Age Classes and Sex Differences in the Skull of the Mediterranean Monk Seal, Monachus monachus (Hermann, 1779). A study Based on Bone Shape and Density



This study analyzes morphometrically 17 skulls of the Mediterranean monk seal Monachus monachus housed in different Italian Museums and collections. We considered several morphometric variables (31 linear, 1 volumetric and 1 surface area measurements). In addition, we identified, measured and compared two nonmorphometric variables, namely, the bone densities of selected areas obtained using a dual-energy X-ray absorptiometry (DXA) device. The high correlation coefficient of all variables indicated continuous growth with the onset of age. The ranking of the hierarchical cluster analysis identified the presence of three main groups containing individuals of similar sizes: lactating pups and yearlings; subadult individuals and adult females; and adult males. Smaller groups were identified within these clusters, and their respective allocations into two subgroups were argued on the basis of skull development and other factors. The discriminant analysis of the three main groups indicated a discriminant diagnostic key, based on condilobasilar length (CBlr-L); maximum mandibular branch height (MB-H); and surface area of the bulla tympanica. The proposed diagnostic key is useful to classify monk seal skulls of unidentified age and sex. The data reported here suggest that in this species certain adult skull growth features (enhanced tympanic bullae surface area extension, occipital bone density) are sexually dimorphic and possibly related to specific anatomical functions. These functions may include an enhanced auditory capacity; an increased development of the cranial musculature capable of supporting a large skull and guaranteeing the mandibular strength necessary for mastication; and male to male social interactions. Anat Rec, 292:544–556, 2009. © 2009 Wiley-Liss, Inc.


Studies on the cranial morphometry of the genus Monachus have been limited due to the rarity of the two extant monk seal species and the difficulty in obtaining complete specimens of known age and gender. Recent advances on the biology of the Mediterranean monk seal (Monachus monachus) however support the hypothesis that an osteological dimorphism may be present. The species exhibits a pelage dimorphism consisting in a dark coat with white ventral patch that is observed in newborn pups as well as in the male adult reproductive phase (Badosa and Grau,1998; Samaranch and Gonzalez,2000). In some locations, use of the terrestrial habitat (caves) appears to be partitioned in groups of individuals composed of one adult male, one or more adult females and several subadults and juveniles (Gucu et al.,2004), while other authors have suggested that adult males establish distinct aquatic territories (Aguilar,1999) which they presumably maintain to guarantee access to the reproductive females of the colony. The establishment of terrestrial and aquatic territories implies a polygynous mating system, a notion that has also been advanced by other authors based on genetic evidence (Pastor et al.,2004,2007). Since sexual dimorphism in the skulls of carnivore species has been shown to be correlated to polygynous or group-living breeding systems (De Marinis,1995; Gittleman and Valkenbourg,1997), and since the Mediterranean monk seal manifests an adult pelage morphological dimorphism, it is hypothesized that the skulls of adult Mediterranean monk seals may also be dimorphic and present osteological features that differ in size between the two sexes.

The Monachus skull was first described by King (1956) who highlighted the relevance of several osteological features. In the large-sized skulls the temporal ridges of adults form a distinct sagittal crest. The occipital crests of these skulls can also reach noticeable dimensions and thickness. Likewise, the tympanic bone is composed of a globular tympanic bulla and an irregularly shaped but prominent mastoid region. Just as in other phocids, the face region grows more quickly than the neurocranium, but the overall cranial growth seems to stop when the skull is ∼90% of its ultimate size (King,1972). This implies that the length as well as the shape of the skull, and thereby its distinctive features described above, should vary with the onset of age. Moreover, since certain features (mastoid region, occipital crest) appear more evident in large sized skulls, it is likely that differences in size and thickness of these features exist between individuals of different ages and gender and that such bone features become more ossified with time.

Considering that deposition of the bone matrix is a characteristic of vertebrate growth, it was postulated that a higher bone density value of certain skull features should be observed in older skulls. Deposition of bone matrix is indeed a dynamic process that continues through the life span of mammals as attested by various studies on bone growth and mineralization processes inferred through the bone mineral density (BMD) measurements of various domestic mammal species (Martin et al.,1981; Munday et al.,1994; Lauten et al.,2000,2001,2002; Mitchell et al.,2001; Nüske et al.,2002; Zotti et al.,2004; Ekici et al.,2005). These studies have shown that body weight is the most important intra and interspecific factor directly affecting bone mineral content and density. More recently this type of information has been considered a useful biological indicator of age in wild mammals such as striped (Guglielmini et al.,2002) and bottlenose dolphins (Butti et al.,2007). The incidence of osteoporosis is, on the other hand, rarely observed as a spontaneous pathology in female nonhuman primates (Champ et al.,1996) and has never been described as occurring naturally in other mammalian females. In light of all of the above, the BMD measurements of specific features of the Mediterranean monk seal skull were considered additional variables to correlate with standard morphometric values. The aim of the present study was therefore to determine which skull features of the Mediterranean monk seal skull change with age and between gender. To this effect we revised the morphometric approach of a previous study (Venturino,1989) and enlarged the sample size by considering further osteological material contained in other museum collections. This study considers also some of the most prominent bone features of the species such as the surface area extension of the tympanic bullae and the cranial volume, in addition to the already mentioned relative bone density of the occipital and mastoid skull regions. Since studies on the changes in shape and size of skulls also need to consider age and gender, the study also considered information on the corresponding stuffed specimens present in museums so as to provide each skull with an approximate age category.


