPhee et al.,1995) or possibly an omomyoid (Ross and Covert,2000). Both phylogentic interpretations are investigated here.
The first results to discuss relate to the association between CL and body mass in living and recent (i.e., subfossil lemurs) taxa. Considering all of the taxonomic groups together (Fig. 3A—light gray line), there is a significant positive relationship between CL and body mass (r2 = 0.388, P < 0.001). However, the amount of variation explained (coefficient of determination = r2) is much higher when extant haplorhines (r2 = 0.692, P < 0.001) are considered separately from the other extant euarchontans (r2 = 0.742, P < 0.001). This is interpreted to indicate that the relative length of the cochlea is divided into two distinct groups (Fig. 3A—dark black lines): Extant haplorhines have the relatively longest cochleae while nonhaplorhine taxa generally have relatively shorter cochleae. There is also a significant positive relationship between body mass and OWA in extant and recent taxa (r2 = 0.825, P < 0.001). However, in this comparison haplorhines do not appear to be distinct from the other euarchontan taxa and are scattered along both sides of the best fit regression line (Fig. 3B).
Although the phylogenetic relationship of tarsiers remains debatable (Perelman et al.,2011), this study reveals that they share certain similarities in cochlear structure with living anthropoids. For example, tarsiers have remarkably long cochleae for their body size (20.1–24.1 mm) and three of the four species investigated fall above the haplorhine regression line (Fig. 3A). They also demonstrate a high number of spiral turns (3 1/2–3 7/8), which is most similar to the catarrhines among the taxa in our study (catarrhines = 2 7/8–3 1/8; platyrrhines = 2 1/2–3; strepsirrhines = 2–2 3/4; dermopterans = 2 1/4–2 7/8; scandentians = 2 3/4 – 3 1/8). In fact, tarsiers have more spiral turns than almost any other mammal that has been examined with the exception of a few animals like guinea pigs that display an average of 4 1/4 turns (West,1985). Tarsiers also exhibit a well developed secondary bony lamina that extends along the radial wall of the cochlear canal for approximately two full spiral turns (Fig. 4). In contrast, secondary bony laminae were not identified in any of the anthropoids examined here (e.g., Saimiri – Fig. 4).
CL values in living strepsirrhines generally cluster around the nonhaplorhine regression line indicating relatively shorter cochleae compared with tarsiers and anthropoids (Fig. 3A). Also unlike anthropoids, some of the strepsirrhines appear to display some indication of a secondary bony lamina. The lamina is most highly expressed in Galago senegalensis which shows a depression on the outer surface of the cochlear endocasts which extends for approximately 1/2 turn (Fig. 4). A secondary spiral (bony) lamina was also identified in galagos in a previous study by Fleischer (1973). Eulemur fulvus also shows indications of a secondary bony lamina, although not as distinct as in galagos and it appears to stretch for only 1/4 – 1/2 turns. Lemur, Nycticebus, Perodicticus, and Loris could also have short secondary laminae (∼1/4 turn) but the current evidence from the cochlear endocasts does not allow for a conclusive determination. Higher resolution CT scans are needed to better evaluate the extent of development of this structure in these and other strepsirrhine taxa.
The recently extinct subfossil lemurs show a similar pattern in CL to their smaller bodied living relatives, although there is more scatter around the nonhaplorhine line. Despite their large body mass, most subfossil lemur cochleae are no longer than New World monkeys like Aotus and Saimiri that are well over an order of magnitude smaller in body mass (Table 2). Archaeolemur sp. presents the main exception to this pattern [often called the “monkey-lemur” because of its post-cranial and dental similarities to Old World monkeys (Fleagle,1999)], and had an estimated relative CL more similar to living haplorhines (Fig. 3A). Subfossil lemurs display 1 7/8–2 5/8 spiral turns of the cochlea and none of the specimens examined here demonstrated indications of a secondary bony lamina. Considering OWA, most subfossil lemurs show relatively small values based on body mass estimates for these species (Fig. 3B). In fact, Megaladapis sp. and Palaeopropithecus sp. had areas that were approximately half the size of chimpanzees (Pan troglodytes), which are roughly equivalent in overall mass. As with CL, Archaeolemur is unusual compared with the other subfossil lemurs and shows a mean value for OWA that falls just above the regression line (Fig. 3A).
