Research Article

Morphology and function of Neandertal and modern human ear ossicles

See allHide authors and affiliations

PNAS first published September 26, 2016; https://doi.org/10.1073/pnas.1605881113

  1. Edited by Richard G. Klein, Stanford University, Stanford, CA, and approved August 3, 2016 (received for review April 12, 2016)

Significance

Middle ear ossicles are critical for audition and rarely preserved in fossils. Based on microcomputed tomography images, our comparative 3D shape analysis of Neandertal ossicles shows striking shape differences between Neandertals and anatomically modern humans (AMHs). However, these morphological differences do not affect the functional properties of the ossicles, potentially indicating consistent aspects of vocal communication in Neandertals and AMHs. Instead, a strong relationship between ossicle morphology and tympanic cavity architecture is found.

Abstract

The diminutive middle ear ossicles (malleus, incus, stapes) housed in the tympanic cavity of the temporal bone play an important role in audition. The few known ossicles of Neandertals are distinctly different from those of anatomically modern humans (AMHs), despite the close relationship between both human species. Although not mutually exclusive, these differences may affect hearing capacity or could reflect covariation with the surrounding temporal bone. Until now, detailed comparisons were hampered by the small sample of Neandertal ossicles and the unavailability of methods combining analyses of ossicles with surrounding structures. Here, we present an analysis of the largest sample of Neandertal ossicles to date, including many previously unknown specimens, covering a wide geographic and temporal range. Microcomputed tomography scans and 3D geometric morphometrics were used to quantify shape and functional properties of the ossicles and the tympanic cavity and make comparisons with recent and extinct AMHs as well as African apes. We find striking morphological differences between ossicles of AMHs and Neandertals. Ossicles of both Neandertals and AMHs appear derived compared with the inferred ancestral morphology, albeit in different ways. Brain size increase evolved separately in AMHs and Neandertals, leading to differences in the tympanic cavity and, consequently, the shape and spatial configuration of the ossicles. Despite these different evolutionary trajectories, functional properties of the middle ear of AMHs and Neandertals are largely similar. The relevance of these functionally equivalent solutions is likely to conserve a similar auditory sensitivity level inherited from their last common ancestor.

The hominin fossil record can only provide indirect information about auditory capacities of our extinct relatives. Inferences about the evolution of the human sense of hearing require understanding of the interplay between form and function in extant species. When auditory capacities are visualized as audiograms, plotting the sensitivity for different frequencies, anatomically modern humans (AMHs) differ from the W-shaped pattern found in most anthropoid primates. AMHs are characterized by a drastically lowered high-frequency cutoff and maintaining high sensitivity in the low to midfrequencies, resulting in a U-shaped audiogram (1⇓⇓⇓⇓⇓–7). In primates, such hearing variability is assumed to be partly related to forms of vocalization and habitat acoustics (8⇓–10). Diverse hearing capabilities are also related to the morphology of the diminutive middle ear ossicles housed in the tympanic cavity (11, 12). The malleus, incus, and stapes form the ossicular chain that connects the tympanic membrane to the oval window of the inner ear. These bones play an important role in audition by amplifying and regulating the sound waves transmitted to the cochlea (11, 13⇓–15). In particular, the middle ear acts as a transformer that matches the impedances between the air and the perilymph of the cochlea (16), participating in the tuning of the sensitivity levels.

Recent analyses have emphasized a distinctly derived morphology of the ossicles of AMHs compared with extant great apes (17, 18), suggesting that the ossicles of extinct hominins may provide insights into the origin of the distinct auditory capacities of AMHs.

Until recently, only a few isolated Neandertal ossicles were known, and their morphology differs consistently from AMH ossicles (19⇓⇓⇓⇓⇓–25). Because the external acoustic meatus and cochlea have nearly identical dimensions in AMHs and Neandertals (26⇓–28), such shape differences of the ossicles could indicate differences in auditory capacities and with it, potential implications for habitat preference and aspects of vocal communication. However, the temporal bone housing the ossicles is well-known to differentiate Neandertals from AMHs, and some of its structures express morphological covariation (29, 30). Therefore, differences in ossicle morphology between Neandertals and AMHs could also reflect variation in the spatial relationship of the ossicles within the tympanic cavity because of differences in the placement of either the oval window or the tympanic membrane (24).

Comparative and functional investigations of Neandertal ossicles were previously limited by the small sample size. High-resolution computed tomography (CT) has made it possible to study the morphology of ossicles trapped in the tympanic cavity. Here, we analyze the largest sample of Neandertal ossicles to date (22 ossicles from 14 individuals), covering a wide geographic and temporal range. We compare these fossils with Holocene and Pleistocene AMHs and African apes and investigate how previously observed ossicle characteristics fit into the Neandertal bauplan and how it evolved. Studying ossicles is methodologically challenging because of their small size and complex 3D shapes. To quantify ossicle shape, we apply a 3D geometric morphometric approach (18). We also test how differences in ossicle morphology affect the impedance matching function of the middle ear. As a functional measure, we chose an ideal transformer ratio (ITR), namely the pressure gain, which is the product of the area ratio between the tympanic membrane and stapes footplate and the lever ratio of the malleus and incus functional lengths (31⇓–33). Finally, we analyze how ossicle shape covaries with the surrounding tympanic cavity, looking into the angular orientation and distance between the tympanic sulcus and oval window as well as their respective sizes.