We carried out a census of the osteological materials and stuffed specimens of Mediterranean monk seal, Monachus monachus (Hermann, 1779) housed in the main Italian museums. Seventeen of these were studied between January 2003 and June 2005 based on their suitability (integral skulls, skulls which could be easily disarticulated from the skeleton) and on the possibility of their being taken on temporary loan outside the museum premises. All examined skulls belonged to individuals originating from Italian or nearby coasts. To correlate skull osteological features to age and gender each skull was allocated, where possible, into an age class category. The categories were allocated by considering: the external morphological characteristics of the stuffed component (pelage colour, appearance and markings, genital orifices) and comparing these to the species' known external morphological characteristics described by Samaranch and Gonzalez (2000), the length of the vertebral column and skull (skeletal length), the developmental stage of the skull (for very young specimens), and any other pertinent information derived from museum records such as the gender of the specimen. An extra length of 10, 8, and 2.5 cm was added to the skeletal length of the presumed adults, subadults and pups based on the tail lengths reported by Carruccio (1893) and this length was considered as estimated standard length. In cases where only the skeleton was available, an approximate age category was hypothesized only on the basis of the estimated standard length. In the two cases in which only the zoological lengths of the stuffed components were available, age class category of the respective skulls was allocated by considering the external morphological features of the pelage such as the first postmoult fur pattern of weaned pups and the light saddle-shape scarring pattern observed in adult females and which results from mating injuries. Gender was allocated to skulls based on museum records or by observing the ventral area posterior to the umbilicus of the respective stuffed components. Females were identified by the presence of an umbilicus and four mammary nipples and a uro-genital opening just under the tail, while males were identified by the presence of a genital opening lying between the umbilicus and the anal opening (King,1956).

The hypothesized age class was allocated on the basis of an age class category matrix table (see Table 1). Age class partitions were defined by considering the available scientific information on the species' external morphology and development, thus obtaining nine age class categories each of which are characterised by specific morphological and size features. The four morphological age classes identified for pups between 0 and 5 months (Dendrinos et al., 2000) were merged with the six classes, encompassing individuals of all ages from newborn pups to adult males, identified by Samaranch and Gonzalez (2000). However, since the fourth age class, which Dendrinos et al. (2000) identify as “fat short pup” (age 1.7–5 months), partially overlaps with the second age class defined by Samaranch and Gonzalez (2000) as “youngster” (age 2–9 months), two age classes were formulated which take into account information from both studies. Age class category four, in the present work, thereby encompasses pups which have moulted the lanugo, appear bloated in shape and are 2–4-month-old which implies that they are still suckling or are presumably about to enter their weaning phase since lactation in this species is known to last for up to 4 months after birth. Age class category five instead, encompasses pups, which have been weaned and are 5–9-month-old. The compended information contributing to the nine age class categories into which museum specimens were allocated is listed in ascending order in Table 1. Females that were known to have given birth from museum records were automatically attributed to age class eight regardless of their length. In pups, age class category was also attributed on the basis of development of dentition and, in addition, information on the pelage morphological appearance. Moreover, museum records indicating the month in which the pups had been collected contributed to the allocation of pups in particular age class categories given that monk seal births peak in the month of September-October. An age class category was therefore allocated to 12 out of the 17 examined skulls repartitioned according to the following age class categories: Category three (2 skulls; unknown gender), Category four (2 skulls; 1♀, 1♂), Category seven (4 skulls; 2♂, 2 unknown gender), Category eight (2♀ skulls), Category nine (2♂ skulls). Skulls are referred to throughout the text with their catalogue number preceded by two letters indicative of the city in which their respective museum is located and by the estimated age category number to which they were allocated.

Table 1. Age class category for Mediterranean monk seals based on morphological descriptions and growth development
Age class categoryAge in monthsCategory nameDentition and descriptionMorphological appearanceStandard length (m)
  • SL, standard length is defined as the straight-line distance from the tip of the snout to the tip of the tail flesh, Scheffer,1967.

  • a

    Neves and Pires,1998.

  • b

    Dendrinos et al.,2000.

