Comparative morphology of the dormouse skull and the influence of size and ecology

Abstract

Dormice are widely dispersed across various ecosystems in Eurasia and Africa and among the oldest extant rodent lineages. Despite distinct morphological variation to be present between groups, comprehensive morphometrical studies on the dormouse skull are limited. Here, the form of eight out of the nine extant dormouse genera was analysed using 3D geometric morphometrics and linear biomechanical measurements, providing a better understanding of the overall morphological variation present within Gliridae. Species‐, genus‐ and family‐specific morphological trends in both the size and shape of the cranium and mandible are linked with specific habitats and feeding strategies. Smaller dormice show adaptations to a more arboreal lifestyle such as a relatively enlarged braincase and an inferiorly reoriented foramen magnum. Larger dormice show cranial modifications, including clear flattening of the skull and a more posteriorly positioned foramen magnum, hinting towards a more rupicolous lifestyle. Furthermore, specimens inhabiting arid areas appear to have more inflated auditory bullae, whereas other variable features, such as the length of the incisive foramen, were not associated with either size changes or climatic variables. Lastly, more robust and horizontally orientated zygomatic arches as well as increased robusticity of the molar row appear to be linked with herbivory in dormice, whereas thinner arches and small concave molars are seen in more insectivorous species. This study reveals clear convergent adaptations between dormouse species and results in a better understanding of ecological drivers underpinning the morphological divergence present within Gliridae.

This study focuses on the morphological variation in the skull and mandible of dormice (Gliridae). Using geometric morphometrics and biomechanical analyses, the shape of eight out of the nine extant dormouse genera is evaluated. Assessing the influence of size and ecology on morphology, this comparative study reveals convergent adaptations between genera, and highlights the strong relationship between form and function within the mammalian skull.

1. INTRODUCTION

Gliridae is considered one of the oldest extant rodent families (Fabre et al., 2012) and consists of nine dormouse genera split over three subfamilies: Glirinae (Glirulis, Glis), Leithiinae (Chaetocauda, Dryomys, Eliomys, Muscardinus, Myomimus and Selevinia), and the monogeneric Graphiurinae (Graphiurus) (Holden‐Musser et al., 2016). Dispersed widely throughout Africa and Eurasia, the family occupies various ecological habitats, including tropical forests in Africa and deserts in central Asia. Dormouse species vary significantly in both cranial and mandibular morphologies, with some of these morphological variations thought to be related to adaptations to specific habitats and diet (Hennekam, Sadler, et al., 2020; Wahlert et al., 1993). The fossil record shows the radiation of Gliridae to start during the Early Eocene (Vianey‐Liaud, 1994), and to peak around the late Early Miocene (Daams & de Bruijn, 1995). During this time, dormice were especially abundant and diverse within the Mediterranean area (Hartenberger, 1994), occupying a variety of different ecological niches and exhibiting a great diversity in diet, as evidenced by, for example, the evolution of hypsodont species (Daams, 1990). The extant dormouse distribution is only a remnant of the success of this family in the past; nonetheless, the various ecological niches occupied by these species today suggest that adaptability to different environments is still retained within the family.

Gliridae is a relatively small family within Rodentia with only 29 species in 9 genera (Holden‐Musser et al., 2016), making it a tractable group for analyzing morphological variations with respect to size, phylogeny and ecology at this taxonomic level. Previous descriptive studies have analysed the cranium within extant dormice (Daams & de Bruijn, 1995; Hautier et al., 2008; Hennekam, Sadler, et al., 2020; Koenigswald, 1995; Potapova, 2001; Wahlert et al., 1993) and found a large variation in, for example, the zygomasseteric construction, morphology of the auditory bullae, the infraorbital foramen (IOF) and dental characteristics. In particular, the subfamily Graphiurinae appears to differ significantly from other dormice, with its zygomatic arch displaying a hystricomorph condition (Hautier et al., 2008; Tullberg, 1899; Wahlert et al., 1993), in contrast to the myomorph condition present within Leithiinae and Glirinae (Simpson, 1945). This is the only rodent family displaying two types of masseter muscle organisation (Hautier et al., 2008), and this distinct morphological differentiation resulted in much debate on the phylogenetic position of Graphiurinae within the family (Simpson, 1945; Vianey‐Liaud & Jaeger, 1996; Winge, 1941). Molecular phylogenetic studies have not been conclusive for Gliridae, resulting in the positioning of some genera and species remaining unresolved (Bentz & Montgelard, 1999; Montgelard et al., 2003; Nunome et al., 2007).

1.1. Aims

The considerable morphological variation in combination with the relatively low number of species within Gliridae (29) makes it an excellent candidate for a case study investigating adaptive morphological features within a group at species, genus and (sub)family level. Using geometric morphometrics, this study aims to characterise the morphological variation present in the crania and mandibles of dormice and to assess the relationship between morphology and ecology. All but one of the genera comprising Gliridae are represented in the study, resulting in a nearly complete morphological overview of this family. By evaluating the morphological variation associated with various ecoregions within genera, it is possible to identify to what extent shape variation within Gliridae is the result of phylogenetic divergence, and which morphologies might indicate adaptive features to specific habitats and lifestyles. In addition, the functionality of various morphologies is evaluated using biomechanical analyses. Furthermore, analysing the effects of size on shape at a family level will enable the identification of cranial and mandibular adaptations for coping with increasing body size.

2. MATERIALS AND METHODS

2.1. Sample and data acquisition

A total of 109 specimens were included in this research, encompassing 8 of the 9 extant dormouse genera (Figure 1). The missing genus, Chaetocauda, is incredibly rare, being known from only five specimens (Holden‐Musser et al., 2016). Multiple populations were included for widely dispersed species. Insular populations were avoided, as morphological variation in dormice linked with insularity is described in a previous study (Hennekam, Benson, et al., 2020). Even though sexual dimorphism is not considered to be present within dormice (Holden‐Musser et al., 2016), whenever possible, an equal representation of both sexes was included within the dataset. The material used is located in the collections of Muséum National d’Histoire Naturelle, Paris (MNHN); the Natural History Museum, London (NHMUK); the Senckenberg‐Forschungsinstitut und Naturmuseum, Frankfurt (SMF); and the Zoological Institute in St Petersburg (ZISP) (Table S1). All material was digitized using micro‐computed tomography (μCT), using facilities at the X‐ray Tomography Facility, University of Bristol (Nikon T H 225 ST CT scanner) and University of Liverpool (SKYSCAN 1272), as well as one scan at the Centre for X‐ray Diffraction Studies of Saint Petersburg State University (SKYSCAN 1172). Isometric voxel dimensions ranged between 4 and 40 μm, depending on the size of the specimen and the scanning facility used. Only adult specimens (fully erupted third molar) were included to exclude ontogenetic variation from the dataset.

FIGURE 1.

Lateral view of crania and mandibles representing all the species used in this study. Colours are associated with datapoints in subsequent figures referring to the genus. The same scale is used for all specimen

2.2. Data collection

3D anatomical landmarks were placed on surface reconstructions derived from the μCT scans, in order to represent cranial and mandibular shape. Landmarks were recorded in the imaging software Avizo Lite v9.2.0 (Thermo Fisher Scientific) using configurations only including type 1 and 2 landmarks (Bookstein, 1991). The cranial landmark set consisted of 42 anatomical landmarks, and the mandibular configuration included 19 landmarks (Figures S1 and S2). Only the left‐hand side of the specimens was landmarked. Whenever this area was incomplete or clearly deformed, the object was mirrored digitally to enable landmarking of the right‐hand side instead, assuming bilateral symmetry within all dormice.