Results

Shape of Middle Ear Ossicles.

Malleus.

In 3D shape space, Neandertals and Holocene AMHs form nonoverlapping clusters fully separated from African apes (Fig. 1A); means of all groups differ significantly (P < 0.01). Average shape distance between Holocene AMHs and Neandertals (0.120) exceeds distances between AMHs and Pan troglodytes (0.099) and between P. troglodytes and Gorilla gorilla (0.098). The mean shape of Pleistocene AMHs is closer to that of Neandertals (0.104) than the mean shape of Holocene AMHs. Pleistocene AMH specimens fall within the range of variation of Holocene AMHs. Compared with AMHs, Neandertals possess mallei with a shorter manubrium, a larger and more anterior–posterior flattened head, a bigger articular facet with a less developed medioinferior part, and a distinctively wide angle between the manubrium and the head (Fig. 1D and Tables S1 and S2).

Fig. 1.
Fig. 1.

Principal component (PC) analysis of the ossicle (semi)landmark sets in shape space. (A) Malleus. (B) Incus. (C) Stapes. Variance explained by the PCs is listed in Table S2. Mean shapes of the (D) malleus, (E) incus, and (F) stapes of AMHs, Neandertals, P. troglodytes, and G. gorilla. Scale has been standardized by centroid size.

View this table:
Table S1.

Mean scores of the first three PCs of AMHs and Neandertals, which are weighted by fossil ages, and estimated mean scores for the LCA of the Homo–Pan dichotomy and the AMH–Neandertal dichotomy

View this table:
Table S2.

Shape changes of the malleus, incus, and stapes associated with the first three PCs

Incus.

Neandertals and AMHs form nonoverlapping clusters in 3D shape space and do not overlap with African apes (Fig. 1B). Mean shapes of all groups differ significantly (P < 0.01). Average shape distance between AMHs and Neandertals (0.128) is slightly smaller than that between AMH and Pan (0.136), AMH and G. gorilla (0.136), and between Pan and G. gorilla (0.135). Mean shape of Pleistocene AMHs is closer to that of Neandertals (0.117) than mean shape of Holocene AMHs. Pleistocene AMHs fall within the range of variation of Holocene AMHs. Compared with AMHs, Neandertals possess incudes with a shorter intercrural distance, a deeper and less symmetrical intercrural curvature, a distinctively straight long process, and a distinctively large articular facet (Fig. 1E and Tables S1 and S2).

Stapes.

Neandertals separate from all other species, whereas AMHs overlap with P. troglodytes (Fig. 1C). Mean shapes of African apes do not differ significantly, but those of Neandertals and AMHs do (P < 0.05). Average shape distance between Neandertals and AMHs (0.152) is higher than that between AMHs and P. troglodytes (0.063). Neandertals separate from AMHs by possessing stapedes with more asymmetrical and long crura, an anterior–posteriorly shorter but inferior–superiorly broader footplate, a distinct angle between the footplate and the crura, and a head facing more anteriorly than in any other species investigated (Fig. 1F and Tables S1 and S2).

Ancestral State Estimation of AMH and Neandertal Ossicle Shapes.

The Euclidean distances between the estimated ancestral scores and the respective mean scores of AMHs and Neandertals in the space of the first three principal components show that the stapes mean shape is more derived in Neandertals than in AMHs compared with their last common ancestor (SI Text, Text S1, Fig. S1, and Table S1). Likewise, malleus and incus shape has significantly changed in both AMHs and Neandertals. Compared with the inferred morphology of the last common ancestor of Homo and Pan, the Neandertal mean shape is consistently more derived than that in AMHs (Fig. S1).

Fig. S1.
Fig. S1.

Two-dimensional shape space of the first two PCs of the (A) malleus, (B) incus, and (C) stapes showing the position of the PC mean scores of P. troglodytes, AMHs, and Neandertals (large colored dots). The PC mean scores of AMHs and Neandertals include weighted fossil ages. The black dots show the positions of the estimated mean scores for the LCA of the Homo–Pan dichotomy and the AMH–Neandertal dichotomy. The lines give a graphical presentation of the distances between the AMH and Neandertal mean shapes and the respective estimated ancestral state (dashed lines, Homo–Pan; solid lines, AMH–Neandertal; another solid line also graphically shows the distance between the LCA of the Homo–Pan and AMH–Neandertal dichotomies and between Homo–Pan and the mean of Pan). The numbers associated with the lines are the calculated Euclidean distances of the first three PCs showing that AMH and Neandertal mean shapes are approximately similarly derived from their estimated LCA regarding malleus and incus but that the Neandertal stapes is more derived. In contrast, when considering the Homo–Pan dichotomy, Neandertals are consistently more derived from the estimated ancestral state.