  • c

    Androukaki et al.,2002.

  • d

    Samaranch and Gonzalez,2000.

  • e


  • f


  • g


10–0.3bNewbornbNonecBlack pelage, hair in clumps0.74b
Thin appearanceb
20.3–1a,bSpindle woollybIncisors first erupt (0.2–.3)cBlack pelage, sometimes yellow patch1.0a,b
Uniform wooly coat
Body spindle shaped
Umbilical cordb
31–1.7bSpindle-patchybEruption continues (canines and molars)cBlack pelage1.0–1.38
Areas of alopecia (starting from head and flippers etc.)b
4Proposed age: 2–4 (1.7–5)bFat short pupbEruption finished (1.0–.1 months)cGray back and lighter abdomen1.4–1.6b
Original abdominal patch only visible on lateral borders
Body appears bloatedb
5Proposed age: 5–9 (2–9)dWeaned Youngster or Yearling Light gray1.4–1.6b
Nape and throat continuous
Frontal hood present
No lateral hood
Belly and back continuousd
68–2Juveniled Dark gray1.46–1.8d
Nape and throat continuous
Frontal hood present
No lateral hood
Belly and back continuousd
718+d to 48Medium Gray Seald (mostly immature males and some females) Medium gray/Dark gray 
Nape and throat continuous
Frontal and lateral hood present
Back discontinuousd2.10–2.62d
Note: Gravid (SL = 2.23;2.26;2.29;2.34d; 2.13; 2.19e,f) and mature females of 3–4 years old (SL = 1.85, 1.95g) have been observed in this size category
8 Large Gray Seald (mostly females) Dark gray/Medium gray 
Nape and throat continuous
Hood present
Frontal and lateral hood present
Back discontinuous
Dorsal sashd
9 Black adult maled Black2.10–2.70d
Nape and throat discontinuous 
No hood
Back discontinuousd

Skull Data Collection

The skull of the Mediterranean monk seal (Fig. 1A–E) possesses the typical morphology of true seals, with absence of the supraorbital process, caudal end of the nasal bones placed posteriorly between the frontal bones, and flat rounded tympanic bullae. In particular, the skull of this species is wide at the zygomatic-temporal arch, and the infraorbital canal is very short and opens with a large infraorbital foramen, suggesting a massive trigeminal innervation of the vibrissae. Morphometric (linear, volumetric and surface area) and bone density data were collected for the 17 examined skulls. The skull osteometric linear variables that were taken into account were those considered by Venturino (1989). These consisted of 31 measurements that were recorded with a plastic manual caliper with maximum length capacity of 130 mm having a 0.1 mm precision capacity and a stainless steel manual caliper with maximum length of 700 mm having a 0.1 mm precision capacity. The morphometric measurements considered are listed in Table 2.

Figure 1.

Composition of images of the skull of Monachus monachus; A, lateral (adult male, age group 9); B, lateral (young male pup, age group 4); C, dorsal (adult female, age group 8); D, frontal (adult male, age group 9); E, caudal (female, age group 7); F, ventral (limited to the basisphenoid area to show the morphology of the tympanic bullae); G, dorsoventral and ventrodorsal radiographs.

Table 2. Variables considered for each skull in the present study
Variable typeVariable numberDescriptionAcronim
Morphometric1Condilo basal length(CB-L)
2Basal length(B-L)
3Palatal length(P-L)
4Distance between the left basion and the infraorbital(BasInf-D)
5Superior dental row length(SupDentR-L)
6Superior maxilla dental row length(SupMDentR-L)
7Left mandible length 1(Mand L1)
8Zygomatic width(Zy-W)
9Acoustic meatus width(AM-W)
10Maxillary width(Mxry-W)
11Palatal width(P-W)
12Maximum right mandibular condyle width(MaxMC-W)
13Palatilar length(Ptlr-L)
14Basilar length(Blr-L)
15Condilo basilar length(CBlr-L)
16Occipito nasal length(ONas-L)
17Nasal bone suture length(NBS-L)
18Nasal bone width(Nas-W)
19Snout width(Sn-W)
20Skull width(Sk-W)
21Distance between articular processes(AP-D)
22Occipital foramen width(OF-W)
23Upper left canine height(Can-H)
24Distance between the articular coronoid condyles(ACorCon-D)
25Maximum mandibular branch height(MB-H)
26Occipital condyle width(OC-W)
27Inferior left maxilla dental row length(Inf MdentR-L)
28Mandibular simphysis length(MandS-L)
29Left mandible length(Mand-L2)
30Inferior dental row length(InfDentR-L)
31Distance between orbital processes(OrbProc-D)
Volumetric32Cranial volume(CC)
Densitometric33Region B density(Bul-Den)
34Region O density(Oc-Den)
Surface area35Tympanic bulla surface area(Bul-SA)

