The materials used for this study consist of a right stapes of Glossotherium robustum (MNHN 914) and a left stapes of Lestodon sp. (MNHN 2149) deposited in the Palaeontological Collection of the MNHN.

Both stapes are very similar in size (Fig. 2) and do not show much novelty in morphology in comparison with other taxa of Mylodontidae (genera Mylodon, Paramylodon, Scelidotherium and Alancitherium). They are robust with a strong muscular process and are thickened dorsoventrally (Guth, 1956 and Woodward, 1900).

As in all Mylodontidae and the genus Megalonyx the two stapes described here possess a large stapedial foramen. This foramen does not exist (or is very reduced) in adults of the living genera Bradypus and Choloepus (Patterson et al., 1992). Like other taxa placed among Mylodontidae, the general morphology of MNHN 914 and MNHN 2149 resemble the typical stapedial morphology found in the anteaters Myrmecophaga and Tamandua (Patterson et al., 1992).

The major and minor diameters of the stapedial base of the Glossotherium robustum (MNHN 914) and Lestodon sp. (MNHN 2149) were measured with a digital vernier caliper. The footplate area was estimated, assuming an elliptical shape.

We found estimates for the low-frequency limit, the best sensitivity frequency and the high-frequency limit by using the corresponding power function fit to audiometric limits and anatomical dimensions from Rosowski (1992). The power function fit equations used here are the following: •

lower-frequency = 0.40 (Footplate Area)^−1.1;

Frequencies are in kHz and areas are in mm². These predictive equations were obtained from data from specimens with a broad range of body masses from tree shrew (Tupaia glis) to horses (Equus caballus) and explain 67% of the variance for low-frequency limit, 32% for best frequency, and 68% for high-frequency limit (Rosowski, 1992). The best frequency was defined as the tone frequency at the best threshold (the best threshold is the lowest sound pressure of a tone of any frequency that produced a positive behavioral response). The limits of hearing mentioned by Rosowski (1992) were defined by the intersection of the audiogram, with an isopressure contour defined by the best threshold plus 30 dB. That definition of limits is slightly arbitrary and for Homo sapiens the lower-frequency value measured is 150 Hz (other definitions imply much lower values as 20 Hz) with best frequency at 4000 Hz. These values are useful for comparison with the results in sloths and to define infrasound sensitivity. The lower-frequency limit in Rosowski's experimental data is roughly in the infrasound level (90 Hz for Chinchilla laniger).

In order to estimate the body size expected differences in tympanum and footplate areas we use the following allometric equations for mammals (Hunt and Korth, 1980): Y = 0.26     X + 1.23 , \[ Y=0.26\text{\hspace{0.28em}}X+1.23\text{,} \] where Y is the base 10 logarithm of the tympanum area in mm² and X is the base 10 logarithm of the body mass in kg. Y = 0.28 X–0.07 where Y is the base 10 logarithm of the stapedial footplate area in mm² and X is the base 10 logarithm of the body mass in kg.

The major and minor diameters of the stapedial base of Glossotherium robustum (MNHN 914) were 3.84 mm and 2.46 mm, respectively, and those of Lestodon (MNHN 2149) were 3.24 mm and 2.40 mm. These values give footplate areas of 7.42 mm² and 6.11 mm² respectively. The footplate area is slightly larger in the smaller Glossotherium. However, the expected values of the footplate area from allometric equations are 6.60 mm² and 8.74 mm² respectively. The tympanum areas expected from allometric equations are 113.7 mm² and 147.7 mm². In a previous work it was determined that in Glossotherium the tympanum area was much larger than expected, that is in the range 180 to 208 mm² (Blanco and Rinderknecht, 2008).

Using the corresponding power function fit to audiometric limits and anatomical dimensions from the literature (Rosowski, 1992), we found that the low-frequency limits are in the infrasound region and the best sensitivity frequencies and the high-frequency limits are relatively low values (Table 1). The results for Glossotherium are very close to those obtained by Blanco and Rinderknecht (2008) and for Lestodon the results are very close to the expected results mentioned above.

The body mass of Glossotherium was estimated at 1500 kg, clearly smaller than the body mass of 4100 kg estimated for Lestodon, suggesting that the ossicle's size, almost the same in both genera, is probably related to the detection of a specific range of frequencies and not strictly dependent on body size. The results show that for a mammal with the same body mass the expected tympanum areas are roughly 60% of the value previously measured in Glossotherium (Blanco and Rinderknecht, 2008). Also, the results show that for mammals with those body masses the expected difference in stapedial footplate areas is larger than the measured values (see results section). Thus, the similarities in frequency ranges are not related to differences in body mass.