Common treeshrews (Tupaia glis) have the absolutely (15.0 mm) and relatively shortest cochleae of any extant species examined (Fig. 3A). Pen-tailed treeshrews (Ptilocercus lowii) actually have slightly longer cochleae and are somewhat more coiled than common tree-shrews despite being over three times smaller in body mass. These two species of treeshrews also have among the smallest values for OWA (Fig. 3B). Tupaia glis displays indications of a secondary bony lamina that is reminiscent of galagos although it appears to proceed for nearly one full turn. The secondary lamina is quite evident on the cochlear endocasts for Ptilocercus lowii and clearly extends for at least one full turn (Fig. 4). Both genera of dermopterans (Cynocephalus and Galeopterus) have CL values that fall very close to the nonhaplorhine line. What appears to be a thin secondary bony lamina extends for about 1/2 turn in Cynocephalus volans, but is not plainly visible on the endocasts of Galeopterus variegatus. A previous analysis found that Cynocephalus variegatus (= Galeopterus variegatus) had a weakly developed ridge along the radial wall of the cochlear canal but that it was structurally different from the secondary lamina of animals like tupaia and galagos and more similar to the condition (spiral ligament) in humans (Fleischer,1973).
Plesiadapiformes, the geologically oldest taxa, had relatively short cochleae compared with living euarchontans and display values that consistently place below the nonhaplorhine line (Fig. 3A). Plesiadapis cookei had the longest cochlea of any plesiadapiform examined but still was relatively shorter than any extant species except common treeshrews. Carpolestes simpsoni, had the shortest cochlea (8.6 mm – Fig. 3A) and smallest number of cochlear spirals (1 1/2 turns – Fig. 4) of any specimen examined, extinct or extant. Most plesiadapiformes had OWA values that were close to the best fit regression line although Carpolestes was relatively smaller than most other taxa (Fig. 3B). Carpolestes also had a clearly developed secondary bony lamina that extended for at least 1/2 turn (Fig. 4). All of the remaining plesiadapiformes also appear to have had a secondary lamina although generally not apparently as developed as in Carpolestes (i.e., shallower depression in the endocasts and extending only 1/4–1/2 turns).
In contrast to plesiadapiformes, the geologically younger adapoids (50–35 Ma) all demonstrate CL values that are very close to the nonhaplorhine regression line, similar to most living strepsirrhines. However, the adapoids we examined were similar to most plesiadapiformes in seemingly possessing a secondary bony lamina that was about 1/4–1/2 turns long (e.g., Adapis - Fig. 4). The roughly coeval omomyoids illustrate a mixed pattern with one species showing similarities to adapoids while the other species resembles plesiadapiformes in some respects. The smaller and older of the two species, Omomys carteri (45 Ma), had a relatively short cochlea, with values falling below the nonhaplorhine line (Fig. 3A), and demonstrably had a secondary lamina that extended for nearly one full turn (Fig. 4). The larger bodied and geologically younger Microchoerus (35 Ma), on the other hand, produced a CL value that fell slightly above the nonhaplorhine line and had a less developed (although still visible) secondary lamina that was approximately 1/4 turn long.
The 45 Ma Shanghuang petrosal, a possible basal anthropoid, had a short cochlea with relative values similar to most plesiadapiformes (Fig. 3A). In fact, the Shanghuang petrosal had the second shortest cochlea of any taxon in our sample (13.3 mm). The cochlear endocast for this specimen also clearly revealed a secondary bony lamina that goes at least one full spiral turn (Fig. 4). In contrast, the 30–20 Ma unambiguous fossil anthropoids show the longest cochleae of any extinct group examined. Three of these specimens fall just below the haplorhine line, while Tremacebus harringtoni actually falls slightly closer to the nonhaplorhine line (Fig. 3A). However, a previous analysis of Tremacebus suggested that this specimen may be distorted, reducing its apparent length (Coleman et al.,2010). Regardless, none of these specimens displayed indications of a secondary bony lamina.
Using the values for CL and OWA, we estimated measures of low-frequency sensitivity (SPL@250 Hz) and high-frequency sensitivity (SPL@32 kHz) for each of the fossil specimens using the predictive equations presented above (Table 3). These values are illustrated in Figure 5 along with the same audiometric parameters for extant anthropoids and strepsirrhines as well as two “primitive” living placental mammals, treeshrews and hedgehogs.