2.3. Analyses

This study analyses the shape and size of various dormouse species at a family, genus and species level. The initial analyses included all specimens within the dataset. Subsequently, statistical analyses were performed at a genus level and eventually qualitative analyses at a species (subspecies) level.

All statistical analyses were undertaken in R Studio v3.5.3 (RStudio, Inc.) using various R packages for data preparation and visualisation (Morpho: Schlager, 2017; geomorph: Adams et al., 2018; Arothron: Profico et al., 2018). All configurations were superimposed using a Generalized Procrustes Analysis (GPA) to provide optimal comparability of shape between specimens (Gower, 1975; Rohlf & Slice, 1990). A GPA involves translating all specimens to the origin, optimally rotating using a least‐square criterion, and isotropic scaling of all shapes to unit‐centroid size in order to align the landmarks as accurately as possible. Centroid size was used as the estimation of size in all subsequent analyses.

2.3.1. Family level

A principal component analysis (PCA) was used to visualise the shape variance within the entire mandibular and cranial dataset. PCAs including all axes representing over 5% of the total shape variation were plotted. The specimen best approximating the mean shape was warped along the axes to represent the morphological shape variation associated with these components.

Allometry within Gliridae was evaluated using Procrustes ANOVAs for the mandibular and cranial dataset. In order to investigate whether unique (genus‐specific) allometries or a common (family‐specific) allometric trajectory best explains the correlation between size and shape, ANOVAs including common and unique allometries were performed to identify the correct model for analysing allometry within the datasets.

$$\text{Unique}\text{allometry}:\text{shape}\sim \text{size}\times \text{genus,}$$

$$\text{Common}\text{allometry}:\text{shape}\sim \text{size}+\text{genus}.$$

2.3.2. Genus level

More detailed analyses were performed per genus, applying ANOVAs showing correlations with shape for both size and location. These analyses evaluate to what extent intrageneric variation occurs between geographically separated groups, for both size and shape. Unfortunately, for numerous specimens only the country of origin was documented. For consistency, the variable “location” refers to the country of origin of the specimens, unless clear differences in populations were made by the assignment of subspecies (e.g. Eliomys quercinus lusitanicus, a subspecies in the South of Spain). Due to the limited available information on the exact location of specimens, statistical analyses, on for example habitat type, were not feasible. Dormice exhibit multiple zygomasseteric conditions, resulting in the shape of the IOF to vary between species (Hautier et al., 2008). This feature is represented by three landmarks in the cranial landmark configuration, and its morphology assessed independently in addition to the morphological analyses of the cranium and mandible.

2.3.3. Species level

In order to evaluate the effects of ecological variance on skull morphology within dormice, one genus per subfamily was investigated in more detail, assessing cranial and mandibular shape per geographical region in more detail. Criteria for the chosen genera included a wide dispersal area with specimens collected from various locations, a significant correlation between location and shape, and relative abundance in the dataset. The subfamily Glirinae includes two species: Glirulus japonicus and Glis glis. Glis glis is used in this study as it is widely distributed across Europe and small parts of Asia, whereas Glirulus is geographically restricted to Japan. Within Leithiinae, Eliomys is considered the most widely dispersed genus and is considerably more abundant in the dataset than other genera belonging to this subfamily. Graphiurinae is a monogeneric subfamily, solely including the genus Graphiurus, which meets all the determined criteria.

A relatively large number of specimens of these three genera were included in the datasets (Glis [12], Eliomys [39], Graphiurus [30]) and all three genera are widely distributed; together, they cover the majority of the total dispersal area of all dormice, excluding the eastern parts of Asia. The method of analysing shape is dependent on the number of specimens within the genus, the species included and how they morphologically differ, and the different ecoregions occupied. Variations in morphology were quantified by comparing species to the mean Procrustes shape per genus. Only a limited number of specimens per location were analysed, as to avoid overrepresentation affecting the mean shape.

2.4. Biomechanical analysis

The mechanical advantage (MA) of three masticatory muscles is estimated for all dormouse genera (Casanovas‐Vilar & van Dam, 2013; Cox et al., 2020). The MA is calculated based on the ratio of in‐ and out‐levers (Figure 2). In‐levers represent the linear distances between the center of the condyle and the extremity of the muscle insertions on the mandible. The location of the center of the condyle is calculated by determining the midpoint between mandibular landmarks 3 and 4 (see Figure S2 for placement of landmarks). The insertions of the extremities are represented in the landmarks used for the geometric morphometric analyses: dorsal part of the temporalis (LM1), posterior part of the superficial masseter (LM6), ventral part of the superficial masseter (LM8), anterior part of the deep masseter (LM18). The out‐levers used in the biomechanical analyses are based on the biting at the tip of the incisor (LM10) and at the second molar (LM15). The MA is calculated by dividing the linear distance of the respective out‐lever with the linear distance of the in‐lever at the incisor for gnawing, and the second molar for representing chewing. The difference in the MA between genera was assessed using post hoc pairwise permutations tests (10,000 permutations), adjusting the p‐values using the Holm‐Bonferroni method to account for multiple comparisons (Holm, 1979).

FIGURE 2.

Lateral view of a right hemi‐mandible indicating the landmarks used for estimating the mechanical efficiency within dormice. Abbreviations refer to the in‐ and out‐levers used in the biomechanical analyses: DM_a, anterior deep masseter in‐lever; I, incisor out‐lever; M₂, second molar out‐lever; SM_d, dorsal superficial masseter in‐lever; SM_v, ventral superficial masseter in‐lever; T, temporalis in‐lever

2.5. Validation and caveats

For the Eurasian specimens, the assignment of location is often not linked with a change in ecoregion. Instead, physical barriers (e.g. rivers, lakes and mountain ranges) are more likely to play a role in the segregation of populations. The Sub‐Saharan ecoregions are more variable and seem to act as physical barriers as well. However, the species‐rich genus Graphiurus is the only dormouse genus inhabiting this part of the world. The species ranges within this genus, according to the IUCN, are clearly associated with climatic variables, such as temperature and precipitation (WorldClim data: Fick & Hijmans, 2017), and the associated terrestrial ecoregions (as defined by the World Wildlife Fund). Unfortunately, the Graphiurus dataset is limited and does not include all species. Furthermore, for some species only one specimen is included, resulting in a poor representation of the species. Lastly, the lack of precise geographical information on some specimens resulted in an inaccurate estimation of climatic variables and in some cases in the assignment of two different ecoregions. Considering the large number of different species and the wide distribution of certain species within the genus Graphiurus, covering a large number of ecoregions, it is likely that this study has captured only a fraction of the morphological variation present. It must be noted that due to the inaccurate determination of the location (country level), ecoregions were not used in statistical analyses, but instead assessed qualitatively.

With respect to the biomechanical analysis, a number of caveats should be noted. Firstly, the use of lever arms is a rather simplistic way of estimating the MA, whereas the use of moment arms would be more appropriate (Cox et al., 2020). Secondly, as this study uses 3D landmarks, the linear distance between the condyle and muscle insertion also include lateral and medial displacement. As the mediolateral orientation of the muscle insertion does not contribute to the biteforce, the effective in‐lever might be overrepresented. Thirdly, the out‐lever for chewing is estimated based on the placement of the second molar, whereas between dormouse species there are clear differences in the length of the molar row, as well as substantial differences in size between specific molars. The effective out‐lever lengths could therefore differ significantly from those used in the model. Lastly, the insertion areas of the muscles are represented by one or two points, whereas the masseteric muscles have rather large muscle insertion areas on the mandible in dormice. Due to the simplistic nature of the biomechanical analysis in this study, the results should solely be interpreted as a first reasonable approximation of the MA.