Skulls were photographed laterally (left side), frontally, posteriorly, dorsally and ventrally with a digital camera. The camera was mounted on a copy stand lying at a 90-degree angle with respect to the photographed horizontal plane. The skull was rested on a horizontal plane having a posterior perpendicular vertical plane. Each skull was rested on the horizontal plane, with the following features lying perfectly flat against the vertical plane: the outer edges of the left maxillary process (lateral view photo), the posterior extremity of the occipital condyles (frontal view photo) and the frontal side of the rostrum (posterior view photo). The dorsal photo was taken by placing the skull on the horizontal plane such that the complete ventral side of the lower mandible was resting on the horizontal plane and by placing a support under the cranial box in such a way to guarantee full contact between the surfaces of the upper mandibular condyle and the respective lower mandibular groove. Ventral photos of the skull were obtained by placing the skull with its dorsal side lying against the horizontal plane. Caudo-rostral and lateral oscillations of the dorsal plane, due to the presence of large sagittal and occipital crests, were avoided by placing a supportive wedge on each side of the frontal bones. A bubble level was then placed on the ventral plane of the skull, at the height which extends from one tympanic bulla to another, and the position of the skull was then modified with minor adjusting tilts in two perpendicular directions until the bubble level indicated that the ventral plane of the skull was level. Image analysis of the photographs was carried out with the image analysis software, Image Pro Plus (© Media Cybernetics, Bethesda, MD). Skull morphometric measurements that could not be measured by manual caliper due to the presence of a metal fixture holding the upper and lower mandibles, (i.e. palatilar and palatal length, palatal width, mandible length, maximum mandible height, maximum mandibular condyle width) were measured, where possible, through image analysis of the skull photographs. The surface area of each tympanic bulla (Bul-SA), intended as its maximum dimensional extension along the horizontal span of the skull's ventral plane, was also measured through computer image analysis of the photograph of the ventral plane of each skull. The bullae's contour was traced along its perimeter so as to include the ducts situated along the four corners of the bulla. Calibration for each computer image analysis measurement was done by calibrating the measurement units to the known units of a 10 cm ruler previously placed in each photograph, at the height of the measured structures.

Cranial capacity (CC) was measured by closing all foramen and fractures with soft putty and then filling the skull with small grained rice until the upper edge of the occipital foramen. As the skull was filled with rice, the rice was slightly compressed and the skull was oscillated back and forth for the rice grains to settle evenly throughout the cavity. The rice was weighed and then placed in a 1000 cc graduated cylinder to calculate the volume of the CC, expressed in cubic centimeters. Four trials of this procedure on two skulls of differing sizes initially indicated a discrepancy of <1cc differences, representing <0.3%–0.4% error differences amongst replicas. This discrepancy was considered marginal so that measurements of the CC were taken only once for each skull.

Each skull was radiographed using a high detail film-screen system both in a dorso-ventral and in a ventro-dorsal projection to evaluate the integrity of the skull bony structures. Subsequently, the BMD of two regions of each skull was measured by means of a DXA device. A dual energy X-ray source (70 kVp; 140 kVp) with a resolution of 0.3 mm was employed (Hologic QDR-1000, Hologic, Waltham, MA) and the Subregion Analysis Lumbar Spine software (version 6.20 D, Hologic,Waltham, MA) was used to analyse the scans. Each skull was thus scanned on a dorso-ventral projection in a caudo-rostral direction. The regions in which BMD was measured can be defined as follows:

  1.  Regions B: The left and right regions of the parietal and squamosal bones overlying the tympanic bullae and stretching outwards until the end of the mastoid process, anteriorly including the mandibular groove and posteriorly until the antero-lateral edge of the occipital condyle. Since these areas include the tympanic bullae the regions are indicated throughout the text as lB and rB (left and right) and their relative density values are indicated as Bul-Dens.

  2. b) Regions O: The left and right regions encompassing the occipital-parietal suture, the supraoccipital bones and the posterior edges of the parietal bones. This latter scan thereby included the sagitto-occipital crests and occipital condyles. Since these areas include the occipital bones the regions are indicated throughout the text as lO and rO (left and right) and their relative density values are indicated as Oc-Dens.

BMD for each of the regions (lB, rB, lO, rO) was calculated and reported as g/cm2. Mean and standard deviations of the left and right values for each region were calculated.

Linear Morphometric Measurements and Preliminary Statistical Treatment of Data

Surface area measurements for each bulla were conducted in four replicas. Mean surface area and standard deviations for each bulla, and for right and left bullae combined, were calculated for each individual.

