Allometry and advancing age significantly structure craniofacial variation in adult female baboons

Abstract

Primate craniofacial growth is traditionally assumed to cease upon maturation or at least be negligible, whereas bony remodeling is typically associated with advanced adult age and, in particular, tooth loss. Therefore, size and shape of the craniofacial skeleton of young and middle‐aged adults should be stable. However, research on both modern and historic human samples suggests that portions of the CFS exhibit age‐related changes in mature individuals, both related to and independent of tooth loss. These results demonstrate that the age‐category ‘adult’ is heterogeneous, containing individuals demonstrating post‐maturational age‐related variation, but the topic remains understudied outside of humans and in the cranial vault and base. Our research quantifies variation in a sample of captive adult female baboons (n = 97) in an effort to understand how advancing age alters the mature CFS. Craniometric landmarks and sliding semilandmarks were collected from computed tomography (CT) scans of adult baboons aged 7–32 years old. To determine whether craniofacial morphology is sensitive to aging mechanisms and whether any such effects are differentially distributed throughout the cranium, geometric morphometric techniques were employed to compare the shapes of various cranial regions among individuals of increasing age. Unexpectedly, the biggest form differences were observed between young and middle‐aged adults, rather than between adults with full dentitions and those with some degree of tooth loss. Shape variation was greatest in masticatory and nuchal musculature attachment areas. Our results indicate that the craniofacial skeleton changes form during adulthood in baboons, raising interesting questions about the molecular and biological mechanisms governing these changes.

Introduction

Bone growth (change in size over time) and development (change in shape over time) are governed by the molecular and hormonal signaling cascades that regulate and coordinate osteoblastic secretion and osteoclastic resorption of the composite protein‐and‐mineral matrix forming the molecular structure of bone (Enlow, 1975). Consequently, skeletal maturation is a complex interplay of epigenetic, hormonal, biochemical, and mechanical factors (Biewener & Bertram, 1993). Post‐natal bone growth occurs via both periosteal expansion (appositional growth) and sutural/synchondroseal/epiphyseal displacement (longitudinal growth). However, modification of the size and shape of the mature skeleton proceeds differently (e.g. Frost, 1969; Currey, 1984), as the growth sites have ossified and longitudinal growth is no longer possible. Additionally, cortical drift (remodeling) can modify the shape of preexisting bone by altering the relative rates of endosteal resorption and periosteal deposition (Boskey & Coleman, 2010). In this way, bone tissue can adapt to the environmental demands of an organism, helping to ensure that an individual's skeleton is biomechanically optimized over the course of their lifetime. However, the mechanotransductive properties (the ability of cells to sense, process, and respond to mechanical stimuli) of bone cells are altered with increasing age (Wu et al. 2011), potentially recurrently modifying the remodeling process throughout adulthood.

Fusion of the spheno‐occipital synchondrosis and eruption of the third molars are traditionally accepted as the craniofacial features in humans and non‐human primates that define an individual as being an ‘adult’ (e.g. Bramblett, 1969; Falk et al. 1989; Wang et al. 2006; Zihlman et al. 2007; Poe, 2011; Franklin & Flavel, 2014; Balolia, 2015). As it is assumed that processes of skeletal maturation largely function the same in the craniofacial skeleton (CFS) and long bones (e.g. Litsas, 2015; Wei et al. 2016), and because the limbs cease growing when individuals reach adulthood, craniofacial growth in adult individuals is traditionally considered to be nonexistent or negligible (Shea, 1998; Bastir et al. 2006; Matros et al. 2009, see also the discussion by G. Northcroft accompanying Hellman, 1927). However, an accumulation of research in the orthodontic, esthetic plastic surgery, and forensic literature suggests this assumption is invalid (see reviews by Israel, 1973b; Behrents, 1985; Mendelson & Wong, 2012; Albert et al. 2007).

There are four main deficits in the literature regarding the study of post‐maturational age‐related effects in the CFS. First, the vast majority of such research is conducted in humans (although, see Hens, 2005; Wang et al. 2007; Balolia et al. 2013; Minh et al. 2015), so it is difficult to determine what effects dietary shifts and the reorganization of the hominin cranium may have had on any general mammalian patterns of post‐maturational craniofacial aging. Second, most research considers only the facial skeleton. There are only a few examples in the clinical literature that include the cranial base or vault (e.g. Zuckerman, 1955; Bondevik, 1995). Third, research conducted on the subject of aging in the adult CFS deals almost exclusively in linear dimensions and angular measurements. Consequently, a considerable amount of data exists that demonstrates size increase of the CFS in adults, but far less is known about its development and any allometric (covariation in shape and size) trends. A few of the more recent studies use geometric morphometric methods, which are often preferred for studying shape variation as they preserve the geometry among landmarks (e.g. Bastir et al. 2006; Williams & Slice, 2010), but much more research on post‐maturational shape variation is needed. Fourth, few studies consider the effects of bone resorption independent of tooth loss (although, see Pessa et al. 1998; Mendelson et al. 2007) because, in general, the majority of research on growth trends in adult crania suggests that ‘craniofacial bony aging, contrary to popular belief, is one of expansion and not of atrophy’ (Bartlett et al. 1992).

CFS aging effects in adult humans

Considerable research on age‐related growth in the human head has been conducted, starting in 1858 with Sir George Murray Humphry who recognized that bones of the cranial vault increase in thickness with age (cited in Bartlett et al. 1992). Additional early work demonstrated that cranial growth continues throughout the life of an individual (Todd, 1924; Hrdlička, 1925; Hellman, 1927; Israel, 1968) and, in the facial skeleton at least, increases in rate later in life (Hooton & Dupertuis, 1951; Lasker, 1953). A study by Israel (1973a) indicated an overall size increase on the order of 5–7%, although a greater expansion is observed in cranial vault thickness than in global width or length (Israel, 1977). These patterns are sexually dimorphic (e.g. Formby et al. 1994; Williams & Slice, 2010) and observable in both modern and historic populations (e.g. Herzberg & Holic, 1943; Ruff, 1980; Bartlett et al. 1992).