Functionally Important Factors and Ratios of the Middle Ear.

All species differ significantly (P < 0.05) in the functional areas ratio between the areas enclosed by the tympanic sulcus and the stapes footplate and the functional lengths ratio (lever ratio) between the malleus and the incus (Fig. 2), except when comparing the functional lengths ratio (P = 0.425) of AMHs with that of Neandertals. All species also differ significantly (P < 0.05) in the ITR, an estimation of the gain in sound energy that is achieved by the middle ear, except when comparing AMHs with Neandertals (P = 0.960) and P. troglodytes with G. gorilla (P = 0.480) (Table S3).

Fig. 2.
Fig. 2.

Box–whiskers plots of the (A) functional areas ratio, (B) functional lengths ratio, and (C) ITR of AMHs, Neandertals, P. troglodytes, and G. gorilla. Sample sizes for each parameter are as follows: (A) AMH n = 54, Neandertals n = 10, P. troglodytes n = 14, and G. gorilla n = 10; (B) AMH n = 27, Neandertals n = 5, P. troglodytes n = 9, and G. gorilla n = 8; and (C) AMH n = 27, Neandertals n = 5, P. troglodytes n = 9, and G. gorilla n = 8.

View this table:
Table S3.

P values obtained from Mann–Whitney–Wilcoxon tests of the comparisons of the middle ear and functional parameters of the specimens possessing both the tympanic sulcus and the oval window

Linking Middle Ear Morphometrics to Ossicle Shape.

Compared with the African apes, AMHs and Neandertals share a smaller tympanic sulcus area and a larger middle ear length. Oval window area is less variable across the entire sample, with significant differences only between AMH and P. troglodytes and between G. gorilla and P. troglodytes (Tables S3 and S4). The two human groups differ significantly in the angles between the middle ear axis and the planes of the tympanic sulcus and oval window (Tables S3 and S4). Neandertals show distinctly smaller angles compared with AMHs, particular when considering the angle of the plane of the oval window and the middle ear axis. The differences in the spatial relationship between the tympanic sulcus and oval window are best visualized when looking at the oval window from a perspective parallel to the plane of the tympanic sulcus (Fig. 3). Here, the position of the oval window relative to the tympanic sulcus is more eccentrically in Neandertals. Table 1 lists middle ear parameters that explain significant amounts (P < 0.01) of ossicle shape. Middle ear length has the highest influence on the shape of the incus and the stapes, whereas malleus shape is most strongly affected by the area enclosed by the tympanic sulcus. Both angles, measured between the middle ear axis and the planes of the tympanic sulcus and oval window, explain a significant amount of the shape variance of each of the three ossicles.

Fig. 3.
Fig. 3.

Surface reconstructions of the right ear region of (A) an AMH (University of Leipzig Anatomy collection 58) and (B) a Neandertal (La Chapelle-aux-Saints 1; original left ear mirror imaged) showing the orientational/positional differences between the planes of the tympanic sulcus (red landmarks) and oval window (blue landmarks). The perspective of this view is parallel to the plane of the tympanic sulcus, and the long axis of the oval window is oriented horizontally.

View this table:
Table 1.

Shape changes of the malleus, incus, and stapes associated with changes in the computed middle ear parameters and shape variance explained by these middle ear parameters

View this table:
Table S4.

Mean value and range of the areas enclosed by the tympanic sulcus and the oval window, middle ear length, and the angles between the middle ear axis and the planes of the tympanic sulcus and the oval window of AMHs, Neandertals, P. troglodytes, and G. gorilla

Discussion

Our results show striking differences between Neandertal and AMH ossicles. Previously shown metric overlap of these groups (24) likely reflects allometric scaling caused by similarities in body size (34, 35). As in the skull (36, 37), the amount of shape distance between AMHs and Neandertals is comparable with (incus) or even exceeds (malleus and stapes) the distances between AMHs and African apes. Hence, just like the inner ear (38, 39), the ossicles are valuable taxonomic discriminators of late Pleistocene fossil human remains.

Despite distinct differences in ossicle morphology, functionally relevant parameters of the ossicles and the surrounding middle ear structures are largely similar between AMHs and Neandertals, particularly compared with the African apes (Fig. 2 and Tables S3 and S4). At first, this similarity seems surprising, because ossicle shape differences between AMHs and Neandertals affect the biomechanical characteristics of the ossicles when analyzed individually. Particularly, the more open angle between the manubrium and the head seen in the Neandertal malleus causes the bones’ centers of mass to shift more superiorly, thereby increasing its functional length. However, we also found an increase in functional length of the incus that is of similar magnitude. Consequently, the ITR as an estimate for the impedance matching function by the middle ear is nearly identical between AMHs and Neandertals.