Replicas of morphometric measurements obtained through computer image analysis of skull photographs, carried out randomly and compared to ones made with a caliper, indicated measurement discrepancies between the two methods of 0.5–1.2 mm. It was assumed that such differences are negligible since they represent <1%–2% error differences with respect to the total value of the observed measurement. On the basis of such marginal differences between the two methods, the missing morphometric values obtained through computer image analysis were considered satisfactory for statistical analysis. In such a way, computer image analysis of the photographed skulls allowed the calculation of 22 morphometric measurement values that could not be measured by caliper because of inaccessibility to part of the bones due to mandible and maxilla fixture.

Correlation coefficients between all variables were calculated for the morphometric measurements and for those variables, on a single basis, where data was missing. Skull morphometric measurements which could not be recorded manually with a caliper or measured by image analysis, due to missing/broken bones of the palate and zygomatic arch or bone formations which could not be observed through image analysis if the mandible was fixed to the skull (i.e. morphometric measurements n.: 12.MaxMC-W; 24. ACorCon-D; 29. Mand-L2), were thereby calculated through a regression analysis between the variable for which values were missing and the variable having the highest correlation coefficient. This was conducted on the scientific basis that, when there is a relationship between a dependent and independent variable, the resulting regression function can be used to predict values of one variable in terms of the other when the magnitude of the correlation coefficient (r), obtained from the regression function's coefficient of determination, indicates that the variables are highly correlated to each other (values close to 1) (Sokal and Rohlf,1995; Johnston and Di Nardo,1997). In cases where more than one variable had a high correlation coefficient, the variable expressing the closest dimensional direction (i.e. length or width) or functional feature (measurements of a particular area of the skull) was used. Preference was given to linear regressions followed by polynomial or exponential ones, based on the r2 value in question for each regression. Measurements obtained through regression analysis involved 13 skulls and were related to seven variables which are listed in Table 3. Once all the values for each incomplete variable set were obtained, the entire collected data set values were standardized so as to allow correct statistical comparison between all measurements.

Table 3. Summary of skulls with missing morphometric values, the highest correlated variable, and the coefficient of correlation between the two variables
Variable with missing valueNumber of skulls with missing valueHighest correlated variableCoefficient of correlation (r)
Zygomatic width (Zy-W)2Condilo basal length (CB-L)0.99
Palatal length (P-L)2Condilo basal length (CB-L)0.97
Maxillary width (Mxry-W)1Snout width (Sm-W)0.95
Palatilar length (Ptlr-L)2Condilo basal length (CB-L)0.94
Basilar length (Blr-L)1Condilo basal length (CB-L)1.0
Nasal bone suture length (NBS-L)1Distance between orbital processes (Orb Proc-D)0.91
Nasal bone width (Nas-W)1Skull width (Sk-W)0.96
Distance between the articular coronoid condyles (ACorCon-D)3Left mandible length 1 (Mand-L1)0.97

All the variables (the 31 linear morphometric variable data), the volumetric CC values, the BMD average values for the regions B and O (Bull Dens = Density of the Bulla, and Oc Dens = Occipital Density, respectively) and the average tympanic bullae surface area values were correlated to obtain a correlation coefficient matrix for all variables considered (N = 35). A Hierarchical Cluster Analysis using Euclidean distances and Ward's linkage method was run on the variables using the statistical software packet SPPS (© SPSS, Chicago, IL). A cluster analysis was used because it is designed to finding homogeneous subgroups of cases in a population or in datasets based on measured characteristics. A Euclidean distance cluster analysis was used because it calculates the geometric distance in a multidimensional space. Ward's method was used on the matrix of the Euclidean distances calculated on the observations of the 35 variables.

A stepwise discriminant analysis was subsequently conducted so as to identify those variables which were most significant in the hierarchical “clustering process” of the examined specimens. The method utilised for the selection of the variables in the linear discrimination function minimized the Lambda Wilks and the criteria for variable insertion was based on a probability level P < 0.05 while that for variable removal was superior to 0.1. A principal component analysis was also carried out on all the variables considering the groupings identified by the hierarchical cluster analysis to figuratively highlight the existence of possible correlations between the examined variables, and the existence of possible patterns between the data records.


Cranial Capacity

Data obtained with filling of the cranial cavity yielded values between 225 and 540 cm3. Figure 2 shows the distribution of values in the different individuals of the experimental group.

Figure 2.

Cranial capacity measured by rice filling of the skull expressed in cm3.

Surface Area of Tympanic Bullae

Computer image analysis of the tympanic bullae from photographs of the skull yielded surface area values with a maximum difference of 0.8 cm2 among replicas of the same bulla, and replica mean standard deviations ranging between 0.05 and 0.53 cm2. The average right and left bulla surface area values were highly correlated (r2 = 0.99), so that average surface values for left and right bullaes were computed and considered for each individual.