Across a sample of five human populations that differ in geographic region of origin and chronological age, Guagliardo (1982) discovered a number of common post‐maturational aging patterns in CFS shape. These included a flattening of the frontal bone profile, increased facial protrusion, widening of various interorbital dimensions, subnasal retraction in relation to the zygomatic arches, and a general increase in the size of both the orbits and the mastoid region. He attributed these changes to a lifetime of strain in response to the stresses of bite forces and exertion of the masticatory muscles.

The majority of research on post‐maturational aging in the CFS, especially as it relates to facial rejuvenation surgery and dental orthotics for edentulous individuals, is conducted by clinicians in contemporary humans (see reviews by Mendelson & Wong, 2012; Wulc et al. 2012) or in non‐human primates, particularly those experimentally manipulated for orthodontic research (e.g. Jaul et al. 1980; Minh et al. 2015). According to a review of the aging literature by Albert et al. (2007), the following changes have been observed in various human skeletal and radiographic samples: horizontal dimensions of the face increase in size, particularly the distance between the cheekbones (i.e. bizygomatic breadth); anterior facial height (chin to nasal bridge) increases, primarily in the lower $\frac{4}{5}$ of the face; continued eruption of the teeth increases facial height, whereas temporomandibular joint (TMJ) arthritis and dental attrition can decrease it; the maxilla is often displaced posteriorly (although, see Levine et al. 2003); absolute cranial bone thickness is greater in females and increases until 65 years of age; variation in cranial bone thickness is greater in males and remains constant over time, yet absolute thickness steadily decreases with age. The authors note that these changes are small (on the order of 1.1–1.6 mm) but noticeable.

The influence of tooth loss

Both tooth loss and subsequent bone resorption are acknowledged to alter the size and shape of the anterior midface, palate, and mandible (e.g. Tallgren, 1974; Jaul et al. 1980; Xie et al. 1997; Sutton et al. 2004; Dechow et al. 2010). However, it is not until individuals reach an advanced age that the effects of such remodeling on the CFS are usually deemed a potential source of intraspecific variation. Consequently, age is often incorporated as a covariate in regression models or landmarks located in areas expected to be affected by resorption, such as alveolar bone, are omitted from further analysis. Alternatively, individuals demonstrating CFS remodeling are often avoided, if possible. Depending on the organism being studied, this practice can result in a sample comprising a range of individuals who are somatically mature but that do not yet demonstrate macroscopic signs of aging. Such a sample of wildly varying ages has the potential to bias any resulting inferences that are made.

Additionally, the potential effects of tooth loss and remodeling of alveolar bone on portions of the CFS not directly associated with the oral cavity are rarely considered. There is debate as to whether the same patterns of craniofacial variation are observed in elderly individuals who retain all their teeth (e.g. Herzberg & Holic, 1943; Pessa et al. 1998). Due to the changing material properties of bone as it ages (Colman et al. 1999; Boskey & Coleman, 2010), one might expect similar aging effects in both elderly individuals who retain all their teeth and their partially edentulous conspecifics. Alternatively, as tooth loss significantly alters the material properties of bone (Dechow et al. 2010), it may be entirely possible that the difference in craniofacial morphology is so drastic as to be easily accounted for when examining morphological variation in aged individuals. It is also unclear whether other factors, such as regional muscle atrophy, may affect the aging cranium in both humans and other mammals.

CFS aging effects in other animals

It is possible that these recognized changes in the human adult CFS are related to the fact that the introduction of cooked and less mechanically demanding foodstuffs into our diets has caused facial shortening and the resultant malocclusion issues that plague modern humans (Begg, 1954; Larsen, 1995; Lieberman et al. 2004). In addition, the conformation of the CFS has been drastically reorganized in humans to accommodate our greatly encephalized brains (Lieberman, 2008). Therefore, it is both interesting and informative to look at patterns of CFS post‐maturational change in other animals to ascertain whether the trends in size and shape change published in the literature thus far are also unique to humans. If they are, we may need to revisit our use of non‐human model organisms in craniofacial research.

Outside of humans, the effects of normal post‐maturational aging on the CFS, and in particular the portions of the skull not immediately involved in the oral cavity, are neither well‐studied nor often considered a potential source of variation in biological, evolutionary or anthropological research. The use of non‐human primates and other mammals, particularly macaques and rabbits, to explore the response of the CFS to orthodontic and endodontic treatments is popular (e.g. Legan & Nanda, 1978; Jaul et al. 1980; Ramfjord & Blankenship, 1981; Rafferty et al. 2012). Consequently, a considerable amount of evidence has been amassed that maxillary growth results from surgical intervention, even in adults, but less research has been done to uncover the natural patterns of CFS change in post‐maturational individuals.

Although ontogenetic growth and development of the primate CFS has been the subject of considerable research (e.g. Collard & O'Higgins, 2001; Mitteroecker et al. 2004a,2004b, 2005; Leigh, 2006; Singleton, 2012), very few studies extend the observational period into adulthood. However, there are three recent studies that provide evidence for post‐maturational changes to the CFS in non‐human primates. Minh et al. (2015) present data on captive adult Japanese macaques indicating that cranial length increases more than breadth, the magnitude of post‐maturational changes is greater in macaques than in humans, and most changes are observed in male, rather than female, macaques. In contrast, Wang et al. (2007) demonstrate that CFS development of rhesus macaques is characterized by faster male growth rates during adolescence (2–8 years) and an extended female growth period (8–15 years). Finally, Balolia et al. (2013) show that hominoid facial dimensions increase in size after the attainment of sexual maturity in male and female orangutans and gorillas, female gibbons, and likely female chimpanzees, although they lacked the power to conclude definitively that the trend is visible in Pan as well. In all cases, because the increases were more pronounced between younger adult age categories, the authors conclude that ontogenetic trajectories extend past the age of adolescence and that this may be a feature of female growth and development with deep evolutionary roots.