ITRs are simple but straightforward estimates for understanding the impedance matching function of the middle ear (33, 40). Although ITRs do not result in an accurate prediction of auditory sensitivity levels of the entire hearing range, they can be seen as an approximation for the pressure gain achieved by the middle ear at lower frequencies (33). An increase in such theoretical measures has been shown to be correlated with an increase in experimentally measured values when comparing a number of mammals (16). At higher frequencies (>10 kHz), ITRs are less reliable, because the rotational behavior of the ossicular chain becomes more complex (41, 42). The impedance matching function of the middle ear is informative about the sensitivity level but not the frequency range of hearing, because the latter depends strongly on other parameters and structures, like the cochlea (43, 44). In light of nearly identical dimensions of the external acoustic meatus and cochlea in AMHs and Neandertals (26⇓–28), our data, thus, show no support for differences in hearing capacities between AMHs and Neandertals. This finding corroborates recent studies showing similar auditory capacities between AMHs and fossils from Atapuerca (45, 46), which are considered to be close to the root of the Neandertal lineage (47). It has been shown that primate species living in different habitats (e.g., rainforest canopy and open landscapes) differ in auditory capacities, and it is likely that habitat acoustics influence vocalization and audition (8⇓–10). Our findings, thus, potentially point to shared aspects of vocal communication in AMHs and Neandertals.

The temporal bone that houses the ear ossicles is known to discriminate between AMHs and Neandertals and among hominids in general (30, 39, 48). It has been shown that other ear structures, like the bony labyrinth, covary with changes in the surrounding temporal bone architecture (29). Our data show a number of ossicle shape changes that are correlated to variation in tympanic cavity morphology (Table 1). Different ossicle shapes might, therefore, indicate a covariation with the temporal bone. Compared with the African apes, AMHs and Neandertals share similar-sized areas enclosed by the tympanic sulcus and an increase in middle ear length. The majority of the ossicle shape aspects shared by AMHs and Neandertals are affected by changes in these linear dimensions.

The distinct ossicles shapes of Neandertals and AMHs suggest differences in their spatial relationship within the tympanic cavity (24). Those ossicle traits discriminating Neandertals from AMHs are related to variation in angular relationships between the tympanic sulcus, the oval window, and the axis of the middle ear (Table 1 and Tables S3 and S4). One of the most marked differences between Neandertals and AMHs concerns the orientation of the head of the stapes relative to its footplate. This configuration is congruent with the off-center position of the oval window relative to the tympanic sulcus in Neandertals compared with AMHs shown in Fig. 3. Our results suggest that in these hominins structural requirements of the tympanic cavity rather than functional differences drive the evolutionary shape changes in ossicle morphology. A correlation to other aspects of the surrounding morphology of the cranial base, like the glenoid fossa, tympanic area, position and orientation of the external auditory meatus, or orientation of the petrous pyramid, therefore, seems likely (30, 39, 49). Changes in cranial base morphology are thought to be associated with changes in relative brain size (50, 51). Hence, distinct differences in middle ear architecture of AMHs and Neandertals may well reflect brain expansion that occurred separately in the two lineages (52, 53).

To interpret our findings, we estimated the ancestral shape of the last common ancestor of AMHs and Neandertals using the African apes as an outgroup and computed the Euclidean distances between the AMH and Neandertal mean shapes and their estimated last common ancestor. Although the stapes seems more derived in Neandertals, the polarity of malleus and incus shape changes relative to the inferred morphology of the last common ancestor of modern humans and Neandertals remains ambiguous. However, estimation of the last common ancestor of AMHs and Neandertals does not take into account the shape variation among extinct hominins after the split from Pan, because only sparse and fragmentary information exists for late Pliocene Australopithecus africanus and early Pleistocene Paranthropus robustus. Findings from these fossils are interpreted as “human-like” in malleus proportions but “great ape-like” in incus and stapes dimensions (54⇓–56). It is, therefore, more reliable to compare AMHs and Neandertal mean shapes with the inferred last common ancestor of Homo and Pan. Here, Neandertals are consistently more derived than AMHs. This finding is concordant with genetic data that show that genes involved in skeletal morphology have changed more than previously thought in the line leading to Neandertals (57) and further adds to the distinctiveness of this archaic human species.

In conclusion, there is no evidence for differences in the auditory sensitivity level in the lower frequencies between AMHs and Neandertals. The ossicles appear to show tight covariation with the surrounding tympanic cavity. Distinctly different morphologies, likely associated with convergent brain expansion, result in similar functional properties of the middle ear of these hominin species. These functionally equivalent solutions could indicate selective pressures acting on the middle ear for conserving a similar auditory sensitivity inherited from the last common ancestor of AMHs and Neandertals and may suggest consistent aspects of vocal communication in the two species. Our findings should also be the basis for future research on the evolution of complex human spoken language.

Materials and Methods

Sample and Imaging.