3. RESULTS

3.1. Shape variation (family level)

The distribution of dormice in shape space is shown in the PCA plots in Figures 3 and 4 and described below. The skulls and mandibles of specimens representing all the species analysed in this study are displayed in Figure 1. The species are colour‐coded in correspondence with the data points in subsequent figures.

FIGURE 3.

Principal component analysis for the cranial dataset and corresponding warps on the first four principal components. The skull of an Eliomys quercinus specimen from France (Muséum National d’Histoire Naturelle, Paris: 1961‐534) is most similar to the mean cranial shape in the dataset and used for the warping along the axes. The size of the dots represents relative centroid size of the specimens. Gc, Graphiurus crassicaudatus; Gn, Graphiurus nagtglasii; Sb, Selevinia betpakdalaensis. Colour key (also see Figure 1): cyan = Graphiurus; grey = Selevinia; black = Dryomys; red = Eliomys; yellow = Myomimus; purple = Muscardinus; green = Glirulus; blue = Glis. See Figure S3 for annotations at a species level

FIGURE 4.

Principal component analysis for the mandibular dataset and corresponding warps on the first four principal components. The mandible of a Dryomys nitedula specimen from Turkey (Natural History Museum, London: 61.362) is most similar to the mean mandibular shape in the dataset and used for the warping along the axes. The size of the dots represents relative centroid size of the specimens. Gc, Graphiurus crassicaudatus; Gn, Graphiurus nagtglasii; Sb, Selevinia betpakdalaensis. Colour key: cyan = Graphiurus; grey = Selevinia; black = Dryomys; red = Eliomys; yellow = Myomimus; purple = Muscardinus; green = Glirulus; blue = Glis. See Figure S3 for annotations at a species level

3.1.1. Crania

The first 32 principal component axes explain 95% of the total shape variance within the cranial dataset. The first four axes explain more than 5% of the total variance and are plotted in Figure 3. Clear morphological variation between genera is apparent in these four components, with every genus being separated from another in at least one of the axes. The first principal component (PC1) does not show a clear size signal (allometry), but instead distinguishes between the three subfamilies; Graphiurinae (Graphiurus) for negative values, Leithiinae (Dryomys, Eliomys, Muscardinus, Myomimus and Selevinia) in the centre and Glirinae (Glis, Glirulus) at more positive values. Some specimens deviate from this pattern, as two species within Graphiurus (G. nagtglasii and G. crassicaudatus) seem to be more similar to Leithiinae specimens, whereas Selevinia is placed within the Graphiurinae on the first principal component. Muscardinus shows relatively positive values on the first axis. For the first three principal components, the cranial morphology of Muscardinus appears to resemble that of the Japanese dormouse G. japonicus, placed within the Glirinae. The second principal component differentiates between genera within sub‐families, as clear separation visible between G. glis (negative) and G. japonicus (positive) for Glirinae and between Eliomys, Dryomys and Muscardinus (negative to positive) in Leithiinae. On this axis, larger species are placed at more negative values (Eliomys, Glis and certain Graphiurus species), and the smaller species at more positive levels (Muscardinus and Glirulus). The third principal component also shows a clear segregation between species; however, the size element is lost as the two largest genera, Glis and Eliomys, are at opposite ends of the spectrum. Myomimus within Leithiinae is now separated from Eliomys and Dryomys, and instead resembles the smaller dormice Glirulus and Muscardinus, as well as some Graphiurus specimens. The fourth principal component indicates morphological similarities between Myomimus and the desert dormouse Selevinia. Furthermore, the Japanese Glirulus resembles certain Graphiurus species (G. crassicaudatus, G. nagtglasii and some G. murinus) at the more negative values on this axis.

3.1.2. Cranial shape variation

The PC1 is characterised by a more open IOF at negative values, and a more slit‐like foramen at the positive end of the component. Negative values are also indicative for a shortened molar row, thinner zygomatic arch and enlarged auditory bullae. Additionally, the foramen magnum is substantially bigger in the dormouse genera at low PC1 values (Selevinia and Graphiurus). The second principal component shows a flattened and relatively elongated skull at more negative values, as well as a horizontally orientated zygomatic arch. Skulls at the positive end of this component have a more tilted zygomatic arch. Furthermore, they are characterised by a bulbous cranial vault, resulting in the overall morphology to appear more compacted. The foramen magnum is positioned ventrally in these dormice, whereas more posteriorly (caudally) orientated in dormice at negative values. PC3 primarily shows a change in rostrum length and the tilting of the zygomatic arch, with a more elongated rostrum and tilted arches at positive values of the component axis. The fourth principal component axis is less distinctive, but a slight tilting of the zygomatic arch is seen at more positive values, as well as more inflated auditory bullae.

3.1.3. Shape variation in the IOF

The shape of the IOF in dormice varies from more laterally extended foramina in Graphiurinae to a thin slit‐like foramen in Leithiinae and Glirinae (Figure 5). The form of the foramen appears to vary within genera, including significant overlap between genera. Within Leithiinae, the IOF in Muscardinus is less slit‐like and instead more similar to the genus Graphiurus. The subfamily Graphiurinae has the most variable IOF of all dormice, being more laterally extended. The most lateral part of the foramen is more superior in larger Graphiurus species, such as G. crassicaudatus and G. nagtglasii.

FIGURE 5.

Principal component analysis for a subset of the cranial dataset, displaying shape variation in the infraorbital foramen on the two principal components. Gc, Graphiurus crassicaudatus; Gn, Graphiurus nagtglasii; Sb, Selevinia betpakdalaensis. Colour key: cyan = Graphiurus; grey = Selevinia; black = Dryomys; red = Eliomys; yellow = Myomimus; purple = Muscardinus; green = Glirulus; blue = Glis

3.1.4. Mandibles

For the mandibular principal component analyses, 21 axes are needed to explain 95% of the total shape variation, with the first four components explaining more than 5% of shape variation each (Figure 4). Clear separation between genera is visible within these axes, whereas size does not appear to be a prominent driver of shape variation. The first principal component separates Muscardinus from all other genera. The second principal component distinguishes between Graphiurinae and Leithiinae. In contrast to the cranial dataset, the mandibular shape of G. nagtglasii and G. crassicaudatus places them neatly within their subfamily. The desert dormouse Selevinia is still located within the Graphiurinae, whereas it belongs to Leithiinae. Glis is associated with more positive values on principal component 3. Furthermore, G. crassicaudatus and G. nagtglasii show more positive values than other Graphiurus species, a trend also seen in the cranial analysis for these two species (Figure 3: PC1 and PC4). The fourth principal component for the mandibular shape variation is less indicative for morphological variation between genera in comparison with the fourth component in the cranial dataset. Clustering of species within genera such as Glirulus and Myomimus is visible, although both genera are relatively underrepresented in the dataset.

3.1.5. Mandibular shape variation

The first principal component represents clear morphological variation in the size of the mandibular body and the masseteric region. Negative values indicate a relatively increased surface area of the mandibular body and the masseteric region. The incisor is orientated more upward and the molar row is tilted downwards posteriorly. Furthermore, the anterior margin of the masseteric ridge is positioned more anteriorly, the coronoid process is in a more upright position, the condyle wider and the condylar process deepened. Dormice on the positive end of this component are more elongated and less widened dorsoventrally. Additionally, the incisors are positioned more anteriorly and less tilted upward, whereas the molar row is not tilted and relatively shortened. The second principal component axis indicates clear morphological variation in the size of the incisor and the posterior part of the mandible. More negative values on this axis are indicative for a longer incisor. The coronoid and condyle are relatively shortened, whereas the angular process is clearly enlarged, also flaring more laterally. The positive end of PC2 represents a smaller angular process, whereas the condylar process is elongated. The anterior margin of the masseteric ridge is located more posteriorly, and the length of the molar row is relatively shortened. Furthermore, the width of the mandibular body is slightly enlarged. The positioning of the angular process relative to the condyle is more anteriorly, whereas at a similar position for mandibles at more negative values on PC2.