Table 3. Predicted hearing sensitivity in fossils
Low-frequency (SPL@250 Hz) and high-frequency (SPL@32 kHz) sensitivity predictions and 95% confidence intervals (±) for all extinct taxa investigated in this study. SPL@250 Hz predicted using CL values and SPL@32 kHz predicted using oval window area based on analyses in Coleman and Colbert (2010).
62.4 (± 12.6)
6.5 (± 14.0)
13.0 (± 12.6)
48.1 (± 10.0)
11.5 (± 12.8)
37.2 (± 9.0)
22.4 (± 13.4)
42.0 (± 4.2)
15.8 (± 3.8)
45.8 (± 5.1)
13.8 (± 12.6)
35.5 (± 8.9)
20.7 (± 13.0)
33.3 (± 8.8)
34.8 (± 8.9)
32.4 (± 8.8)
16.4 (± 12.5)
46.1 (± 9.8)
8.2 (± 13.5)
7.9 (± 7.6)
40.9 (± 18.1)
32.1 (± 2.9)
19.8 (± 12.9)
7.3 (± 11.4)
34.2 (± 13.6)
25.0 (± 9.1)
25.9 (± 3.4)
31.6 (± 11.9)
24.1 (± 9.1)
23.2 (± 13.6)
29.9 (± 8.9)
23.6 (± 9.1)
17.5 (± 12.6)
20.9 (± 9.4)
26.8 (± 14.8)
51.1 (± 10.5)
9.1 (± 13.3)
Carpolestes simpsoni and the Shanghuang petrosal are both predicted to have had relatively poor low-frequency sensitivity that was outside the range of living primates but intermediate between the values for treeshrews and hedgehogs. Nannodectes, Omomys, and Pronothodectes were also apparently less sensitive to low-frequency sounds with predicted values of SPL@250 Hz between the values for treeshrews and the upper limits for living strepsirrhines. On the other side of the auditory spectrum, all of these taxa (in addition to Ignacius) appear to have had good high-frequency sensitivity, based on predictions from OWA, that was above the range of any of the living anthropoids that have had their hearing tested. In fact, the predicted values of SPL@32 kHz for Carpolestes, Shanghuang and Omomys were all below 10 dB SPL. In living primates, there is a significant correlation between SPL@32 kHz and high-frequency cutoff (r = −0.833, P = 0.003), and the few euarchontans that have SPL@32 kHz values below 10 dB SPL also have a high-frequency cutoff of 60 kHz or higher (Table 4).
Table 4. Values for SPL@32 kHz and high-frequency cutoff in living euarchontans with known hearing sensitivity
Sound pressure level at 32 kHz (SPL@32 kHz) in decibels and high-frequency cutoff (Hi-Cut) in kilohertz. There is a significant correlation between SPL@32 kHz and high-frequency cutoff in these euarchontan taxa (r = −0.833, P = 0.003). Note that taxa with a value of less than 10 dB for SPL@32 kHz also have a high-frequency cutoff of 60 kHz or higher.
What's more, all of these taxa apparently had relatively well developed secondary lamina that extended between 1/2 and one full spiral turn. The expression of this feature also suggests that these taxa were well adapted to hear high-frequencies based on the observation that Galago senegalensis and Tupaia glis show similar development of the secondary lamina (1/2 and one full turn, respectively) and have among the best known high-frequency sensitivity of any extant euarchontan. The relationship between secondary laminae development and good high-frequency sensitivity in primates is further supported by the recent finding that Tarsiussyrichta has a high-frequency cutoff of ∼75 kHz (Ramsier et al.,2011), which is the highest of any primate tested, and also demonstrates the greatest development of secondary laminae among the primates in our sample (˜two full turns – Fig. 4).