Research questions

To address the stated deficits in the aging literature and extend what little work has been done to an additional primate species, we evaluate multivariate age‐related changes in craniofacial morphology by quantifying normal variation in the aging CFS from a large number of known‐aged and diet‐controlled adult female baboons. A major benefit of using this sample is that all individuals were maintained in a similar and constant environment during the entirety of their lives, thereby greatly reducing the effects of diet and activity pattern variability. Sexual size and shape dimorphism in the CFS is extreme in baboons (e.g. Leigh & Cheverud, 1991; O'Higgins & Collard, 2002; Singleton, 2002). As the focus of this paper is on age‐related variation, we elected to study only one sex and chose females because the presence of a C‐P3 honing complex and the variable sagittal crest size in males unnecessarily introduces additional sources of variation. Allometry, or shape variation that is a consequence of differences in the size of individuals, is also an important source of intraspecific variation, particularly in baboons (e.g. Frost et al. 2003; Porto et al. 2009; Joganic et al. 2018). Thus, it will be especially important to determine whether any demonstrated age‐related shape variation in the baboon CFS can actually be attributed to allometry.

Using geometric morphometric techniques and multivariate statistics, we address three specific questions. First, are there measurable differences in the patterns of craniofacial variation as individuals age? Secondly, what role does allometry play in the manifestation of any age‐related shape variation? And, thirdly, if patterns of CFS size and shape variation are observable across adulthood, are different regions of the skull affected in the same manner? Based on previous research (e.g. Zuckerman, 1955; Bartlett et al. 1992; Bondevik, 1995), some regions are expected to demonstrate age‐related changes (e.g. the midface), whereas others are expected to remain invariant across age categories (e.g. the basicranium). These analyses will tell us whether craniofacial morphology is sensitive to aging mechanisms and whether any such effect is differentially distributed throughout the skull.

Methodology

Baboon sample

A large colony of baboons (genus Papio) is maintained by the Southwest National Primate Research Center in San Antonio, Texas (USA). The colony was founded with wild‐caught animals from a region of southwestern Kenya overlapping Tsavo East National Park (Maples et al. 1967). Careful breeding to maintain as much of the original genetic diversity as possible has been practiced and has resulted in a seven‐generation pedigree preserving 12 founding sire lines. The skulls from a portion (n = 985) of the pedigreed population were curated and scanned with computed tomography (CT) (0.6–0.75 mm resolution) at Washington University in St. Louis and are now housed at Loyola University Chicago (LUC) under the care of Dr. James Cheverud (see Joganic et al. 2018 for more details).

There are 653 adult females in the LUC baboon collection. Individuals are considered adult if they possess a fused spheno‐occipital synchondrosis and fully occluded M3s. The average age of this total sample of adult females is 19.83 ± 5.5 years (range: 6.23–33.70). Four groups were created: individuals in the bottom age quartile (Young Adult, YA), within 0.5 standard deviations of the mean age (Adult, AD), in the top age quartile with full dentitions (Elderly, ER), and in the top age quartile with some degree of maxillary tooth loss (Tooth Loss, TL). A subset of 25 individuals was randomly selected from each group, except in the case of the TL category, for which only 22 individuals could be identified who were missing at least one maxillary tooth and had complete crania (n = 97; Table 1, Supporting Information Table S1). The modal number of maxillary teeth missing in TL individuals is 1 (mean: 2.14, range: 1–5), whereas all individuals in the three other age categories have full maxillary dentitions.

Table 1.

Sample demographics

Age category	n	Mean (SD)	Min	Max	Human
Young adult (YA)	25	11.56 (3.0)	7.12	16.20	15.7–35.8
Adult (AD)	25	20.00 (1.3)	17.97	22.55	39.7–49.9
Elderly (ER)	25	27.23 (2.5)	23.76	31.69	52.5–70.0
Tooth loss (TL)	22	26.92 (2.3)	23.74	32.47	52.5–71.8

Sample composition of the different age categories, the age ranges of included individuals in years, and the human equivalent age range, based on the worldwide average female human life expectancy of 73.1 years (United Nations, 2017).

Landmarking

Our three research questions deal specifically with the effect of post‐maturational aging on differences in CFS shape. Shape is the portion of an object's geometry that is independent of overall size, position in space or relative orientation (Mitteroecker & Gunz, 2009). Linear distances calculated between landmark pairs are particularly good at capturing variation in size, both relative and overall, but are less useful for quantifying shape differences. Traditional geometric morphometrics utilizes configurations of homologous landmarks strategically placed on an object that are then compared with a reference configuration (Bookstein, 1996). This method of studying shape can be problematic, as the researcher is limited to using easily identifiable and biologically homologous traits, such as the intersection between two bones or the point of the chin. Bones of the cranial vault are relatively featureless, making the identification and placement of anatomical landmarks difficult. Semilandmarks (or sliding landmarks, SL), on the other hand, are ideal for studying smooth curves and surfaces. For our purposes here, semilandmarks allow us to capture changes in concavity/convexity (e.g. maxillary retrusion and nasal protrusion can alter the relationship between the midface and the upper and lower face as individuals age; Mendelson et al. 2007) and in profile (different CFS regions have been shown to become more or less prominent with age, properties best studied in profile; Bartlett et al. 1992) of regions displaying a very small number of traditional landmarks.

We examined three anatomical regions of the CFS in our subsample of 97 female baboons: facial skeleton (including the orbits, zygomatic arches, and palate), cranial vault, and cranial base (Fig. 1). These were selected because they are the traditionally recognized modules of the cranium relating to functional and developmental specializations (Cheverud, 1982, 1995).