A similar trend is seen in the third principal component. More negative values are associated with a relatively anterior position of the angular process, with respect to the condyle, whereas more positive values indicate the posterior part of the angular process to be at the same anterolateral position as the condyle. More positive values also represent a horizontally orientated molar row, a shortened diastema, and a more upright position of the incisor. Also the coronoid process is positioned more upright, whereas at negative values this element flares laterally. PC4 shows shortened and broader incisor at more negative values. Furthermore, the coronoid process is elongated and the angular process enlarged, flaring laterally. Positive values represent long and slim incisors, as well as an elongation of the condylar process.

3.2. Size variation in the skull and mandible (family level)

Based on the landmark configuration, the centroid size of each dormouse genus was determined for both skull and mandible. Clear size variation is present between genera, with the largest dormouse being G. glis (centroid size = 93), and the smallest dormouse being Muscardinus avellanarius (46). These size differences are even more pronounced within the mandibular dataset (min = 18, Graphiurus murinus; max = 43, G. glis). Graphiurus, being by far the most species‐rich dormouse genus, showed the largest intrageneric variation in centroid size, with the largest skull being 56% larger than the smallest specimen, and the largest mandible being 69% larger. This size variation within Graphiurus is the result of the exceptionally large species G. nagtglasii (outlier in Figure 6).

FIGURE 6.

Size variation between dormouse genera in both the cranial (left) and mandibular (right) datasets

3.3. Allometry (family level)

The model comparisons between the unique allometries (shape ~ size × genus) and the common allometries (shape ~ size + genus) were not significant (cranium p = 0.10; mandible p = 0.07). These results indicate the common allometric component (without grouping) to be the most accurate model for analysing allometry within dormice, rather than using unique allometric trajectories per genus (Figure 7, and see Table S2 for test). In total, 7.5% of the shape variance in the mandibular dataset was explained by size, with 16.7% explaining the variance in cranial shape. These relatively low values clarify why the effects of size on shape are not well represented in the first principal components (Figures 3 and 4). The linear relationship between size and shape was tested using a Pearson product‐moment correlation coefficient, showing strong correlations in both datasets (cranium: r = 0.934, 95% CI = 0.91–0.95; mandible: r = 0.829, 95% CI = 0.76–0.88). The amount of shape variance is greater and the linear relationship between shape and size is considerably stronger in the cranial dataset, indicating that allometry influences the morphology of cranial features to a greater degree than in the mandible.

FIGURE 7.

Common allometric component analyses for the cranial (a) and mandibular (c) datasets, including corresponding warps (b) representing the morphology of smaller dormice (left) and larger specimens (right). Colour key (also see Figure 1): cyan = Graphiurus; grey = Selevinia; black = Dryomys; red = Eliomys; yellow = Myomimus; purple = Muscardinus; green = Glirulus; blue = Glis

3.3.1. Cranial allometric variation

Smaller individuals are characterised by a short and robust rostrum and a more bulbous cranial vault (Figure 7). Larger specimens display a more elongated snout and a clear flattening of the skull. The IOF is more triangular in shape in smaller dormice, whereas it is represented by an elongated opening in large dormice. Furthermore, the auditory bulla appears to be relatively small in larger specimens. Lastly, the foramen magnum is large and ventrally orientated in small dormice, but relatively smaller and more posteriorly orientated as size increases.

3.3.2. Mandibular allometric variation

With increasing size, the angular process in the mandible moves slightly posteriorly as the condylar process moves more anteriorly, resulting in the posterior margin of both processes to align with each other (Figure 7). Furthermore, there is a small decrease in the relative length of the mandibular diastema in larger specimens.

3.3.3. Shape and size variation (genus level)

Procrustes ANOVAs per genus resulted in significant correlations between shape and independent variables size and location for some genera, but not in others (Table 1; Table S3). For the genera Selevinia, Myomimus and Glirulus, only specimens originating from the same location (country) were present in the dataset. As no variable for locality could be included, size was used as the single independent variable for the analysis. In addition, these genera were relatively underrepresented within the dataset (1, 3 and 4 specimens, respectively); therefore, the correlation between size and shape could not be assessed. Muscardinus showed a significant correlation for both cranial and mandibular shape with location, but not with size. The same trend was observed in the genus Dryomys, with location being significant in both instances, even though size was not significant for the mandibular dataset and only just significant for the cranial dataset (p = 0.041). The larger dormouse genera Glis and Eliomys showed significant correlations for both location and size on shape, as well as significant interaction terms. The species‐rich genus Graphiurus indicated significant shape correlations as well as in both datasets, although the interaction term between location and size for the mandibular term was not significant.

TABLE 1.

Procrustes ANOVAs per dormouse genus for both cranial (top) and mandibular (bottom) datasets, evaluating the shape variation with respect to centroid size and location: shape ~ size × location

Subfamily	Genus	# specimens	R 2 size	R 2 location	R 2 interaction
Leithiinae	Dryomys	15	0.11*	0.52*	0.13**
	Eliomys	35	0.06**	0.48**	0.19**
	Muscardinus	11	0.11	0.46**	0.09*
	Myomimus	3	0.68	NA
	Selevinia	1	NA	NA
Glirinae	Glirulus	4	0.49	NA
	Glis	12	0.28**	0.60**	0.04**
Graphiurinae	Graphiurus	30	0.17**	0.39**	0.1*
Leithiinae	Dryomys	15	0.07	0.63**	0.11
	Eliomys	37	0.05*	0.50**	0.13**
	Muscardinus	11	0.15	0.44**	0.05*
	Myomimus	3	0.56	NA
	Selevinia	1	NA	NA
Glirinae	Glirulus	4	0.48	NA
	Glis	12	0.18**	0.72**	0.04**
Graphiurinae	Graphiurus	30	0.06**	0.41**	0.1

Relevant and statistically significant values are bold, with the p‐values summarised with asterisks (*p = 0.01–0.05; **p = 0.001–0.01).

Abbreviation: NA, not applicable (only one specimen and/or location included).

3.4. Biomechanics

Between the various dormouse genera, a significant variation of MA is present (Figure 8; Table S4). The genus Muscardinus appears to have the most effective mandibular morphology for mastication, showing the highest ratios in five out of the eight analyses. The MA of the temporalis in Glirulus is relatively high for both gnawing and chewing, whereas the posterior part of the superficial masseter results in a high MA for Glis. The highest outlier in the MAs for Graphiurus appears to be the mandible of G. crassicaudatus. The outlier for Glis is the relatively large G. g. persicus.

FIGURE 8.

Boxplots indicating the mechanical advantage (MA) in dormice based on in‐ and out‐lever analyses of the mandible. The muscle MA is calculated for the Temporalis (a, b), most ventral attachment site of the superficial masseter (c, d), the most dorsal attachment site of the superficial masseter (e, f), and the most anterior point of the deep masseter (g, h). Plots on the left represent MA in incisor biting (gnawing). Plots on the right represent a molar biting scenario (chewing)

3.5. Shape and size variation (species level)

Eliomys, Glis and Graphiurus were evaluated as representatives of their associated subfamilies (Leithiinae, Glirinae and Graphiurinae, respectively). Shape variation within all three genera is substantial, albeit that some genera are more speciose than others. Depending on the genus, the number of specimens within the dataset and the various locations represented, different methods for assessing within‐group shape variation based on GPAs and PCAs were applied. The descriptions of the monospecific genus Glis were focused on the most distinctive specimens within the dataset. The ten most distinctive specimens were described for the genus Eliomys. This study follows Pavlinov and Potapova (2003) by splitting the genus Graphiurus into three subgenera. These are Aethoglis, which contains G. nagtglasii; Claviglis, containing G. crassicaudatus; and Graphiurus, which contains all other Graphiurus species. The subgenera Graphiurus (Aethoglis) and Graphiurus (Claviglis) are morphologically very distinct from the subgenus Graphiurus (Graphiurus). Their skull morphology was described individually, whereas a GPA and PCA were used to determine the ten most distinct Graphiurus (Graphiurus) specimens.