Tables S5 and S6 list specimens used in this study and provide information about morphological structures available for analysis. Table S7 summarizes the state of preservation of ossicles and surrounding middle ear structures in the fossil specimens. The complete fossil sample includes 16 Neandertals and 4 Pleistocene AMHs. The comparative extant sample comprises 81 mallei, 78 incudes, 37 stapedes, and 78 temporal bones, including the tympanic sulcus and oval window from AMH, P. troglodytes, Pan paniscus, Gorilla beringei, and G. gorilla. CT images of specimens housed in the American Museum of Natural History were provided by Rolf M. Quam, Binghamton University, Binghamton, NY. CT images of fossil specimens housed at the Muséum National d'Histoire Naturelle were scanned at AST-RX (Accès Scientifique à la Tomographie à Rayons X). All other specimens were scanned with the BIR ACTIS 225/300 or the Skyscan 1173 housed at the Max Planck Institute for Evolutionary Anthropology in Leipzig. Whenever possible, middle ear structures from the right side of the skull were analyzed. When necessary, left ones were mirrored. Avizo 7.1 (Visualization Science Group) was used to create 3D surface models of the ossicles and the temporal bone and place landmarks. In the case of isolated ossicles and temporal bones free of sediment, the Isosurface module was used using a single threshold value. Tympanic sulci and oval windows of sediment-filled temporals as well as ossicles scanned inside the temporal bone were isolated and visualized using the Segmentation Editor.

View this table:
Table S5.

Sample of Neandertal and Pleistocene AMH middle ear ossicles, tympanic sulcus, and oval window

View this table:
Table S6.

Precise sample description of the extant specimens analyzed in the study

View this table:
Table S7.

Summary about the state of preservation of the fossil specimens

Ossicle Landmarks and Shape Analysis.

Landmark coordinates were analyzed using Mathematica 8 (Wolfram Research, Inc.), with software routines developed by PG. The measurement protocol for ossicles is described in detail in ref. 18 and also summarized in SI Text, Text S2.

Measurement Protocol and Computation of Middle Ear Parameters.

Using surface models of the temporal bone, ∼25 landmarks were placed along the tympanic sulcus starting at the level of the anterior tympanic spine (slightly laterally from this point) and going in a clockwise direction when seen from medial. No landmarks were placed in the area of the tympanic notch. Along the inwardly projecting edge of the oval window, ∼15 landmarks were digitized starting from its most posterior oriented protrusion and going in a clockwise direction when seen from lateral. From these landmark sets, the following parameters were computed: maximal areas enclosed by the tympanic sulcus, the oval window, and the stapes footplate; centers of maximal areas enclosed by the tympanic sulcus and the oval window; the distance between these centers (e.g., middle ear length) and the axis running through these centers (e.g., middle ear axis); and angles between tympanic sulcus and oval window planes, with the middle ear axis. Detailed formulas for these computations are provided within SI Text, Text S3. It is noted that the maximal area enclosed by the tympanic sulcus is not equal to the functional surface area of the tympanic membrane for reasons outlined in SI Text, Text S3. However, as detailed in SI Text, Text S3, we show that the maximal area enclosed by the tympanic sulcus provides an estimate for the functional area of the tympanic membrane.

To increase the sample of temporal bones, infant and juvenile (erupted M1) specimens were included. Because ontogenetic erection of the tympanic ring ends at 4–5 y in AMHs (58), values for these specimens were compared with those of adults. No difference was found for the angle between the tympanic sulcus plane and the middle ear axis (AMHs, P = 0.522; Neandertals, P = 0.762) and between the oval window plane and the middle ear axis (AMHs, P = 0.681; Neandertals, P = 0.143).

To assess whether the area enclosed by the oval window can be used to assess stapes footplate area, the ratio between these two parameters was calculated in AMH, Neandertals, P. troglodytes, and G. gorilla. On average, the stapes footplate area was found to represent 90% (±5.2%) of the area enclosed by the oval window (SI Text, Text S4)—confirming what is found in the literature (54). This value was then used to estimate stapes footplate area from oval window area.

Functional Parameters.

The functional areas ratio RΛ was computed based on the work in ref. 31:RΛ=ΛTSΛST,

where ΛTS is the area enclosed by the tympanic sulcus, and ΛST is the area enclosed by the stapes footplate outline.

The functional lengths ratio RFL was computed based on the work in ref. 11:RFL=FLMFLI,

where FLM is the length from the malleus’ center of mass (CoM) to the tip of the manubrium, and FLI is the length from the incus’ CoM to the tip of the long processus.

CoM determination was done by transforming surface mesh models into volumetric models and averaging coordinates of individual volume elements (59). Computation was done in MATLAB (The MathWorks, Inc.) using the implementation of Aitkenhead (www.mathworks.com/matlabcentral/fileexchange/27390-mesh-voxelisation).