Figure 1.

Craniometric landmarks, semilandmarks, and cranial regions. Locations of the craniometric landmarks color‐coded by type: anatomical (black); orbital and zygomatic curve semilandmarks, SL (white); facial, vault, and occipital surface SL (pink); frontal surface SL (gold). Only the anatomical landmarks are identified and the corresponding abbreviations are provided in Table 2. Descriptions of anatomical features used to define the surface SL patches are provided in the text. The three anatomical regions are indicated by colored shading: face (green), base (yellow), and vault (lavender).

We used avizo standard v 8.1.1 (FEI Visualization Sciences Group) to visualize the CT images, reconstruct the skulls, identify 43 craniometric landmarks (Table 2), and make surface mesh files. viewbox v 4.0.1.7 (dHAL software, Athens, Greece) was then used to measure curve and surface SL on these surface meshes by applying a thin plate spline deformation to a template landmark configuration (Gunz & Mitteroecker, 2013). This protocol has been described elsewhere (Heuzé et al. 2010); briefly, after being manually placed on a single individual to create a template, the reference landmark configuration is fitted to other individuals according to an algorithm that minimizes the bending energy while sliding SL along their identified curves and surfaces.

Table 2.

Anatomical landmarks

Landmark	Symbol	Description	Region*	Located†
Inion	INI	Most posterior point in the midline of the nuchal crest, often on an inionic hook	B	M
Opisthion	OPI	Most posterior point on the foramen magnum margin	B	M
Basion	BAS	Most anterior point on the foramen magnum margin	B	M
Hormion	HOR	Junction of vomer posterior border with the sphenoid	B	M
Posterior nasal spine	PNS	Most posterior point on the palate in the transverse plane	B	M
Incisive Canal	INC	Midline point on the incisive canal posterior margin	F	M
Greater palatine foramen	GPF	Most posterior point on the greater palatine foramen margin	F	B
Anterior petrous temporal	APT	Most anterior point on the ectocranial surface of the petrous temporal	B	B
Pterygomaxillary Septum	PMX	In the middle of the pterygomaxillary septum or on the maxillary projection if the septum is incomplete	F	B
Stylomastoid Foramen	SMF	Most lateral point on the stylomastoid foramen margin	B	B
Porion	POR	Anterior junction of the external auditory meatus and the nuchal crest	B	B
Zygotemporale inferior	ZTI	Most inferior point on the zygotemporal suture	F	B
Frontomalar foramen	FMF	Most inferior point on the frontozygomatic foramen margin	F	B
Nasion	NAS	Intersection of the internasal suture and the frontal bone	F	M
Masseter origin	MAS	Most anterior point on the inferior surface of the zygomatic arch where the masseter muscle originates	F	B
Nasomaxillare inferior	NMI	Most anterior point on the nasomaxillary suture	F	B
Acanthion	ACN	Most anterior point on the nasal aperature inferior margin	F	M
M3‐M2 Septum	MM3	Most inferior point on the bony septum between upper M3 & M2	F	B
M2‐M1 Septum	MM2	Most inferior point on the bony septum between upper M2 & M1	F	B
M1‐P4 Septum	MM1	Most inferior point on the bony septum between upper M1 & P4	F	B
P4‐P3 Septum	PM4	Most inferior point on the bony septum between upper P4 & P3	F	B
P3‐CN Septum	PM3	Most posterior point on the upper canine alveolus	F	B
Premaxillary‐maxillary junction	PMM	Most anterior point on the superior premaxillary‐maxillary suture	F	B
CN‐I2	II2	Most distal point on the upper I2 alveolus	F	B
I2‐I1	II1	Most inferior point on the bony septum between upper I2 & I1	F	B
Alveolare	ALV	Most inferior point on the bony septum between the upper I1s	F	M

Names, abbreviations, descriptions, and characteristics of the craniometric landmarks measured on female baboon crania.

^*The region of the cranium in which the landmark is located: vault (V), base (B), face (F).

^†Whether the landmark is located in the midline (M) or bilaterally (B).

The template we constructed consisted of four curves and six surfaces. An open curve of 15 SL was fit to each orbital margin (supraorbital notch to a point on the medial orbital margin at the level of nasion) and an open curve of 9 SL to each zygomatic arch (frontozygomatic foramen to superior zygotemporal suture) for a total of 48 curve SL. The 210 surface SL were divided among six patches located on the left and right face (30 SL each), forehead (30 SL), left and right vault (40 SL each), and occiput (40 SL). Facial patches covered the entirety of the anterior surfaces of the malars, nasals, maxillae, and premaxillae, except for the alveoli, pyriform aperture, orbital margins, and zygomatic temporal processes. The forehead patch was defined as the triangular area bordered by the superior orbital margin and superior temporal lines. Vault patches covered the surface of the neurocranium from the superior temporal line/crest to the nuchal crest. The occiput consisted of the trapezoidal region bordered by the nuchal crest, posterior margin of the foramen magnum, and the mastoid processes. To construct the template, each patch was identified on the surface mesh by highlighting the desired area using the paintbrush tool in viewbox. Then, we used the functions Distribute on mesh (random) followed by Diffuse points uniformly to associate the SLs with their corresponding patches. Our protocol resulted in a template with a total of 301 landmarks and SLs to be identified for each individual (Fig. 1, Table 2).