3.5.1. Glirininae: Glis glis

This species shows a significant correlation between shape and size, as well as shape and location, including the interaction term (Table 1). The edible dormouse G. glis is the only extant species in the genus Glis and is considered the largest extant dormouse species (Storch, 1978). The distribution of the species is mainly characterized by temperate broadleaf, coniferous and mixed forests throughout Europe and southwestern parts of Asia.

Specimens representing populations from seven different localities across Eurasia were analysed based on shape, size and ecoregion. The genus is represented by 12 individual specimens within the dataset, resulting in a relatively low number of specimens per locality. It is therefore difficult to assess whether these individuals are representative for their population. The skull and mandible size of G. glis is quite variable within populations, especially with respect to relatively small‐ to medium‐sized specimens (centroid skull size: 76.3–86.7). Dormice from Azerbaijan, Italy and Germany show similar skull sizes (86.6–86.7), although climate varies significantly between these localities. The second largest Glis specimen is significantly larger than the rest of the mainland dormice (92.4) and belongs to the subspecies G. g. persicus, originating from Iran and occupying a desert‐like habitat (Figure 9).

FIGURE 9.

Mean crania (Senckenberg‐Forschungsinstitut und Naturmuseum, Frankfurt [SMF]: 82070) and mandible (SMF: 40073) of Glis glis on the left, and the distinctive Glis glis persicus (Natural History Museum, London: 27.10.26.21) from a frontal, lateral, dorsal and ventral perspective on the right

The largest G. glis is found on the island of Sicily (92.5). Due to the tendency of dormice to become larger in isolated habitats, insular dormice were initially excluded from this study. However, the relative scarcity of G. glis within the dataset justified the inclusion of this particular specimen. Interestingly, the Sicilian dormouse is reported to have the smallest body size among all Italian populations (Milazzo et al., 2003), which does not correspond to the findings here, nor the observations by Storch (1978). The skull of this specimen is considerably flatter compared to skulls in other populations. The largest Procrustes distance to the mean is present in the Iranian dormouse (0.0540 for the skull, 0.0559 for the mandible), which is the only specimen inhabiting a desert environment. Morphologically, it is most similar to the nearby Azerbaijan population with both populations having a relatively broad rostrum. The nasal bone in the Iranian dormouse flares more laterally anteriorly compared to other populations. Furthermore, the orbit is enlarged and the midorbital constriction is narrower in this specimen, resulting in the fusion of the cranial crests, whereas they are separated in all other populations. When orientated ventrally, the muscle attachment site of the superficial masseter is expanded medially, which is indicative for more strongly developed musculature. The IOF in this specimen is elongated superiorly and the zygomatic plate less angled inferoposteriorly. The molars are severely worn, especially lingually, which has resulted in a more ventral orientation of the occlusal surface. The muscle attachment area for the temporalis muscle on the vault is less pronounced. The ventral margin of the masseteric ridge on the mandible is positioned slightly more inferior, effectively enlarging the masseteric region and attachment area for the masseter muscle. Furthermore, the angular process in the Iranian dormouse is more robust.

3.5.2. Leithiinae: Eliomys

Eliomys, Dryomys and Muscardinus all show shape variation significantly correlating with location (Table 1). Of the three genera, Eliomys is considerably more abundant within the dataset (39, 15 and 11, respectively). Furthermore, this is the only genus with significant correlations between shape and both size and location, as well as the interaction term, for both datasets. The garden dormouse genus Eliomys contains three extant species. The species are separated geographically from each other, occupying parts of Northern Africa, Europe and the Middle East. E. quercinus inhabits forests across Europe and, in contrast to Glis glis, also the Iberian peninsula. E. munbyanus occupies the Maghreb region, inhabiting both forest and desert‐like environments (Holden‐Musser et al., 2016). The patchy distribution of the Asian garden dormouse E. melanurus ranges from the north east of Libya to the Middle East, where it occupies very arid environments.

The initial six principal components all explain more than 5% of the total shape variation, which can be an indication that morphological variance within this group is relatively subtle. Segregation between the European Eliomys quercinus and the other two species is visible on the first principal component (Figure 10). The cranial morphology of the ten most distinct specimens was evaluated by comparing the landmark positioning with that of the mean shape derived from the GPA. The Procrustes distances from the mean in all specimens varied from 0.043 to 0.027, indicating the overall morphological variation within this genus to be relatively small. Most variation is present in the rostrum, the zygomatic arch and the foramen magnum. E. quercinus tends to have an anteriorly displaced its incisive foramen with respect to the other species. Furthermore, the auditory bulla in this species is slightly less inflated. The dorsalmost point of the zygomatic arch in both E. melanurus and E. munbyanus is positioned more anterosuperiorly. The incisors in E. munbyanus appear to be more opisthodont and the foramen magnum is oriented more inferiorly. The largest Eliomys specimen in the dataset originates from Sevilla and shows clear flattening of the skull, a feature also seen in the larger G. glis specimens. Within E. quercinus, the Swiss populations and, to a lesser extent, the northern Italian populations have the dorsal most section of the zygomatic arch located more posteriorly compared to all other populations. Furthermore, in the majority of these specimens, the orientation of the foramen magnum appears more caudal.

FIGURE 10.

(a) Principal component analysis of the cranial dataset for the genus Eliomys, displaying the first two principal components describing 29.5% of the total shape variation within this group. (b) The common allometric component analysis on the cranial dataset

The PCA on the mandibular dataset of Eliomys also has six principal components explaining more than 5% of the total shape variation. The Procrustes distances are larger than those seen in the cranial dataset, varying from 0.079 to 0.033 from mean. The PCA shows clustering of the Maghreb dormice at more positive values of the first principal component, whereas one specimen of E. munbyanus is not placed here (Figure 11a). Two E. quercinus mandibles from Sevilla are the largest mandibles within the dataset and clustered with the Maghreb dormice. Furthermore, other relatively large Eliomys specimens are associated with positive values on the PC1. The common allometric component analysis indicates that most E. munbyanus and E. melanurus represent larger E. quercinus specimens, whereas not increasing in size themselves (Figure 11b).

FIGURE 11.

(a) Principal component analysis of the mandibular dataset for the genus Eliomys, displaying the first two principal components describing 35.6% of the total shape variation within this group. (b) The common allometric component analysis on the mandibular dataset, indicating that E. munbyanus and E. melanurus are morphologically similar to larger E. quercinus specimens

3.5.3. Graphiurinae: Graphiurus

Within the Graphiurus subgenus, including a total of 13 different species, significant diversity in cranial morphology is apparent (Figures 1 and 3). The dataset used in this study incorporates eight out of the thirteen extant species. With 14 specimens, the species G. murinus is overrepresented in the dataset, but does show clear morphological variation between geographically separated populations (Table 1).