Bone Mineral Density

The mean and standard deviations of the cumulative left and right BMD values were very similar within each subregion (Fig. 3). The means of the cumulative lB and rB values differed by only 0.06 cm2 with standard deviations differing by only 0.18 cm2 between the right and left regions. The mean of the cumulative lO, rO values differed by 0.05 cm2 with standard deviations differing by only 0.07 cm2 between the right and left regions. Not surprisingly, lB and rB values were highly correlated (r2 = 0.97) as well as lO and rO (r2 = 0.97). The above results are indicative of the low variability in the sampling of both right and left subregions. The computed average lB and rB (Bul-Dens) and lO and rO (Oc-Dens) density values for each individual were therefore considered as single variables for further statistical analysis.

Figure 3.

Report of a skull bone densitometry showing the global region and all the subregions of interest. (R1 and R2 correspond to lO and RO, respectively; R3 and R4 correspond to rB and lB, respectively—BMC, bone mineral content; BMD, bone mineral density).

Statistical Analysis

The correlation coefficient matrix of all variables (morphometric, volumetric, densitometric and geometric) indicated that most variables are highly correlated to each other which is indicative of continuous growth among all variables with the advancement of age and implies an age related correlation amongst the variables measured in all the skulls. The results obtained from the Hierarchical Cluster analysis using Ward's method are reported in the dendrogram in Figure 4. The individuals examined appear “clustered” in three natural groupings: the first group (cluster “a”) composed of six individuals, a second group (cluster “b”) composed of eight individuals and a third group (cluster “c”), composed of three individuals. Two smaller subgroups are present within clusters “a” and “b”, though the distance between each subgroup and another is too small for the cluster analysis to identify two further distinct groups. In particular, cluster “a” is characterized by a three-individual subgroup composed of skulls belonging to pups allocated to age category three and four and another subgroup composed of one individual in category four and two individuals of undetermined age class category. Cluster “b” has a subgroup composed of three individuals in age category seven (GE714 and TS759 which are two known females, and PV379 of unknown gender), two females in age category eight and one individual of undetermined age class category (PV6307). The second subgroup of cluster “b” is instead composed of an individual hypothesized as being in age category seven and a male of undetermined age category. Cluster “c” is composed of two male individuals allocated to age class category nine and one with undetermined age class category.

Figure 4.

Hierarchical cluster dendogram obtained through the application of Ward's method on the Euclidean distances between observations of the 17 examined Mediterranean monk seal skulls.

The stepwise discriminant analysis indicated the relevance of certain variables in the discriminant model generated by the standardized discriminant functions. In particular, the analysis pointeds out that the discrimination of the three clusters (a–c) depends principally on the following variables: variable n. 15—Condilo basilar length (CBlr-L), variable n. 25—Maximum MB-H, variable n. 35—Tympanic bulla surface area (Bul-SA). The results of the analysis and the graphic representation of the discriminant analysis (Fig. 5) in fact indicate that:

Figure 5.

Graphic representation of the discriminant analysis. 100% of the cases are discriminated correctly.

  • 1Variables CBlr-L and MB-H are significantly and positively correlated with the first discriminant function.
  • 2Variable Bul-SA is significantly and positively correlated to the second discriminant function.

Figure 5 also highlights that the individuals associated to clusters “a”, “b”, and “c” all fall very close to each group's centroid and 100% of the cases were correctly discriminated.

The discriminant analysis yielded coefficients for each of the discriminating variables per function which are listed in Table 4. The abovementioned coefficients can thus be utilized to calculate the values of the variates (discriminant functions) for each case. This transformation allows one to calculate the coordinates of the points which may be plotted on the discriminating plane (as in Fig. 5). The two required discriminating functions are:

  • I° discriminating function = −21.828 + 0.059 CBlr-L + 0.291 MB-H + 0.026 Bul-SA

  • II° discriminating function = 5.258 − 0.116 CBlr-L + 0.58 MB-H + 0.53 Bul-SA

Table 4. Canonical variable coefficients
VariablesCanonical discriminant function coefficients
Condilobasilar Length (CBlr-L)0.059−0.116
Mandibular branch Height (MB-H)0.2910.580
Bullae surface area (Bul-SA)0.0260.520

The stepwise discriminant function also allows one to obtain the estimated coefficients of the classification function which was calculated in the process for each considered cluster identified through the Hierarchical Cluster Analysis (cluster a, b, and c). The coefficients of the classification function using Fisher's linear discriminating function are listed in Table 5.