Low-frequency sensitivity in Plesiadapis tricuspidens, P. cookei, Microchoerus, and the three adapoids we investigated appear to have been somewhat better than the fossil taxa described above. The predicted values of SPL@250 Hz are below those for treeshrews and are similar to the middle and lower values displayed by living strepsirrhines but also overlap the upper values for extant anthropoids. In other words, the estimated values are toward the middle of the distribution of values in living forms (Fig. 5). Conversely, the predicted SPL@32 kHz for these taxa suggests reduced high-frequency sensitivity compared with the first group of fossils described and in this case overlaps the range of values for living anthropoids. P. tricuspidens and Microchoerus produced values that are found in all three comparative groups of living taxa, whereas the values for P. cookei and Adapis were higher than those for treeshrews (indicating less sensitivity) but still within the ranges of extant strepsirrhines and anthropoids. These predicted values of SPL@32 kHz plus the apparent presence of moderately developed secondary laminae in the fossils (1/4–1/2 turns) suggests an upper frequency limit (high-frequency cutoff) between ∼41–58 kHz based on modern analogues (Table 4).
The four species of fossil anthropoids we examined all produced predicted values of SPL@250 Hz that are similar to those found in living anthropoids but below the range of strepsirrhines, treeshrews or hedgehogs (Fig. 5). In contrast, the estimated high-frequency sensitivity for this group was less uniform. The predicted value of SPL@32 kHz for Aegyptopithecus falls exclusively within the range for extant anthropoids. In contrast, the value for Dolichocebus also overlaps the range of values for strepsirrhines and the value for Homunculus overlaps the ranges of all three comparative groups. However, considering the finding that secondary bony laminae were not detected in any of these fossils (or in living anthropoids) and the fact that the predicted values were all encompassed by the range of living anthropoids, suggests a high-frequency cutoff between 34 and 45 kHz for this group of fossils.
It should be noted that the predictive equation used to estimate high-frequency sensitivity (based on OWA) has a higher margin of error than the equation used to predict low-frequency sensitivity (based on CL). However, these two traits may in fact be interrelated to some degreee. Although Coleman and Colbert (2010) did not find a significant relationship between high-frequency sensitivity and CL, studies from other researchers have identified such a relationship. Echteler et al. (1994) found that the high-frequency cutoff in mammals goes up as basilar membrane length decreases and Kirk and Gosselin-Ildari (2009) found that smaller cochlear volumes are also related to increased high-frequency sensitivity. Therefore, the general pattern identified here that the primates with smaller (shorter) cochleae also had smaller OWAs strengthens the interpretations for changes in high-frequency hearing based on oval window area alone.
To further investigate evolutionary trends in OWA, CL and development of the secondary bony lamina, we mapped the values for these characters onto a phylogenetic tree and reconstructed the values for key transitional points along the tree not represented by fossils (Fig. 6). One major challenge when constructing a phylogenetic tree of primates that includes the fossils in our sample relates to the phylogenetic position of the Shanghuang petrosal. As briefly described in the Materials and Methods Section, various authors have suggested that the Shanghuang petrosal could be either a basal anthropoid (MacPhee et al.,1995) or a member of the Omomyiformes (Ross and Covert,2000). As revealed by our study, the cochlear endocast of the Shanghuang petrosal is superficially similar to the cochlea of Omomys carteri in several characteristics (Fig. 4). For example, they both have relatively short cochleae (Fig. 3A), have a similar number of cochlear spirals (2 3/4 vs. 2 3/8), and both appear to have a secondary bony lamina that extended for about one turn. In addition, both specimens have similar values for OWA (0.37 mm2 vs. 0.41 mm2). However, they do appear to differ somewhat in the apical height of the cochlear spiral (Fig. 4). Regardless of phylogenetic affinity, the Shanghuang petrosal likely had hearing capabilities that were very similar to that of Omomys based on the predictions in sensitivity presented here. Since the taxonomic identity of the Shanghuang petrosal remains uncertain, we will tentatively consider it to belong to an Omomyiform although we will also discuss the implications in the case that it actually came from a basal anthropoid.
Evolutionary changes in OWA, and presumably high-frequency sensitivity, seem to be largely related to overall increases in body size (Fig. 3B, Fig. 6A). Small mammals (with small heads) need heightened high-frequency sensitivity to take advantage of spectral cues that aid in the ability to localize the source of a sound (Heffner and Heffner,2010). However, as animals get larger, there is likely reduced selection to maintain heightened high-frequency sensitivity for localization purposes. Therefore, if OWA is one of the proximate mechanisms governing high-frequency sensitivity, then it makes sense that increases in body size (and head size) will be paralleled by increases in stapedial footplate size.