Based on the landmark configuration, the positioning of various cranial elements was analysed. Larger specimens appear to be characterized by a flattening of the vault, a trait seen also within Glis and Eliomys. Furthermore, enlarged specimens tend to have the foramen magnum positioned more superiorly, resulting in the foramen being oriented more caudally. Smaller specimens are characterised by a shortened rostrum, with the anterior part of the nasal bone deflected more inferiorly. The cranial vault in these specimens is relatively enlarged and more rounded. The auditory bullae are more inflated in specimens occupying dry areas. The rupicolous species G. rupicola and G. ocularis both show a relatively narrow zygomatic arch and an inflated auditory bulla, whereas only the larger G. rupicola has a clearly flattened skull. The G. ocularis in this study is rather small, whereas this species is often regarded as the second largest Graphiurus species (after G. nagtglasii; Holden‐Musser et al., 2016). It is unclear if this specimen was labeled incorrectly, or simply represents a relatively small individual. The overrepresented G. murinus is known from three localities: Democratic Republic of the Congo, Tanzania and a single specimen from Kenya. The Tanzanian specimens appear to be larger, with the exception of specimens inhabiting ecoregion 7 (Tropical and Subtropical Grasslands, Savannas and Shrublands). The populations from Congo are also assumed to inhabit this ecoregion, whereas the average‐sized Tanzanian populations occupy ecoregion 1 (Tropical and Subtropical Moist Broadleaf Forests) and the largest specimens ecoregion 14 (Mangroves). The specimen from Kenya is from a desert environment and is of average size. The cranial morphology of this specimen is very similar to the mean shape of G. murinus. However, the mandibular shape of this particular specimen is very distinctive compared to other G. murinus mandibles. Other mandibular and cranial features, including the incisive foramen and the dorsal bending of the zygomatic arch, appear variable within Graphiurus, although not distinguishing between species and/or ecoregion.

The monospecific subgenus Aethoglis includes the largest extant Graphiurus species, G. nagtglasii. This specimen is known to be primarily arboreal and rarely seen on the ground (Holden‐Musser et al., 2016). Alongside its relatively large size, the skull of G. nagtglasii has a large rostrum. The zygomatic arch is relatively robust and the posterior section of the skull appears to be angled inferiorly, resulting in a bulbous cranial vault and a more ventral orientation of the foramen magnum. The auditory bullae are relatively small. The molars within this species appear to be less concave and more complex (molar ridges) and robust compared to most other Graphiurus specimens. The mandibular shape of this species is not as distinctive as the cranium (Figures 3 and 4).

Graphiurus crassicaudatus is considered as the sole member of the subgenus Claviglis and the most herbivorous of all Graphiurus species. The zygomatic arch in this species is very pronounced, being more robust than in any other species, including muscle attachment sites on the zygomatic plate that are very well defined. Furthermore, distinct temporal crests are present, as also seen G. glis, although the crest morphology in the two species is different. The orbital constriction in G. crassicaudatus is very broad and the rostrum relatively shortened and quite narrow. Lastly, the species shows clear lateral ridges on its robust molars and displays molarised premolars, whereas molars in other Graphiurus species tend to be relatively small and concave and lacking distinctive molar ridges. The size of the G. crassicaudatus mandible is similar to that of the relatively large G. angolensis, whereas the skull of this species is quite small, only slightly larger than the small G. lorraineus (see also Figure 1).

4. DISCUSSION

Clear morphological variation between and within dormouse populations is present. As well as distinguishing between genera, this study identifies particular changes in shape associated with size (allometry), whereas other morphologies are linked with specific ecological parameters such as habitat and diet.

4.1. Intrageneric versus intergeneric morphological variation

With clearly separated genera in shape space, it is evident that the intergeneric variation across the whole family is more pronounced compared to shape variation within genera. However, intrageneric variation in morphology is present and sometimes exceeds intergeneric variation. This is the case for cranial morphology in two subgenera within Graphiurus, Claviglis and Aethoglis, on the first two principal components (Figure 3). Instead, these dormice appear more similar in shape to the Leithiinae genera Dryomys and Eliomys. This is presumably the result of the relative broad zygomatic arches and long molar rows in these species. Furthermore, G. nagtglasii (Aethoglis) is the largest Graphiurus species, similar in size with Eliomys specimens. The desert dormouse Selevinia betpakdalaensis best resembles dormice within Graphiurinae in the PCA, whereas, phylogenetically, it is placed within the Leithiinae (Hennekam, Sadler, et al., 2020). This is the result of the extremely reduced molars within this specimen, corresponding to the relatively short molar rows seen in Graphiurus. The Selevinia skull also suggests morphological similarities with Dryomys (PC2) and Myomimus (PC4), in line with previous studies of this specimen (Hennekam, Sadler, et al., 2020). In both datasets, the hazel dormouse M. avellanarius is very different compared to other species. This is especially clear when examining shape variation within the mandible. This small arboreal dormouse has highly modified molars, displaying pronounced lateral ridges specialised for eating the tough food it was named after, hazelnuts. Other than this unusual species and the Graphiurus subgenera Claviglis and Aethoglis, the first principal component for skulls and the first two principal components for mandibles distinguish between the three dormouse subfamilies. However, the first principal components do not represent changes in size, even though size variation within the family is significant. This suggests that the allometric effect on shape is relatively small within dormice.

4.2. Allometric patterns

Size is a distinguishing feature of dormouse genera (Storch, 1978), and the extent of size variation differs between species (Figures 1 and 5). It should be noted that due to the scarcity of certain dormouse species, some genera are underrepresented. The maximum within‐genus size difference is therefore expected to be larger than displayed in Figure 6. This absence within the dataset is exceptionally notable for the genus Eliomys, for which insular specimens were excluded from this study, but the largest population of which is found on the island of Formentera (Hennekam, Benson, et al., 2020; Storch, 1978). The results indicate that relatively large dormouse genera (Eliomys and Glis) are more variable in size than smaller genera (Muscardinus and Glirulus). This can partly be explained by the large sampling of Eliomys (39 specimens) and the poor sampling for Glirulus (4 specimens). Nonetheless, Muscardinus and Glis are relatively equally represented in the dataset (11 and 12 specimens respectively), with size variation in the larger Glis species being clearly more pronounced.

Analyses indicated that the interaction effect of size and genus was significant, and had a smaller residual sum of squares than the analyses without the interaction. However, as the differences between the models were not significant, the common allometric trajectory can be used to evaluate the allometric signals within the datasets (Table S2). This is probably the result of non‐significant correlations between shape and size within certain genera. Significant correlations between size and shape are more common in larger dormouse species (Glis and Eliomys; Table 1). Smaller‐ to medium‐sized dormice show significant correlations between shape and locality and no significant correlation between shape and size, although this could partly be explained by the relatively small sample size. Shape variation within some genera represents morphological variations between populations. This appears to be the case for Dryomys and Muscardinus. This could not be tested for other small dormice analysed in this study, as the dataset did not include spatially separated populations for these species (Table 1).

For both datasets, the covariate ‘location’ explains a substantial amount of the total shape variation. Size shows a significant, although smaller, impact on the morphology within dormice as well. The cranial dataset indicates more significant correlations between size and shape compared to the mandibular dataset (Figure 7), with size explaining a larger amount of the total shape variation in the cranium (Table 1). This is also evident in the medium‐sized genus Dryomys, which shows significant correlations of both size and location with its cranial morphology, whereas location was the only significant variable for the mandibular shape.

The common allometric component for the cranial dataset shows a clear distinction in size between the large dormouse genera (Glis and Eliomys) and all other genera, with the exception of the peculiar Graphiurus nagtglasii (Figure 7). The warps along the allometric trajectory show an elongation of the rostrum, flattening of the cranial vault and relative reduction of the auditory bullae and foramen magnum corresponding to an increase in size. However, the allometric signal in the mandibular dataset is less pronounced. The relatively small genus Myomimus actually resembles larger genera (Figure 7). Muscardinus appears to be morphologically very different from other small dormice, only resembling the desert dormouse Selevinia to some extent. Increasing the size of mandibles is associated with a relative shortening of the condylar process and of the diastema. This shape change is very similar to that observed along the third principal component, having the small Muscardinus and large Glis on either end of the axis (Figure 4). These shape changes are seen to a much lesser extent in the relatively large genus Eliomys. This indicates that the allometric trajectory is not just depicting morphological variation associated with size, but could represent shape variations related to genera occupying various ecological niches. The mandibular allometric trajectory is not nearly as well defined compared to the cranial dataset, even though mandibular size is quite variable within dormice.