Statistical shape analysis

A generalized Procrustes superimposition (Rohlf & Slice, 1990; Goodall, 1991) to remove the aspects of variation in which we are not interested (differences in overall size, relative orientation, and spatial position) results in shape data in the form of Procrustes shape coordinates (PSCs). To address the research questions, we performed a multivariate multiple regression (‘primary shape regression’) of the PSCs on centroid size (CS), chronological age (hereafter referred to as age), and number of teeth missing (Mitteroecker & Gunz, 2009). This analysis tests the null hypothesis that post‐maturational CFS variation is not structured by cranial size, age or tooth loss. A Type II multivariate analysis of variance (manova) is then used to test whether each explanatory variable is significant after controlling for the others. Centroid size is a method of approximating size by calculating the square root of the sum of the squared distances of a set of landmarks from their centroid, the centroid being the mean x,y,z of all coordinates across all landmarks (Klingenberg, 2016). This is important for determining whether CFS shape changes with age simply because adult individuals increase in size with age or because there are shape differences between younger and older adults that cannot be accounted for by size differences among the age categories. All analyses were performed in morphoj v 1.06d (Klingenberg, 2011) and rstudio using the Morpho v 2.6 (Schlager, 2017) and geomorph v 3.0.6 (Adams et al. 2018) packages.

To visualize the shape changes related to CS and age, heat maps were created by deforming the template surface mesh in avizo (Heuzé et al. 2014). The template surface mesh was shifted into shape space by morphing it according to the coordinates of the Procrustes mean shape. The PSCs for an individual at the extreme small end of the multivariate regression of shape on CS were calculated from the regression equation (small PSCs) and the same was done for an individual on the extreme large end (large PSCs). Then, ‘small’ and ‘large’ surface meshes were created by morphing the Procrustes mean shape mesh by the small and large PSCs, respectively. Finally, a displacement field was created using the Vertex Difference function, which calculates vectors between the vertices of each triangle in the reference surface mesh and the corresponding vertices of the homologous triangle in the target surface mesh. The reference surface mesh was then colored according to the magnitude of the vectors, allowing the viewer to discern ‘hot spots’ of change between the two surface meshes. The small surface mesh was selected as the reference and the large as the target, so the visualized displacement field demonstrates regions that experience major change as individuals increase in size. The same procedure was done for age, with the reference surface representing a younger adult and the target an older one, thus demonstrating regions where age‐related differences are concentrated. It is important to emphasize that the differences represented by the heat maps are not those between an actual young/small individual and an old/large one, but represent the extremes of the shape variation vectors that are statistically correlated with age and size, respectively.

Growth and allometry

We addressed the question of post‐maturational growth in female baboons in three ways. First, we performed an ordinary least squares (OLS) regression of CS on age to test the null hypothesis that the size of the CFS is static once sexual maturity is achieved. Secondly, we performed a one‐way anova to determine which, if any, of the age‐groups differed in CS. Thirdly, we looked at the effect of size on the shape of the cranium as a whole (i.e. allometry) by examining the coefficient of the CS term in the primary shape regression.

Age‐related shape variation

We examined both age effects and the effect of tooth loss on global cranial shape variation. To identify whether shape differences exist among age groups, we calculated the consensus configuration (or mean shape) specific to each group. For example, a YA mean shape was computed as the coordinate‐wise average of the PSCs of the individuals in the youngest age category. These mean shapes were compared between the different age groups using the metric of Procrustes distance, the square root of the sum of the squared Euclidean distances between each landmark in a configuration and its homologous landmark in the sample mean configuration. To test for significant group differentiation, a Procrustes anova was performed, which models the Procrustes distances among specimens, rather than explained covariance among variables (Goodall, 1991). Then, to determine how much of the global variation in CFS shape can be explained adequately by age, we examined the coefficient of the age term from the primary shape regression.

Finally, we tested the null hypothesis that tooth loss has no observable effect on the CFS in two ways. We compared the Procrustes distances of the ER and TL groups using a permutation test to determine statistical significance. This type of analysis considers tooth loss as a binary variable, as ER individuals have all maxillary teeth present and TL individuals have lost at least one maxillary tooth. To perform a more nuanced analysis of the effect of tooth loss on CFS shape, we also coded TL individuals by the number of maxillary teeth missing and performed a multivariate multiple regression (‘dental shape regression’) using the PSCs as dependent variables and CS, age, and number of teeth lost as independent variables.

Region‐specific aging effects

PSCs were calculated separately for each of the three craniofacial regions by performing three distinct generalized Procrustes superimpositions. The resulting alignments are referred to as the vault PSCs (n = 110 landmarks), base PSCs (n = 50), and face PSCs (n = 141). A multivariate multiple regression was performed for each of the regional PSCs [‘(region) shape regression’]. All three regressions contained the same three explanatory variables as the primary shape regression (CS, age, number of teeth lost). Procrustes anova was used to test for significant age‐group differentiation in the morpho space of each craniofacial region.

Results

The effects of age on size

First, we assessed the impact that aging has on cranial size. There is a 1.8% increase in CS between the YA and AD groups, a 2.8% increase between the YA and ER groups, and a 2.4% increase between the YA and TL groups (Fig. 2). If the three older groups are combined, the pooled mean CS is 2.4% larger than the YA mean CS, and this difference is statistically significant (Welch's t‐test: ${\overline{X}}_{\mathit{YA}}=927.5,{\overline{X}}_{\mathrm{pooled}}=949.9$, t = −3.76, P = 0.0004). Age explains 14.5% of the variance in CS (OLS: t = 4.2, P < 0.001; Fig. 3).

Figure 2.

Centroid size boxplot. Comparison of centroid size distribution across age groups. Maximum: top whisker and number above it, 3rd quantile: top hinge, median: waist, confidence interval of the median: notch, mean: boldface number above the waist, standard deviation of the mean: italic typeface number below the waist, 1st quantile: bottom hinge, minimum: bottom whisker and the number below it. The confidence interval of the median is calculated as: $\text{median}\pm 1.58\times \frac{\text{IQR}}{\sqrt{n}}$, where IQR is the interquartile range.

Figure 3.

Growth maturation. Centroid size regressed on chronological age. Individuals are color‐coded by age category, as indicated by the key in the top left corner. The OLS regression line is indicated with a solid black line and its formula and associated regression statistics are provided in the bottom right corner.