In many ways, the loss of secondary bony laminae in some groups may also reflect this trend toward larger body size in mammals. This may explain why the large bodied subfossil lemurs appear to be devoid of secondary laminae while at least some of the relatively smaller living strepsirrhines (e.g., Galago, Eulemur) apparently possess such structures. However, the absence or presence of the secondary laminae is not strictly tied to body size since small-bodied monkeys like tamarins and marmosets apparently lack them while similarly-sized primates like galagos and lorises possess them (Fig. 6B). It also does not appear that the development of secondary laminae is directly related to either CL or the number of cochlear spirals since tarsiers show the greatest expression of laminae development, yet have a high number of cochlear coils and relatively long cochleae, similar to most monkeys and apes which lack laminae. Regardless, the finding that all of the fossils older than 30 Ma in our sample appear to show some development of a secondary bony lamina (Fig. 6B) supports the notion that the presence of this structure is the ancestral condition for the cladotherian clade (Ruf et al.,2009).
When examining changes in CL, some interesting phylogenetic and temporal patterns become evident (Fig. 6B). Note that only the geologically youngest fossils show CL values that approach those found in living forms. Although increases in body size through time have no doubt resulted in overall increases in CL, changes in body size alone cannot explain these patterns. For example, Plesiadapis cookei, the plesiadapiform with the longest cochlea (19.0 mm) and among the largest in body mass (∼2,059 g) in our sample, still possessed a cochlea that was shorter than a modern species like an owl monkey (Aotus trivirgatus – 22.4 mm), which is less than half the body mass (∼775 g). In addition, the geologically older omomyoid Omomys had a relatively shorter cochlea than the geologically younger omomyoid Microchoerus (Fig. 3A). Furthermore, if the Shanghuang petrosal is that of a basal anthropoid (or basal haplorhine for that matter), then there is a large difference in relative CL between this 45 Ma specimen compared with the 30–20 Ma fossil anthropoids we investigated from Africa and South America.
The reconstructed values at the branch leading to all primates (1 – Fig. 6) for OWA was 0.66 mm2 and that for CL was 17.9 mm. These values suggest that basal primates had low-frequency sensitivity that was intermediate between the mean values for extant strepsirrhines and treeshrews and high-frequency sensitivity similar to living strepsirrhines (Fig. 7). Considering the Shanghuang petrosal to belong to a stem anthropoid only moderately influences these reconstructed values (0.63 mm2 and 17.5 mm) and consequently does not significantly alter this interpretation. The reconstructed values at the branch leading to strepsirrhines (2 – Fig. 6) was 0.77 mm2 for OWA and 19.6 mm for CL. These values suggest slightly less high-frequency sensitivity but slightly better low-frequency hearing than found in living strepsirrhines (Fig. 7). Again, evaluating this pattern with Shanghuang as a basal anthropoid had little effect on the reconstructed values (0.74 mm2 and 19.3 mm). The reconstructed values at the branch leading to haplorrhines (3 – Fig. 6) were 0.69 mm2 and 18.7 mm suggesting both high- and low-frequency sensitivity that was very similar to the mean for the extant strepsirrhines for which hearing sensitivity has been tested (Fig. 7). As before, there are only minor changes if Shanghuang is positioned as a basal anthropoid (0.64 mm2 and 18.1 mm).
In contrast to the minimal influence of the Shanghuang petrosal on the reconstructed values at the branches labeled 1, 2, 3 discussed above, its phylogenetic position does significantly alter the interpretations associated with the branch leading to anthropoids (4 – Fig. 6). If the Shanghuang petrosal was that of an Omomyiform, then the reconstructed values at the branch labeled 4 were 1.12 mm2 and 24.4 mm for OWA and CL, respectively. This suggests that the relatively long cochleae characteristic of anthropoids and tarsiers is a shared derived trait with an origin that likely stretches back to the diversification of these two groups. However, if the Shanghuang was actually from a basal anthropoid, then the values at branch 4 are reconstructed to have been 0.70 mm2 and 18.2 mm, similar to the inferred values for basal haplorrhines. This scenario implies that the relatively long cochleae of tarsiers developed independently to that of anthropoids. Regardless of when anthropoids and tarsiers began to develop their relatively long cochleae, it appears that their early ancestors (basal haplorhines) were characterized by reduced low-frequency and increased high-frequency sensitivity compared with modern and more recent species. The earliest fossil evidence for hearing sensitivity that is on par with that of living haplorrhines does not appear until 30–20 Ma as witnessed in definitive fossil anthropoids like Aegyptopithecus and Dolichocebus (Fig. 5).