Table 5. Coefficients of the classification functions using Fisher's linear discriminating function
VariablesCluster groups based on Ward's Method
Condilobasilar length (CBlr-L)1.4702.2511.921
Mandibular branch height (MB-H)2.1642.4226.861
Bullae surface area (Bul-SA)−1.569−2.999−0.131

The analysis of the principal component was applied to all the variables with the aim of identifying links and correlations among variables and records. The upper part of Figure 6 visually represents the correlation distances among the variables while the lower representation of the figure is indicative of the correlation distance of the variables with respect to the three clusters indicated along the axis and four quadrants. The principal components of cluster “a” are thus found in the upper left quadrant of the graph. Variables expressing length such as SupDentR-L, SupMDentR-L, Mand-L1, CBlr-L, Ptlr-L, (variable nos. 5, 6, 7, 15, 17) as well as width AM-W, P-W, Sn-W (variable nos. 9, 11, 19) appear closely related for individuals in group “a”. Group “b's” principal components are located in the intersection of the quadrants. Values such as CB-L, MB-H, BulDens, InfMDentR-L, Mand-L2 (variable nos. 1, 25, 27, 29) appear closely related but negatively related to variables such as AP-D or MandS-L (variable nos. 21, 28). Individuals belonging to group “c” instead, have principal components in the lower right quadrant with variables such as Bul-SA, Oc-Dens, CC very closely related, as well as Zy-W, ONas-L, OF-W, B-L and AcorCon-D (variable nos. 8, 16, 2, 24).

Figure 6.

Principal component analysis of all variables among the three groups; Bul Dens, density of the tympanic bulla; Bul Surf, surface of the tympanic bulla; CBL, condilo-basal length; Oc Dens, density of the occipital subregion; Cran cap, cranial capacity.


The morphological analysis of the skull confirms the features typical of this species. The measurements of the CC indicate values that are considerably high, even considering the possible weight of the specimens. The highest reported Mediterranean monk seal weight is 360 kg (Gamulin-Brida et al.,1965; sex of the specimen was not reported) and, although further studies are required to clarify this point, it appears as though the Mediterranean monk seal possesses an apparently highly developed brain, if compared to the brain: body weight ratio of other highly evolved vertebrates (Van Dongen,1998).

The ranking of the hierarchical cluster analysis identifies the presence of three main groups or clusters containing individuals of similar sizes. Group “a” encompasses the youngest individuals, group “b” encompasses adult females and some subadult sized individuals, while group “c” encompasses two adult males and one most likely adult male. Cluster “a” is actually composed of a subgroup of lactating pups while its kin group contains an individual (4-FI8643) that is grouped together with two individuals having larger skull dimensions. Skull FI-8643 had no associated pelage morphological or skeletal data, and given that monk seal births generally peak in September-October, this particular skull was allocated to category four based on the sole information that the pup was caught in the month of January, implying it could have been 4 months old. The likelihood that this individual may have been older than 4 months would explain its clustering in the subgroup with slightly larger skulls belonging to age class category five. Unfortunately, there is no skeletal, external morphological nor specific museum record data for individual TO5423 to hypothesize whether this may be an individual belonging to age category five or six. It is however possible to postulate that the first subgrouping (GE22, TS394, TS395) consists in lactating pups while the second subgroup involves yearlings.

Cluster “b” also seems to contain two subgroups. The largest is composed of at least two females allocated to category seven, two females in category 7 (TS759, GE714) and one skull of unknown gender allocated to category seven (PV3799). Given the majority of females in this subgroup, it is likely that PV3799 is also a female as well as PV6307 for which no data exists. The second subgroup, composed of a male (FI8646) and of an unknown individual allocated to category seven (PV3402) seems to imply that this group could also be composed of immature males and females defined by Samaranch and Gonzalez (2000) as “medium grey seals” (category seven). A closer look at the details of the skeletal component of PV3402 however, indicates that the vertebral length of this specimen is particularly long considering the skull's short condilobasal length. This could imply that the skull in question is of a younger animal and that its vertebral column has been assembled erroneously. If this were to be true, this latter subgroup could represent smaller sized males belonging to age category seven or yet even younger individuals (belonging to age category six). The latter could be credible considering that traces of Wormian bones were observed in individual FI8646 and that it is supposed that Wormian bones are symptomatic of relatively young individuals (King,1956). The third cluster, “c”, is composed of two known adult males and an unknown specimen which, given its clustering allocation, is also assumed to be an individual in age category nine.

In summary, the cluster analysis indicates the presence of three main clusters, or groups, which may be defined, according to the Samaranch and Gonzalez (2000) morphological definition, as follows: lactating and weaned pups (group “a”), large and medium grey seals consisting in subadult specimens and adult females (group “b”), and black adult males (group “c”). Within these clusters smaller groups are identified whose allocation in two smaller subgroups may be argued both on the basis of skull development, dentition, skeletal length and pelage morphology. Statistical confirmation of these two subgroupings would however require further scrutiny and a larger skull sample size including skulls evenly distributed throughout all proposed nine age class categories.