The effects of size on shape

In the primary shape regression, the CS term is highly significant (F = 4.37, P = 0.001) and accounts for 4.3% of the variance in the PSCs (Table 3; Supporting Information Fig. S1). Larger individuals have a longer snout, a narrower midface, and shorter posterior facial height, smaller and more ellipsoid orbits, and a neurocranium that is more ovoid in shape. Smaller individuals are more brachycephalic and have large, round orbits and more flaring malars (Supporting Information Videos S1–S4).

Table 3.

Shape regression statistics

Shape Reg.	Centroid size			Age			Tooth loss
	R 2	F	P	R 2	F	P	R 2	F	P
Primary	0.043	4.37	0.001	0.027	2.77	0.005	0.010	1.02	0.317
Dental	0.082	1.90	0.041	0.036	0.83	0.392	0.090	2.08	0.038
Vault	0.028	2.76	0.007	0.018	1.73	0.059	0.005	0.45	0.939
Base	0.025	2.49	0.006	0.014	1.37	0.121	0.007	0.69	0.728
Face	0.099	10.97	0.001	0.038	4.21	0.001	0.013	1.46	0.066

Type II manova results from primary, dental, and region‐specific shape regressions of Procrustes shape coordinates (PSCs) on centroid size (CS), age, and the number of teeth missing. The coefficient of determination (R ²), F‐statistic, and P‐value are provided for each independent variable included in the models. Values indicated in bold are significant at an α = 0.05/3 = 0.017, a Bonferroni correction applied to account for multiple comparisons within each model (Mundfrom et al. 2006).

The effects of age on shape

To determine whether the four groups differed in CFS shape, Procrustes distances between the PSCs of the groups were calculated (Table 4). The YA group is significantly distinct from the other three in the morphospace, which cluster together (Supporting Information Fig. S2). This result is confirmed with a Procrustes anova. When comparing all four groups, the F‐statistic is marginally significant (F = 1.48, P = 0.06), but when pooling the AD, ER, and TL groups, the shape of the YA group is significantly different from that of the older individuals (F = 3.02, P = 0.006).

Table 4.

Procrustes distance matrix

Age group	Age group
	Young adult	Adult	Elderly	Tooth loss
Young adult		0.017	0.019	0.020
Adult	0.05		0.010	0.012
Elderly	0.02	0.90		0.011
Tooth loss	0.02	0.54	0.80

Intergroup Procrustes distances (above the diagonal) and their significance based on permutation tests (below the diagonal). Distances that are significantly large at α = 0.05 and their corresponding P‐values are shown in bold.

In the primary shape regression, age accounts for a modest but significant 2.7% (P = 0.005) of the sample variation (see Table 3. Young individuals are characterized by globular neurocrania, wide midfaces, and a more flexed cranial base, whereas elderly individuals have a more elongated and widened snout, prominent inion, concave nasal bridge with a retruded midface, and a flattened nuchal plane (Supporting Information Videos S5–S8). Although the size and age effects share some commonalities (e.g. both young and small individuals have shorter snouts than older and larger individuals; Fig. 4), there are some important distinguishing features. Overall, the allometric trend gives a sense of a small cranium being stretched in the sagittal plane, primarily in the facial skeleton, to produce a large one. Additionally, the premaxillary dentition is displaced inferiorly, resulting in an increasing degree of klinorhynchy as crania increase in size. In contrast, the age trend is best characterized as an anterior and superior movement of the anterior face in the sagittal plane and a recession of the midface with a concomitant protrusion of the midline upper face.

Figure 4.

Vertex difference surface morphs. Heat maps of the distances between vectors comprising the triangular surface mesh of a small individual and the corresponding vectors of a large one (left panels) and between those of a young adult and the corresponding ones of an older individual (right panels). Regions shown in red are further apart between the two meshes than those shown in blue. In other words, these are areas that change the most with increasing age and size. Because the color scales are relative, the magnitude of the differences should not be compared between the allometric and age‐related meshes.

Two methods were applied to determine the effect of tooth loss as quantified in either a binary or an ordinal manner. A Procrustes anova performed only on the ER and TL groups indicates the two are not significantly different (F = 0.64, P = 0.80). This suggests that, in a global sense, whether an individual has lost any teeth has no marked effect on the shape of the CFS. The second method, the dental shape regression, provides a more nuanced examination by testing the effect of the number of teeth lost. Although 9.0% of CFS shape variation within the TL category is accounted for by the number of teeth missing, after correcting for multiple comparisons (within‐model Bonferroni; Mundfrom et al. 2006) this effect is rendered non‐significant (α = 0.05/3 = 0.017, P = 0.038; see Table 3). As statistical power may be limited by the small number of individuals within each TL sub‐category (range: 1–10, mean: 4.6), future research may yet provide stronger evidence for a link between tooth loss and facial shape. Our results suggest that individuals with a greater number of teeth missing tend to have larger positive regression scores for the dental shape regression (Supporting Information Fig. S3), which are associated with a more arched palate, a superior and posterior shift of all alveolar landmarks, a more horizontal premaxillary profile, and a slightly wider breadth across the midface and orbits.

Size and age effects by region

According to the region‐specific Procrustes anova, there were no group differences in shape of the vault (F = 0.89, P = 0.62) or base (F = 1.16, P = 0.23), but the face PSCs differed significantly among the age categories (F = 2.10, P = 0.008). Using Procrustes distance as a metric, YA individuals are separate from all other groups in the facial morphospace (Supporting Information Table S2).

Centroid size explains 2.8% (P = 0.007) of the total variation in vault PSCs (see Table 3), with a contrast between dolicocephalic large individuals and brachycephalic small ones. Age accounts for a marginally insignificant 1.8% (P = 0.059) of the variation, with the vaults of older animals demonstrating increased postorbital constriction, a protruding glabellar region, an anterior flattening, and a particularly pointy inionic region.