The pattern of increased cochlear coiling and elongation documented in Cretaceous mammals (see introduction) is continued in the geologically younger fossil primates examined in this study. Comparable with the Bug Creek Anthills specimens, Carpolestes simpsoni from the late Paleocene had about 1 1/2 spiral turns, similar to living hedgehogs that also display 1 1/2 turns of the cochlea. The predicted low-frequency sensitivity for C. simpsoni (also similar to the known low-frequency sensitivity of hedgehogs) implies that this animal likely could not have perceived sounds much below 250 Hz (based on the traditional low-frequency cutoff of 60 dB). A recent study of cochlear volume in this specimen also suggests that C. simpsoni had a hearing range that was shifted toward higher frequencies (Armstrong et al.,2011).
The similarly aged Pronothodectes gaoi (middle-late Paleocene) and Nannodectes intermedius (late Paleocene) had cochleae that spiraled for 2 1/8–2 3/8 turns and ranged in length from 14.5–15.6 mm. The slightly younger Plesiadapis tricuspidens and Plesiadapis cookei (late Paleocene-early Eocene) had essentially the same number of spirals (2 1/4–2 3/8) but slightly longer cochleae (17.0–19.0 mm). The three early-late Eocene adapoids we investigated show a slight increase in cochlear turns (2 1/4–2 5/8) and CL (19.7–20.6) compared with Plesiadapis spp. The omomyid, Omomys carteri from the middle Eocene displays 2 3/4 spirals but a relatively short length of 15.4 mm compared with the late Eocene-early Oligocene Microchoerus sp. that exhibits a cochlea with 3 1/8 turns and a length of 21 mm. Finally, the fossil anthropoids in our sample show cochlear characteristics that are within the range of extant anthropoids (Table 2). Aegyptopithecus zeuxis (early Oligocene) had 2 7/8 turns (25.7 mm) and fossil platyrrhines (early Miocene) had 2 3/4–3 1/8 turns (22–24.5 mm) (Coleman et al.,2010).
Our results provide paleontological evidence that basal primates (late Cretaceous) had good high-frequency sensitivity but relatively poor low-frequency sensitivity compared with modern members of the order. Then, they began to develop good low-frequency sensitivity with slight decreases in high-frequency sensitivity starting during the late Paleocene-early Eocene. Finally, essentially modern patterns had evolved by the Oligocene (∼30 Ma), although slight modifications to the entire auditory apparatus appear to have continued through the early part of the Miocene (Coleman et al.,2010).
These findings, in combination with other studies on geologically older mammals, are in accordance with the hypothesis articulated by Masterson et al. (1969) that mammals went through a period of reduced low-frequency sensitivity during the evolution of modern hearing patterns, although the timing of events are different than originally proposed. This purported sequence of changes in hearing sensitivity has considerable implications for the evolution of vocal communication, predator-prey interactions and the development of hearing specializations in mammals. For example, the transition to a primarily high-frequency hearing pattern was likely paralleled by a shift to higher vocalization frequencies making it improbable that early primates (and possibly mammals as well) used low-frequency, long-range communication signals like those that are utilized by many species of primates today. This also raises the possibility that at least some early primate (and mammalian) vocalizations were above the upper limit of hearing of contemporaneous nonmammalian predators (i.e., dinosaurs), resulting in opportunities to exploit new behavioral and ecological niches and potentially altering predator-prey dynamics. Furthermore, the ancestral hearing phase of good high-frequency and poor low-frequency sensitivity (∼90–45 Ma) in primates may be typical of other mammalian orders and would have been a critical first step toward developing the unique hearing patterns like those of echolocating bats. In fact, low-frequency sensitivity in many bats is very similar to that in opossums (Fig. 1) suggesting that the “specialized” hearing of bats may not be that far removed from the pattern that characterized mammals during the reduced low-frequency sensitivity phase.