Comparing estimated middle ear impedance function of fossil Neandertals with extant species was done using a simple ideal transformer ratio (ITR) (31, 33):ITR=RΛ⋅RFL.

Statistics.

Permutation tests based on Procrustes distance between group means were used to test statistical significance of mean shape differences between groups (60). To this end, we compared the Procrustes distance between two group means, with average differences computed after randomly reshuffling group affiliations 10,000 times (61). To assess the influence of the morphology of the tympanic cavity on ossicle shape, we performed a multivariate regression of the Procrustes shape coordinates on each of the middle ear parameters that we have computed. Statistical significance of these correlations was tested using a permutation test based on the explained variance of the regression. Mann–Whitney–Wilcoxon tests were used in R, version 3.1.0 (The R Foundation for Statistical Computing) to assess statistical significance of interspecific differences in tympanic cavity parameters and functional ratios, ontogenetic differences in middle ear angles, and differences between stapes footplate and oval window areas (SI Text, Text S4 and Table S3). Dataset S1 lists the values of the morphological and functional parameters measured in our sample.

Ancestral State Estimations.

To estimate ancestral coordinates along the first three principal components describing malleus, incus, and stapes shape variation, for the successive nodes of the African hominid tree, we applied the Phylogenetic Comparative Method and particularly, the models in refs. 62⇓–64. For AMHs and Neandertals, the mean principal component scores weighted by fossil age were used. Comparison of the Euclidean distance in the shape space of the first three principal components makes it possible to compare how each group differs from the estimated ancestral state. A detailed description of the process is provided in SI Text, Text S5.

SI Text

Text S1. Phylogenetic Polarity of Ossicle Shapes.

Malleus.

Regarding shape changes carried over principal component 1 (PC 1) and PC 3, the mean shape of Neandertals seems more derived from the last common ancestor (LCA) of Pan and Homo than the mean shape of AMH. Regarding PC 2, mean shapes of both AMHs and Neandertals appear to be derived from the LCA of Pan and Homo, albeit differently. Mean shapes of both AMHs and Neandertals seem to be derived from their LCA along PCs 1–3, although again, differently. Overall (based on the Euclidean distance in the shape space of the first three PCs), the mean shape of Neandertals can arguably be considered to be more derived from the LCA of Homo and Pan than the mean shape of AMH, but both species seem to be differently but equivalently derived from their LCA.

Incus.

Regarding shape changes carried over PC 1, the mean shape of AMH seems more derived from the LCA of Pan and Homo than the mean shape of Neandertals. Regarding shape changes carried over PCs 2 and 3, the mean shape of Neandertals seems more derived from the LCA of Homo and Pan than the mean shape of AMH, particularly for PC 2. Mean shapes of both AMHs and Neandertals seem to be equivalently but oppositely derived from their LCA along PCs 1–3. Overall (based on the Euclidean distance in the shape space of the first three PCs), the mean shape of Neandertals can arguably be considered to be more derived from the LCA of Homo and Pan than the mean shape of AMH, but both species seem to be derived from their LCA, albeit differently.

Stapes.

Regarding shape changes carried over PC 1, the mean shape of Neandertals seems strikingly more derived from the LCA of Pan and Homo than the mean shape of AMH. Regarding shape changes carried over PC 2, the mean shape of AMH seems more derived from the LCA of Homo and Pan than the mean shape of Neandertals. On PC 3, the mean shape of Neandertals is confounded with the mean shape of the LCA of Homo and Pan. Mean shapes of both AMHs and Neandertals seem to be equivalently but oppositely derived from their LCA along PCs 1–3. Overall (based on the Euclidean distance in the shape space of the first three PCs), the mean shape of Neandertals can arguably be considered to be more derived from the LCA of Homo and Pan than the mean shape of AMH, but both species seem to be derived from their LCA, albeit differently.

Text S2. Ossicle Landmarks and Ossicle Shape Analysis.

Landmark coordinates were exported from Avizo 7.1 and analyzed using Mathematica 8 (Wolfram Research, Inc.), with software routines developed by PG and Philipp Mitteroecker. The measurement protocol for the ossicles is described in detail in ref. 18. The protocol was developed so that the landmarks and a large number of semilandmarks represent the overall shape and size of the ossicles and quantify functionally and structurally important features. After semilandmark sliding, the shape of the ossicles was assessed following the methods discussed in ref. 18, including Procrustes superimposition, and PC analysis of the landmark data. Visualization of the shape changes related to the individual PC axes was done by surface warping a reference surface to a target configuration using thin-plate spline interpolation (18). The same principle was used to calculate the mean shapes of the Neandertal, AMH, Pan troglodytes, and Gorilla gorilla ossicle samples. This calculation was done after averaging the Procrustes shape coordinates of this sample. Visualization of the shape changes related to changes in the surrounding morphology was done by landmark warping a reference landmark configuration to a target configuration using thin-plate spline interpolation. Unspecified reference to size in descriptions and comparisons given below (e.g., the manubrium is long) concerns relative size, because absolute size has been removed.