Only CS has a significant effect on cranial base shape variation. A modest 2.5% (P = 0.006) of the total cranial base variation is size‐related and an insignificant 1.4% (P = 0.12) is age‐related. The major effect of allometry is a vertical compression of the basicranium in larger individuals. In contrast, age has the effect of rotating the basicranium with the fulcrum essentially located at the landmark OPI. The result is an occipital bone that is tilted backward along its superior margin at the nuchal crest, external auditory meati that extend posteriorly, a foramen magnum that is displaced superiorly at its anterior margin, and a less flexed posterior cranial base (i.e. a more horizontally positioned basioccipital).

Finally, the facial skeleton is affected to varying degrees by all three independent variables. Of the total facial variation, CS accounts for 9.9% (P = 0.001), age for 3.8% (P = 0.001), and tooth loss for 1.3% (P = 0.066). Older individuals have a longer snout that is also wider across the anterior palate compared with their younger conspecifics. Large individuals also have a longer snout, but there is no difference in width when compared with smaller individuals. Both advanced age and increased size are associated with a narrower upper face, but this is due to a disproportionate decrease in the width of the fronto‐zygomatic process in older individuals, whereas it is the width of the orbit itself that is reduced in large individuals. The glabellar and supraorbital regions are only variable along the age axis, with both increasing in prominence in older individuals. The maxillary dentition is placed more anteriorly in large and old individuals relative to small and young ones. However, the allometric trend affects the entire oral cavity, displacing both the palate and dentition forward, and age alters the depth of the palate, the relationship between the palate and maxillary teeth, and the angle between the palate and the bones making up the posterior nasopharynx.

Discussion

Much research has been conducted to describe patterns of growth and development in ontogenetic series of non‐human primates, but the effect of post‐maturational aging on the CFS remains understudied. Researchers are well aware that cranial morphology changes, often drastically, before individuals reach sexual maturity (e.g. in Papio: Leigh & Cheverud, 1991; Collard & O'Higgins, 2001; Singleton, 2012). Consequently, such ontogenetic patterns are typically accounted for, or at least acknowledged, in studies of both extant and fossil taxa. However, research on modern and historic geriatric human samples and on a select few non‐human primate populations indicates that portions of the CFS demonstrate age‐related changes in mature individuals. Our data add to this body of literature by demonstrating age‐related craniofacial variation in a sample of sexually mature female baboons.

We suggest the combination of our data and those presented in the literature lends support to the idea that there exists in primates a more general post‐maturational aging pattern that has heretofore been inadequately studied. Whether such a pattern extends to other mammalian orders largely remains to be seen, although there is some evidence for variation in both adolescent and post‐maturational growth patterns of the carnivoran CFS, with some felid genera developing past the age of sexual maturity (e.g. Segura & Prevosti, 2012; Segura et al. 2013; Segura, 2015).

Surprisingly, our data do not document a significant degree of tooth loss‐related modification of the midface and palate. Primarily, we observed morphological change near muscle attachment sites of the CFS and in individuals who retain all their teeth. These results suggest that the age‐category ‘adult’ is heterogeneous and we should be aware that it encompasses individuals demonstrating the effects of age on their CFS, regardless of their dental status.

Allometric patterns and how they differ from age‐related ones

When visualizing the age‐ and size‐related changes in the female baboon CFS, it appears that some of the regions experiencing the biggest shape changes are the same in both cases (see Fig. 4). However, there are subtle differences between the two. For example, the patterns of large‐scale increase in facial length differ when considering only age‐ or size‐related variation. The snout of older individuals is primarily elongated in the sagittal plane only. In contrast, allometry displaces the premaxilla obliquely, both forward and downward. Additionally, with the major exception of the neurocranium, which is more globular in small females and narrows with increasing size, form differences associated with allometry are almost exclusively the result of increase in size of various cranial dimensions (i.e. osteoblastic activity on ectocranial surfaces coupled with osteoclastic activity on endocranial ones). The same cannot be said for age‐related shape variation, which appears to result from osteoclastic activity on both endo‐ and ectocranial surfaces. The most obvious example of this difference is the midfacial remodeling of older females, which produces a marked concavity that is not seen in younger individuals and does not differ between small and large individuals.

As many of the regions showing localized differences between small and large crania are where major muscles attach (e.g. nuchal crest, inferior zygomatic arch, temporal surface, and pterygoid plates), it seems reasonable to suggest that an allometric increase in muscle mass with increasing body size (e.g. Alexander et al. 1981; Pollock & Shadwick, 1994) may contribute to size‐related variation in the CFS. Because the variance in adult male baboon bodyweight (27.01 ± 4.72 kg, n = 247; Leigh, 2009) is significantly greater (F = 0.513, P < 0.0001) than that of females (15.96 ± 3.38 kg, n = 215; Leigh, 2009), we would expect allometry to account for a larger proportion of craniofacial variation in a dataset of adult male baboons that was analyzed in a comparable manner to our present analysis of adult female baboons. Additionally, given that (1) numerous studies have identified sexual dimorphism in patterns of age‐related variation of human and non‐human primate CFS (e.g. Wang et al. 2007; Williams & Slice, 2010; Balolia et al. 2013) and (2) sexual dimorphism in both cranial size and shape is extremely large in baboons (e.g. Leigh & Cheverud, 1991; O'Higgins & Collard, 2002; Singleton, 2002), we could expect to identify different patterns of age‐related variation in adult male baboon crania than those we have observed in females.