In both genera, the ossicles are very large and in a previous work Blanco and Rinderknecht (2008) suggested that they were probably designed to detect airborne sound transmission or low-frequency seismic waves. The main argument proposed was that these large ossicles, which imply loss of acuity for hearing high frequencies, were useful for better localization of the sound source (Heffner and Heffner, 1992) and could imply some advantage. Mammals with small heads need to hear higher frequencies than larger mammals in order to use binaural and monaural spectral clues for sound localization. That is reflected in the correlation (r = –0.84) between maximum interaural distance and the high-frequency limit (Heffner and Heffner, 1992). If it is assumed that15,000 Hz corresponds to the high-frequency limit for the studied ground sloths, the observed correlation between functional head size and high-frequency hearing limit (Fig. 34,7 in Heffner and Heffner, 1992) gives a rough estimation of the expected interaural distance for good sound localization. In that case the expected interaural distance is around 3000 μs (from data in Heffner and Heffner, 1992). Assuming a sound speed of 340 m/s the interaural distance is 1 m. The anatomical interaural distances measured in the fossil skulls with a digital caliper are 114 mm in Glossotherium MNHN 441, 93.5 mm in Glossotherium MNHN 914, 147.6 mm in Lestodon MNHN 1637, and 144.5 mm in Lestodon MNHN 1638. In both genera, the interaural distances are low for mammals of such body size. Elongated and thin skulls are characteristic of mylodontid ground sloths (Pascual, 1967 and Paula Couto, 1979). Even in the larger skull of Lestodon, the interaural distance in time units (that means the interaural distance divided by sound speed) is around 440 μs and in the larger skull of Glossotherium is around 320 μs. For these values of interaural distance, the general tendency in Figure 34,7 in Heffner and Heffner (1992) implies a higher-frequency limit for good sound localization around 40 kHz. In the most extreme case Heffners’ figure 34,7 shows extant animals with head sizes similar to the two extinct sloths with high-frequency limits of 20 kHz. Even in that case the value is larger than the high-frequency limits obtained here.

The combination of frequency range shifted toward low tones and the small interaural distances of the ground sloths studied here suggests a very poor performance in sound localization. Generally, sound localization is used to enable an animal to direct its gaze to the source for further scrutiny. In a case where this function is not relevant for fitness, low-frequency sensitivity can be selected for. As Heffner and Heffner (1992) pointed out, it would be expected that any species of mammals that could not localize sound would lose its ability to hear high frequencies and this seems to be the case of fossorial mammals such as pocket gophers and blind mole rats that detect low-frequency seismic signals.

In a previous biomechanical study (Bargo et al., 2000), it was suggested that Lestodon, Glossotherium, and Scelidotherium (body mass estimated of around 800 kg) were well adapted to perform strenuous activities with their forelimbs, in which force is enhanced over velocity, useful for functions such as digging. Large Pleistocene caves are known in the Pampean region with a size consistent with the size of Glossotherium and Scelidotherium scratches found on the walls and roofs of some caves agree well with the manus morphology of both genera (Vizcaíno et al., 2001 and Zárate et al., 1998). Fossorial habits are a possible explanation for the utility of low-frequency sensitivity in this group.

In the case of some desert rodents an increased low-frequency sensitivity can also be related to hearing low-frequency sounds produced by some nocturnal predators such as snakes and owls just prior to striking Webster and Plassmann, 1992 and references therein). The detection of sound clues from specific predators of ground sloths, perhaps the larger ones as Smilodon or Arctotherium, is another possibility. However, this hypothesis has no support at present because there is not enough knowledge about the predator-prey relationships of ground sloths (Spencer et al., 2003).

Another possibility is related to an adaptation for semi-aquatic habits. Large ossicles in some living mammals with semi-aquatic habits appear to be an adaptation for underwater hearing (Reuter and Nummela, 1998 and references therein; Shipley et al., 1992). A semi-aquatic sloth (Thalassocnus spp.) was discovered in the Miocene-Pliocene of Perú and Chile (Muizon and de McDonald, 1995 and Muizon et al., 2004) but it was not closely related with Mylodontidae (Gaudin, 2004) and all the paleontological evidence shows that Mylodontidae were completely terrestrial animals (Bargo and Vizcaíno, 2008 and Paula Couto, 1979).

As previously suggested by Blanco and Rinderknecht (2008), low-frequency sensitivity could be an adaptation for long-range communication as observed in living elephants. Low-frequency sound is useful for long-range communication because it is less affected by scattering from vegetation, and atmospheric absorption is essentially nonexistent (Garstang et al., 1995 and references therein). It was claimed that infrasound with frequencies from 14 Hz to 35 Hz is used for communication at ranges up to 4 km (Garstang et al., 1995). However, recent analyses (McComb et al., 2003) suggest that the most important frequency components for long-distance communication of social identity may be well above the infrasonic range and the authors have doubts about the use of airborne sound much below 100 Hz for long-distance hearing. The present result implies that ground sloths have good sensitivity at such low frequencies (the lower-frequency limit is close to 50 Hz in both Glossotherium and Lestodon). Present results and the use of low-frequency sounds for long-range communication in living large mammals such as elephants suggest the possibility that ground sloths could also have used low-frequency sounds for long-range communication.

The new results seem to support previous speculations that both genera have similar frequency ranges. From all the previous discussion it is possible to conclude that there probably existed in some ground sloths an adaptive selection for some specific low-frequency window. The selective pressure for such adaptation is unknown, but detection of specific predators, fossoriality, long-range communication, and seismic wave detection, must be considered. The long-range communication hypothesis is probably the best suited at present to explain the size-independent low-frequency sensitivity in some ground sloths.