The three group stepwise discriminant analysis indicated a discriminating diagnostic key, for identifying monk seal skulls in the three age categories, based on the variables: CBlr-L, maximum MB-H and bulla surface area (Bul-SA). According to our results, new morphometric data deriving from new individuals, in the future, may be used, to position new individuals on the discriminating plane in a manner very analogous to that demonstrated in Figure 5 for the 17 skulls in the present study. Functions I° and II° will therefore enable one to envisage the individual's positioning on the discriminant plane with reference to the two functions and the three groups' positions. The coefficients of Fisher's linear discriminating function (Table 5) can instead be used to create functions capable of assigning any new Mediterranean monk seal skull to the three clusters based on the classification function generated by the seventeen analysed skulls. The conjunct use of the information provided in the two tables (4 and 5) therefore allows the classification of observations of potential new specimens on the base of the discriminating model obtained through the discriminating stepwise analysis.

The estimated classification functions derived from the linear discriminating function for each of the three groups are:

  • Cluster “a”: −144.293 + 1.47 CBlr-L + 2.164 MB-H − 1.569 Bul-SA

  • Cluster “b”: −303.451 + 2.251 CBlr-L + 2.422 MB-H − 2.999 Bul-SA

  • Cluster “c”: −402.291 + 1.921 CBlr-L + 6.861 MB-H − 0.131 BulSA

The three resulting scores, when compared, will give the predicted allocation of a given unknown individual to one of the three groups, based on the highest score obtained for any of the three given functions.

The identification of closely related variables for each of the three above mentioned groups indicated through the PCA graphic representation seems to point to certain features of skull growth that characterise the three groupings/clusters. The close relationship between certain length variables of the mouth region and skull and width variables, corresponding to the acoustic meatus, palate and snout regions of group “a”, suggests that skull growth in lactating and weaned pups up to the age of at least 9 months is occurring concurrently between the different dimensional features. Such a trend is consistent with the findings reported by King (1972) for other phocids although it does not seem to be such a marked phenomenon as it is for species like the southern elephant seal (Mirounga leonina) which is characterised by a distinct skull dimorphism.

Closely related variables of the skulls in group “b” suggests that certain skull features of immature males and reproductive females in this category are growing proportionally and with similar trends. In particular, some striking features of the skulls, like the mandible height, condyle width, and length, are growing proportionally with the total length of the skull. At the same time, the area of the skull including the tympanic bulla and the mastoid process are undergoing a distinct ossification process, as indicated by the highly correlated values of bone density in region B. This seems to imply that skulls of this age category are developing a particularly strong masticatory apparatus which is facilitated by the development of skull regions that are functional to the development of muscular features necessary for mastication and, perhaps, to tympanic bulla development.

Analysis of the variables that characterise adult male skulls (cluster c) indicate that bulla surface areas and skull bone density of the occipital region values are not only high but closely correlated to CC, zygomatic width, skull length and articular coronoid condyle distance, which implies that adult male skulls are characterised by distinctly larger and longer skulls, having a wider zygomatic and therefore suborbital face area, with features allowing enhanced masticatory power such as that represented by the high correlation of the measurement of the distance between the articular coronoid condyles. This latter feature is reflected in the observed enhanced occipital bone growth, as testified by the higher bone density of this region, which is most likely to serve the function of providing a surface area for muscular attachment able to guarantee the support of a larger skull, and of a stronger mandibular power that is probably necessary for mastication purposes as well as for male-to-male social confrontations.

The analysis of data on bone density thus yields some insight into the possible patterns of development of the skull. In particular, the bone density value of the mastoid region of the skull is closely related to specific skull features of adult females and subadult individuals, which implies that as these features increase in size, with the onset of age, so does the mastoid bone density. Similarly, the bone density of the occipital bones increases with the growth of the tympanic bullae surface area and of the skull length and width of adult male skulls. Such findings support the notion that deposition of bone matrix, in the two studied regions of the Mediterranean monk seal skull, is a continuous process at least until the attainment of the reproductive age.

The study highlights a sexual dimorphism of the Mediterranean monk seal adult male skull, the most distinctive feature of which being a strikingly well developed tympanic bulla as attested by the large surface area extension observed in the samples. The functional aspect of the tympanic bulla, and its growth over time, have not been studied in this species. Though the function of the mammalian tympanic bulla is not known with certainty, it is thought that it may improve the perception of sounds of very low and high frequencies (Dyce et al.,2002). Given that male adult M. monachus hold aquatic territories which may extend over forty kilometres of coast (Aguilar,1999; Gucu et al.,2004), it is possible that the development of the increased surface area of the bulla may be functionally correlated to an auditory capacity required for adult males in maintaining their aquatic territories for reproductive purposes.


The authors thank the personnel of the Museums of Natural History of Genoa, Florence, Pisa, Trieste, and Venice, the Museum of Animal Biology of Pavia and the Museum of Natural Sciences of Turin for their collaboration and kind concession of the museum specimens.