It appears that although some cranial structures within dormice can be explained by allometry, a variety of features appear to be species‐specific. In particular, small dormice vary significantly in morphology, from a very reduced molar row in Selevinia to extremely large molars in Muscardinus. Selevinia is also characterised by exceptionally inflated auditory bullae, whereas these are heavily reduced in Glirulus. Larger dormouse genera, like Glis and Eliomys, are morphologically very different from each other. However, shape changes within these genera as size increases appear to be relatively similar, indicating similar allometric trajectories. Smaller individuals are characterised by a shortened rostrum and a relatively bulbous cranial vault. Larger specimens show a clear flattening of the cranial vault and a relatively small auditory bulla and foramen magnum. It seems that mandibular morphology in Gliridae is strongly influenced by parameters other than size, including diet, climate and habitat.

4.3. Associations between cranio‐mandibular morphology and ecology

4.3.1. Geographical divergence and habitat‐specific morphologies

Location explains a large portion of the total shape variation within dormice. However, this variation could be the result of shape variation between species. Then again, speciation within dormice, unsurprisingly, appears to be closely linked with geographical separation of populations and variation in occupied niches. It is therefore of interest to understand the shape variation linked with locality, as it might be indicative for modes of life. Within the family, the shape of the zygomatic arch appears highly variable within species, but only appears to separate species and populations for the genus Eliomys, and to some extent in Graphiurus. The same applies for the position of the incisive foramina. A morphological feature more clearly linked with ecology is the inflated auditory bullae in specimens associated with dry desert‐like environments, as described by Lay (1972) and especially visible within the desert dormouse S. betpakdalaensis and certain Graphiurus species. Interestingly, the auditory bullae appear not to be inflated in the enlarged G. glis occupying a desert‐like habitat (Figure 9), which might indicate that the relative increase in auditory bullae as an adaptation to a desert‐like environment is not or less applicable to larger species.

4.4. Diet‐related morphologies

Specific cranial and mandibular morphologies are associated with preferred dietary resources. Landry (1970) stated the ancestral state for Rodentia was omnivory, rather than herbivory, based on the mandible. All species within Gliridae appear to be omnivorous to some degree, although some are clearly more faunivorous or herbivorous than others (Holden‐Musser et al., 2016; Potapova & Rossolimo, 2008). The reduced molar rows in Selevinia and Graphiurus ocularis are thought to be related to a highly insectivorous diet (Hennekam, Sadler, et al., 2020; Holden‐Musser et al., 2016; Webb & Skinner, 1995). Glis, Glirulus, Muscardinus and the thick‐tailed African dormouse G. crassicaudatus are considered to be the more herbivorous dormice. However, Glirulus is known to prefer insects in captivity, suggesting it is more an opportunistic omnivore than an herbivorous dormouse. All other species are simply classified as omnivorous, even though clear variation in diet between species is present (Holden‐Musser et al., 2016). The rupicolous dormouse Graphiurus rupicola is regarded to be more insectivorous than the arboreal G. nagtglasii. Species like G. murinus and E. quercinus are known to have a variable diet, related to their location (Holden‐Musser et al., 2016; Kahmann & Lau, 1972).

Dormice exhibit pseudomyomorphy with respect to their ways of mastication (Vianey‐Liaud, 1985, 1989). The classical sense of masseteric conditions within rodents are protrogomorphy, hystricomorphy, sciuromorphy, and myomorphy. The family Gliridae is unique in that it is the only rodent family in which multiple masseteric conditions exist. This distinct variation in the masseteric configuration results in the IOF to be highly variable between dormouse genera (Hautier et al., 2008). The primitive protrogomorph condition is present in extinct dormice species like Gliravus majori, based on the small IOF and zygomatic arch (Hartenberger, 1971). Two forms are present within extant dormice, with the subfamilies Glirinae and Leithiinae being myomorph (Simpson, 1945); and the Sub‐Saharan Graphiurinae showing a more hystricomorph condition (Hautier et al., 2008; Tullberg, 1899; Wahlert et al., 1993). The development of the myomorph condition within dormice is convergent and evolved independently compared to the myomorph conditions in muroids. This myomorph characteristic in modern glirids is therefore considered a pseudomyomorphy (Vianey‐Liaud, 1985). The landmark configuration used in this study included three landmarks along the borders of the IOF (Figure S1). When comparing the shape of this feature between genera, Graphiurus shows a more laterally flared foramen (Figure 7), in agreement with the hystricomorph condition proposed for this group. However, alongside lateral flaring, clear variation in the shape and relative size of the foramen exists both between and within genera. This is an indication that this feature is not morphologically restricted, which in turn could explain why two forms are present within Gliridae. Interestingly, G. crassicaudatus and G. nagtglasii both have the most lateral point of the foramen superiorly orientated, again showing distinctly different morphology in comparison with other Graphiurus species. The form of the foramen in Muscardinus is also unique within the subfamily Leithiinae, showing relative resemblance with members of the subfamily Graphiurinae instead.

More herbivorous dormice are characterised by a relatively robust mandible, with Muscardinus displaying a highly unusual mandibular shape for a dormouse (Figures 4 and 6), including a very distinctive molar row (Wahlert et al., 1993). Herbivorous rodents are characterised by a more robust zygomatic arch and a more massive skull in order to withstand the stresses imposed by mastication (Samuels, 2009). In dormice, the zygomatic arch is more robust in herbivorous species like G. glis and G. crassicaudatus, and orientated more horizontally than in most other species. This results in the arch being in line with the occlusal surface of the molar row within these species. Furthermore, pronounced temporal crests are seen in various herbivorous dormice, whereas these are absent in specimens with a more faunivorous diet.

4.4.1. Dental morphology

Dentition in dormice is very variable and seems to correlate well with dietary preferences (Wahlert et al., 1993). Small concave and simplistic molars are associated with the more insectivorous species, whereas large robust molars, often including pronounced lateral ridges, are seen in dormice with a more herbivorous diet. The molar row within most Graphiurus specimens is relatively small with respect to other genera, with the exception of Selevinia, in which this feature is extremely reduced. The two genera are phylogenetically distant from each other and reside in two different parts of the world (Hennekam, Sadler, et al., 2020). Graphiurus species occupy a wide range of ecological habitats in Sub‐Saharan Africa, whereas the monospecific Selevinia is only known from the desert plains in Kazakhstan. However, both genera are considered to be relatively insectivorous (Holden‐Musser et al., 2016). Molars in both genera are generally concave, lacking the complex architecture of ridges seen in many other dormouse species (Hennekam, Sadler, et al., 2020; Wahlert et al., 1993).

The degree of molar development appears to be connected with dietary preference within the highly variable Graphiurus genus. The genus is relatively speciose and displays differences in molar characteristics between some populations. Despite being considered omnivorous in general, the dietary preferences within Graphiurus span from the herbivorous G. crassicaudatus to the insectivorous G. ocularis and G. rupicola species. More herbivorous species show clear adaptation to their diets: G. nagtglasii appears to be more herbivorous than most other Graphiurus species and displays more highly developed molar ridges (Figure 12). These ridges are even more pronounced in the predominantly herbivorous G. crassicaudatus, but lacking in the more omnivorous and insectivorous species (e.g. G. murinus, G. rupicola). The patterns seen in Graphiurus resemble morphological adaptations in other genera, with the extremely reduced molars in the insectivorous G. ocularis and the desert dormouse Selevinia, and more developed molar structures in herbivorous specimens such as G. crassicaudatus and genera like Glis and Muscardinus.