SUMMARY AND FUTURE DIRECTIONS
Plesiadapiformes, the earliest fossil primates for which we have evidence were characterized by having small oval window areas and relatively short cochlea that housed moderate to well-developed secondary bony laminae. These traits are interpreted to suggest that these taxa had good high-frequency but relatively poor low-frequency sensitivity, somewhat intermediate between extant strepsirrhines and primitive living mammals like treeshrews and hedgehogs. Then, with the origin of haplorrhines and strepsirrhines (Euprimates), primates began to develop relatively longer cochleae and reduced expression of secondary bony laminae indicating an increase in low-frequency sensitivity and modest reductions in high-frequency sensitivity. This “strepsirrhine” stage of hearing dates back to at least the Eocene (∼50 Ma) based on fossil taxa like adapoids but could date well back into the late Cretaceous based on molecular evidence.
Finally, sometime after the origin of Euprimates, haplorrhines continued the pattern of cochlear elongation and reduction (or loss) of secondary bony laminae (except in tarsiers). The African and South American fossil evidence suggest that this process was completed by the early Miocene, but the origin of this stage of primate hearing remains unresolved. Early fossil haplorrhines like Omomys and the Shanghuang petrosal suggest that haplorrhines and strepsirrhines (and possibly anthropoids and tarsiers as well) may have developed cochlear elongation independently, relative to the basal primate condition. Among living haplorrhines, only tarsiers have developed relatively long cochleae while still retaining well developed secondary bony laminae—adaptations that apparently confer both heightened high- and low- frequency sensitivity to this unique genus.
Paleontological evidence and comparative studies on auditory function promise to continue refining our understanding of hearing evolution. Investigating more primate specimens from the Eocene and Oligocene will shed more light on the evolution of the hearing patterns in primates and help put these findings in a larger theoretical framework. In particular, analyzing the auditory region of (definitive) basal anthropoids could help us understand the development of the unique traits of this group of primates (to which humans belong). It will also be interesting to begin examining other orders of mammals (e.g., rodents, carnivores) from the early Cenozoic to see if similar patterns of auditory evolution have occurred. Ultimately, this type of information contributes to a more complete understanding of the environmental processes that have resulted in the behavioral and ecological diversity seen among primates, mammals, and other vertebrate groups.
We thank the following individuals and institutions for access to fossil CT data: J.I. Bloch (U. Florida) and M.T. Silcox (U. Winnipeg) – Carpolestes, Ignacius; H.H. Covert (U. Colorado) – Microchoerus, Omomys; R. Emry (Smithsonian) – Nannodectes; R.C. Fox (U. Alberta) – Pronothedectes; P.D. Gingerich (U. Michigan) – P. cookei; M. Godinot (National d'Histoire Naturelle, Paris) – Adapis, P. tricuspidens; R. F. Kay (Duke U.) – Dolichocebus, Homunculus, Tremacebus; E.R. Seiffert (Stony Brook U.) – Aegyptopithecus; E.L. Simons (Duke U.) – Aegyptopithecus, Archaeolemur, Babakotia, Megaladapis, Mesopropithecus, Palaeopropithecus; A. Walker (Pennsylvania State U.) – Notharctus and Smilodectes (used with permission of the Division of Paleontology, AMNH), Adapis, P. tricuspidens; K.C. Beard (Carnegie) – Shanghuang petrosal. J. Rossie provided scans for Loris and Saguinus (NSF BCS-0100825). The following people also helped with scanning and obtaining CT data: M. Colbert, L. Gordan, T. Ryan, A. Walker, E. Westwig. We would also like to acknowledge B. Demes, J. Georgi, A. Grossman, C. Heesy, W. Jungers, J. Rosowski, and C. Ross for helpful discussions and comments about the analysis and early versions of the manuscript. Lastly, we thank two anonymous reviewers for useful comments on the manuscript.
†Technically, the correct designation for the infraorder consisting of New World monkeys + Old World monkeys + Apes may be Simiiformes (Groves, 2005), but in this article we use the traditional designation of Anthropoidea to avoid confusion with the recent literature involving this group.
APPENDIX: OVAL WINDOW AREA—STAPEDIAL FOOTPLATE AREA FOR EXTANT EUARCHONTANS
OWA-SFA estimates based on Coleman et al. (2010) unless indicated by asterisks. Taxa noted with one asterisk (*) indicate individual species values that were previously published as the mean for the genus, those with two asterisks (**) indicate updated values based on additional specimens, and those with three asterisks (***) indicate previously unpublished values.