Text S3. Computation of the Middle Ear Parameters.

Maximal areas enclosed by the tympanic sulcus, the oval window, and the stapes footplate were computed by applying the following formulas:Λ→k=−12∑i=1Nk‖r→k,i+1−r→k,i‖⋅((r→k,i+r→k,i+1)2×(r→k,i+1−r→k,i‖r→k,i+1−r→k,i‖))andΛk=‖Λ→k‖,where Λk corresponds to the maximal area enclosed by the landmarks of the tympanic sulcus, the oval window, or the stapes footplate sets as projected onto their respective plane; Nk corresponds to the number of landmarks contained in the tympanic sulcus, the oval window, or the stapes footplate sets, the first landmark being counted twice to close the loop; and r→k,i corresponds to a vector running from the origin of the reference frame to the ith landmark of the tympanic sulcus, the oval window, or the stapes footplate sets.

Centers of maximal areas enclosed by the tympanic sulcus and the oval window were computed by applying the following formulas:Lk=∑i=1Nk‖r→k,i+1−r→k,i‖andC→k=1Lk∑i=1Nk‖r→k,i+1−r→k,i‖⋅(r→k,i+r→k,i+1)2,where Lk corresponds to the length of the outline of either the tympanic sulcus or the oval window, and C→k corresponds to a vector running from the origin of the reference frame to the center of the maximal area enclosed by the landmarks of either the tympanic sulcus or the oval window.

It should be noted that the maximal area enclosed by the tympanic sulcus is not equal to the functional surface area of the tympanic membrane because of two reasons. First, the tympanic membrane is not planar but rather, conically shaped with an apex that points medially, and thus, the 3D surface is inevitable larger than any planar 2D surface area (65). Second, the tympanic sulcus houses the fibrotic annulus, on which the pars tensa (functional area) inserts. Hence, the enclosed area of the tympanic sulcus is consequently larger than a planar 2D measurement of the functional area of the tympanic membrane. However, using CT scans of five human cadaveric specimens (from the Comparative Ear Bank collection housed at the Max Planck Institute for Evolutionary Anthropology) that have been stained with phosphotungstic acid prior to scanning in order to enhance soft-tissue contrast (e.g., ref. 65), we made surface reconstructions of the tympanic area. We then computed the 3D surface area of the tympanic membrane using in-house software and also, measured the area enclosed by the tympanic sulcus with the method detailed above. In these five specimens, the average 3D surface area of the tympanic membrane has a value of 68.2 mm2 (±6.55 mm2), whereas the area enclosed by the tympanic sulcus is 68.4 mm2 (±8.23 mm2). The average difference between the two values is 2.1 mm2 (±1.44 mm2). These data suggest that measuring the area enclosed by the tympanic sulcus provides a reasonable estimate for the actual functional surface area of the tympanic membrane and is appropriate for the aim of this study, because only this measure can be analyzed in fossils.

The distance between these centers, the middle ear length, was then computed by applying the following formula:MEC=‖C→TS−C→OW‖.Finally, the angles between the tympanic sulcus and oval window planes with the middle ear axis were computed following the formulasu→TS=Λ→TS‖Λ→TS‖, u→OW=Λ→OW‖Λ→OW‖, u→MEC=C→TS−C→OWMEC,u→TS∧u→MEC=90−(arc⁡cos(u→MEC⋅−u→TS)×57.3),andu→OW∧u→MEC=90−(arc⁡cos(u→MEC⋅u→OW)×57.3),where u→TS and u→OW correspond to the unit vectors that are normal to the planes of the tympanic sulcus and the oval window, respectively; u→MEC corresponds to the unit vector that is collinear to the middle ear axis and directed from the tympanic sulcus to the oval window; and u→TS∧u→MEC and u→OW∧u→MEC correspond to the angles between the tympanic sulcus and oval window planes with the middle ear axis, respectively.

Text S4. Proportion Stapes Footplate Area—Oval Window Area.

We tested if individual species significantly differed from the average of the entire sample (stapes footplate represents 90% of the area enclosed by the oval window) but found no support for this (AMH n = 16, P = 0.275; Neandertals n = 5; P = 0.737; P. troglodytes n = 5; P = 0.873; and G. gorilla n = 2; P = 0.133).

Text S5. Ancestral State Computation.

In each case of ancestral state estimation, we studied the modelY=Xβ+ε,where Y is the dependent variable that corresponds to a matrix with n rows and one column (the PC coordinates in this case), X is a matrix with n rows and one column only filled with one, β corresponds to the phylogenetic mean (that is, the expected ancestral state of Y at the root of the tree considered), and ε is the error term, a matrix with n rows and one column that is assumed to be normally distributed with a mean equal to zero and a variance–covariance matrix equal to σ2Σ (62, 63). The parameter n corresponds to the number of taxa used in the model.