FIGURE 12.

Ventral orientation of dormice allocated to specific dietary ecologies. Genus abbreviations: D, Dryomys; S, Selevinia; G, Graphiurus

4.4.2. Arboreal/terrestrial locomotion and cranial form

Larger specimens are generally characterized by a flattening of the cranial vault and presumably a relatively small endocranial volume. This is a pattern often seen in mammalian lineages and associated with the negative allometric scaling of the brain with respect to the increased body size (Craniofacial evolutionary allometry “CREA”; Cardini et al., 2015). Furthermore, a more caudal orientation of the foramen magnum is observed in larger specimens, whereas smaller individuals have small robust rostra, an enlarged and rounded cranial vault and a more ventrally oriented foramen magnum. The enlarged cranial vault suggests a relatively larger brain volume, an adaptation generally seen in arboreal mammals (Eisenberg, 1981; Rensch, 1959). The ventral position of the foramen magnum in lagomorphs is believed to be an adaptation to the sitting behavior (DuBrul, 1950), whereas this morphology in hominids is linked with an increased use of bipedal locomotion (Russo & Kirk, 2013). However, foramen magnum orientation in non‐primates, including rodents, is not always correlated with bipedalism (Russo & Kirk, 2013; Ruth et al., 2016). Satoh and Iwaku (2008) proposed ventral positioning of the foramen in Apodemus spp. to be an adaptation to a semi‐arboreal niche, in which head posture increases the field of vision. The majority of small‐sized dormouse species are considered primarily arboreal, which corresponds to the morphological features seen in smaller specimens. A more caudally orientated foramen magnum is also seen in populations of E. quercinus from montane woodlands, a region in which this animal is known to be more terrestrial (Holden‐Musser et al., 2016).

A recurring pattern of cranial vault flattening and a caudally orientated foramen magnum is identified in large terrestrial dormice. The flattening of the skull could be the result of enlarged masticatory muscles in bigger specimens (Penrose et al., 2016) or CREA (Cardini et al., 2015). More agile dormice (arboreal) are relatively small and have a bulbous cranial vault and shortened rostrum. The foramen magnum in these species is orientated ventrally and is relatively large. Similar features are seen in the largest Graphiurus species, G. nagtglasii (Figure 13), which is documented to be predominantly arboreal and rarely ventures on the ground (Holden‐Musser et al., 2016). The smaller, but very terrestrial dormouse, G. rupicola, shows clear flattening of the skull and more caudally orientated foramen magnum. The cranial morphology of these large African dormice suggests that the position and relative size of the foramen magnum are indicators of locomotion, and not solely an artifact of differences in size.

FIGURE 13.

Posterior and lateral view of the arboreal Graphiurus nagtglasii and the terrestrial Graphiurus rupicola

4.5. Functional implications of morphological variations

4.5.1. Biomechanical analysis of the mandible

The temporalis muscle is expected to best represent the MA in gnawing at the incisor. The biomechanical analysis indicates that both Muscardinus and Glirulus show high MA during a biting scenario at the incisor. The masseteric muscles better reflect the effective mechanical efficiency during chewing at the molar row. Here Muscardinus has the highest MA when evaluating the deep masseteric muscle and the medial part of superficial masseter (Figure 8). Only for the posterior part of the superficial masseter is Muscardinus slightly less efficient than Glis. In overall, Muscardinus appears to have the highest MA on average. This genus, known to eat tough foods including hazelnuts, appears to have developed a very peculiar (see Figures 4 and 6) and efficient mandibular morphology for both chewing at the molars, and gnawing at the incisors (Figure 8). Within genera, specimens showing features hinting towards a more herbivorous (abrasive) diet score relatively high (e.g. G. g. persicus, G. crassicaudatus). Interestingly, these are all large specimens, whereas the highly efficient genus Muscardinus is a relatively small for dormouse standards. It is evident that clear differences in biomechanical efficiency are present within Gliridae, and warrants a more in depth study.

In addition to the biomechanical analyses of the mandible, specific morphological features within dormice are associated with dietary preferences. For example, the angular process of the mandible is positioned more anteriorly in the relatively insectivorous dormouse genera Graphiurus and Selevinia. The function of this process is associated with the attachment of masseteric muscles on the mandible. A similar anterior positioning of the angular process is seen in omnivorous genera with a preference to insects, including Glirulus and Eliomys. This feature seems to correspond to a lateral flaring of the coronoid process and a more angled zygomatic arch, resulting a relatively lower glenoid fossa. More insectivorous dormice, like G. ocularis and G. rupicola, show very thin zygomatic arches. The variation in location of the angular process and the orientation and thickness of the zygomatic arch are presumably connected with the masseter muscle architecture. The more anterior position of the angular process assumes a relative shortening of this muscle, as well as a reduced attachment area on the mandible. The angled zygomatic arch also influences the way this muscle attaches to the skull and potentially the direction of the muscle fibers. A more horizontally orientated zygomatic arch, an elongated angular process, and a less inverted mandibular angular process all result in an enlarged area of muscle attachment for the masseter muscles. The enlarged masseter muscles result in an increase in bite force (Maynard Smith & Savage, 1959). The zygomatic arch and the angular process are more in line with molar rows, which potentially affects the biomechanical efficiency during chewing. These features are appear to be present in more herbivorous dormice. The reduction and shortening of the masseteric muscles due to the anterior repositioning of the angular process can be nullified by increasing the lateral flaring of this feature, resulting in similar length lever arms under a different angle. This feature further addresses the need for more in depth biomechanical analyses of this group, as the current estimation of the MA is based on lever arms, rather than moment arms.

The lowering of the glenoid fossa and the lateral flaring of the coronoid process seen in relatively insectivorous dormice may have implications on the maximum gape of these animals. As the auditory bullae in dormice are considered inflated with respect to other rodents, these mandibular morphologies might be needed in smaller dormice in order to increase potential gape (Nikolai & Bramble, 1983). Specimens lacking these morphological features, including Glis, might therefore not be dependent on an increased potential gape, or are not as constrained by the inflated auditory bullae compared to smaller dormice.

5. CONCLUSION

The variation in cranial and mandibular morphology of eight out of nine extant dormouse genera was evaluated. Dormice appear to vary in shape and size significantly between genera, and to a lesser extent between and within species. Larger dormice are more variable in size than smaller dormouse genera, with the exception of the speciose genus Graphiurus. The common allometric component identifies a short rostrum, a bulbous cranial vault and a large ventrally orientated foramen magnum as characteristics for smaller dormice, whereas flattening of the cranial vault and a more caudally positioned foramen are seen in larger specimens. Mandibular morphology is less driven by allometry, but does show relatively enlarged angular processes in larger dormice. Specific morphologies seem to be associated with specific locomotor and dietary ecologies, and result in clear biomechanical differences between genera. This study connects the large variety of habitats and behaviours in dormice with morphological features, and clearly highlights the strong relationship between form and function in the rodent skull.

Supporting information

Supplementary Material

Hennekam, J.J. (2022) Comparative morphology of the dormouse skull and the influence of size and ecology. Journal of Anatomy, 240, 914–935. Available from: 10.1111/joa.13596

DATA AVAILABILITY STATEMENT

The 3D scans and reconstructions used in this study were uploaded onto the MorphoSource online repository: morphosource.org/projects/00000C941.