The main difference between phylogenetically corrected ancestral state estimations and classical averaging is that the matrix of variance–covariance of the error term ε is assumed to contain information pertaining to shared phylogenetic history, which is taken into account for the estimation. This shared phylogenetic history is embodied by the matrix Σ, which is composed of n rows and n columns, with diagonal values corresponding to the branch length between each taxa and the root of the tree, whereas off-diagonal values correspond to the branch length shared by each pair of taxa from the root of the studied tree (that is, the branch length between the root and the LCA between each pair of taxa). The matrix Σ expresses the a priori structure of dependency and independency between taxa and was taken here to be strictly equal to the estimated time of divergence between the extant taxa.

Using this methodology, β^ and ε^, the estimators of β and ε, are computed by the usual method of generalized least squares, whereas s2, the estimator of σ2, is given by the following equation (64):s2=[1(n−1)](Y−Y^)tΣ−1(Y−Y^),where Y^=Xβ^. ε^, the estimator of ε, can be computed asε^=(Y−Y^).

When phylogenetic relationships are known, we can then use our knowledge of ε^ in studied taxa to predict ancestral states based on phylogenetic relationships (62, 63). If we define a node n, we then haveYn=Xnβ^+ε^n,where Yn corresponds to the maximum likely value of the dependent variable at the level of the node, Xn equals one in our case, β^ corresponds to the already computed estimated phylogenetic mean, and ε^n corresponds to the unknown phylogenetic correction. This last parameter is normally distributed with a mean μn and a variance s2En. To predict the most likely value of a dependent variable at the level of a node along with its uncertainty, we then have to estimate μn and s2En based on the phylogenetic relationships between the taxa and the error term ε^.

Following the works in refs. 62 and 63, the mean μn of the phylogenetic correction term ε^n can be estimated asμ^n=μroot+wn⋅[μtips−μroot1],where [μtips−μroot1] is a matrix of n rows and one column, [μtips] corresponds to the values of ε^, and one corresponds to a matrix of one row and n columns. Also,μroot=(1tΣ−11)−11tΣ−1[μtips](see ref. 62),and wn is a matrix of one row and n columns and corresponds to the phylogenetically weight matrix that can be calculated aswn={(1tΣ−11)−11t+Σn[I−Σ−111t(1tΣ−11)−1]}Σ−1(see ref. 63),where I corresponds to the identity matrix of n rows and n columns, whereas Σn corresponds to a matrix of one row and n columns that represents the shared phylogenetic history between the ancestral node and each of the taxa that were used to establish the phylogenetically corrected regression.

Following the works in refs. 62 and 63, the variance s2En of the phylogenetic correction term ε^n can be estimated ass2En=s2droot>n−s2wnΣnt,where droot>n corresponds to the branch length between the root of the studied tree and the studied node.

Interspecific relationship and divergence dates of the extant species analyzed here are similar to those in ref. 18. The divergence data of AMHs and Neandertals are obtained from ref. 66. References for the fossil ages are ref. 67 for Abri Pataud, ref. 68 for Amud, ref. 69 for Cro-Magnon, ref. 70 for Ehringsdorf, ref. 71 for Krapina, ref. 72 for La Chapelle-aux-Saints, ref. 73 for La Quina and Le Moustier, ref. 74 for Pech de l’Azé, and ref. 75 for Qafzeh.

Acknowledgments

We thank A. Balzeau, I. Bechmann, C. Boesch, C. Feja, M. S. Fischer, S. Flohr, C. Funk, I. Hershkovitz, A. Hoffmann, F. Mayer, R. M. Quam, J. Radovčić, T. Schüler, J. F. Tournepiche, B. Vandermeersch, and Y. Rak for access to specimens and H. Temming and D. Plotzki for scanning. This research was funded by the Max Planck Society.

Footnotes

  • ↵1To whom correspondence should be addressed. Email: alexander_stoessel{at}eva.mpg.de.

  • Author contributions: A.S. and J.-J.H. designed research; A.S., R.D., P.G., F.S., and J.-J.H. performed research; A.S., R.D., P.G., and T.S. analyzed data; and A.S., R.D., P.G., F.S., and J.-J.H. wrote the paper.

  • The authors declare no conflict of interest.

  • This article is a PNAS Direct Submission.

  • This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1605881113/-/DCSupplemental.

Freely available online through the PNAS open access option.

Back to top
Article Alerts
Email Article
Citation Tools
Request Permissions
Neandertal ear ossicles
Alexander Stoessel, Romain David, Philipp Gunz, Tobias Schmidt, Fred Spoor, Jean-Jacques Hublin
Proceedings of the National Academy of Sciences Sep 2016, 201605881; DOI: 10.1073/pnas.1605881113
del.icio.us logo Digg logo Reddit logo Twitter logo CiteULike logo Facebook logo Google logo Mendeley logo
Proceedings of the National Academy of Sciences: 118 (15)
Current Issue

Submit

Jump to section

You May Also be Interested in

Similar Articles