The remaining hot‐spot not related to a muscle attachment site on the size‐related panel of Fig. 4 is the anterior rostrum. This is the same region that is highly sexually dimorphic in baboons (e.g. Collard & O'Higgins, 2001; Singleton, 2002; Leigh, 2006). As males are also significantly larger than females in body size (e.g. Leigh, 2009), it is not surprising that the crania of larger females appear more ‘male‐like’ in having a relatively long snout. This supports the suggestion that the extreme prognathism of Papio and its sister taxon Mandrillus is related to their increased body size relative to smaller‐bodied papionins, such as Macaca, Cercocebus, and Lophocebus. We were able to test for covariation between body and cranial size in our sample of female baboons because longitudinal bodyweight records were collected opportunistically from birth to 15 years of age (see Joganic et al. 2018 for more details). Female adult bodyweight explains a significant 14.8% (OLS: F = 17.7, P < 0.0001) of the variation in overall CFS centroid size and a varying amount of the variation within each region (face: R ² = 0.171, P < 0.0001; base: R ² = 0.124, P = 0.0003; vault: R ² = 0.054, P = 0.02). In female baboons at least, these findings are consistent with the hypothesis of Joganic et al. (2018), which posits that larger‐bodied papionin species tend to have similar‐looking longer faces because approximately 15% of the genetic variation underlying morphological variation in both body size and CFS shape is shared.

Age‐effects are global but differ by region in magnitude

According to Israel (1973a), human crania exhibit ‘virtual magnification’, whereby the percent size increase (5–7%) is uniform and symmetric across regions as adults age. The only exceptions he identified were in the sizes of the sella turcica (9–22%) and paranasal sinuses (7–12%), and in cranial vault thickness (1–11%). All other dimensions demonstrated less gain (0–5%). We found a subtly different pattern in our sample of female baboons (Supporting Information Fig. S4). The centroid size of both the cranial vault and base increased 1.8 and 1.5%, respectively, whereas in the facial skeleton it increased 3.3%. Even though the differences are slight, we suggest they may reflect what we know of baboon biology. Most of the ontogenetic change that occurs late in the CFS is primarily in the face, particularly the lengthening of the snout, especially in males (e.g. Collard & O'Higgins, 2001; Leigh, 2006). Although the focus in the literature is primarily on the effect that this phenomenon has on the male CFS, our results suggest that it may be relevant when considering female cranial morphology as well.

Most of the age‐related changes we observed in our sample are concentrated in the facial skeleton, although the nuchal region is affected also (Fig. 5A–D). As sutural fusion and subsequent obliteration likely prevent depositional growth at sutural margins, especially in the cranial vault and basicranium (see Opperman, 2000 for a review), the size and shape variation associated with increasing age in our sample are likely the result of appositional growth along ectocranial surfaces. In some regions, such as the midsagittal profile of the superior neurocranium, this is coupled with endocranial resorption that serves to maintain the same cranial vault thickness in YA and older age groups. Other areas, such as the supraorbital and inionic regions, experience a disproportionate degree of ectocranial deposition relative to endocranial resorption (remodeling), resulting in a thickening of the involved cranial bones (see Fig. 5E).

Figure 5.

Detailed depictions of age‐related changes in the baboon CFS. In each panel, two surface meshes are superimposed. They were obtained by morphing the mean shape by the fitted regression coordinates for the youngest (pink mesh) and oldest (gray mesh) individuals from the multivariate regression of Procrustes shape coordinates (PSCs) on age. In (A,B,D,F) the young mesh is partially transparent; in (C) it is the old mesh; and in (E) both meshes are fully opaque. The symbols in each panel are as follows: (A) Green triangles, nuchal ligament and medial trapezius attachment; orange triangles, deep masseter attachment; magenta arrows, primary direction of lateral pterygoid muscle force and direction of pterygoid plate movement from youngest to oldest individuals. (B) Light blue triangle, anterior temporalis attachment; magenta triangles, lateral pterygoid attachment; gray arrows, direction of premaxillary change from youngest to oldest adult female baboons; colored arrows, primary direction of posterior face movement from youngest to oldest individuals and of muscle force for deep masseter (orange), anterior temporalis (light blue), and superficial masseter and medial pterygoid (yellow). Note that the zygomatic arch has been cut away to enhance visualization. (C) Blue diamond, supraorbital ridge, and glabella region; white triangle, posterior temporalis attachment. (D) Yellow triangles, superficial masseter attachment; arrows, concavity of the nasal bridge (light green) and midface (black) in the oldest individuals. (E) Gold line segment, cranial vault thickness in youngest individual; other line segments, old individual's cranial vault thickness when it is greater (purple) or less (lavender) than that of the youngest individual. (F) Open circles, anterior‐most extent of the malars; line segments, I1‐M3 length; gold, young individual; purple, old individual.

Potential role for extended ontogenetic trajectories

Previous research in baboons has suggested that skeletal growth, including upper facial height and cranial length, generally ceases in females by the age of 5 years (Leigh, 2009), around the time when sexual maturity is achieved (Altmann et al. 1981). Our data provide an opportunity to test this hypothesis more thoroughly, as well as to examine potential changes in shape that may occur in the CFS of female baboons after this age. We found that female baboon cranial shape changes and its size increases between young adulthood (ages 7–16 years) and older age (>18 years).

Our results raise interesting questions about the potential for various intrinsic and extrinsic factors to pattern variation in the post‐maturational CFS. Due to large canine size, especially in males, canine development is relatively extended in baboons. Research on mandibular teeth in captive baboons by Swindler & Meekins (1991) demonstrates that time from initial crown calcification to apical closure in canines is ~ 5.5 years, whereas for all other tooth types it ranges from 2.4 to 3.7 years. However, despite this prolonged canine development, the size and shape changes we have observed in the YA female face cannot be linked to anterior dental development because maxillary canine length has reached its adult size by 5 years of age in captive females (Table 5), and apical closure, at least of the mandibular canine, occurs by 6.3 years of age (Swindler & Meekins, 1991). Although age at apical closure for the maxillary dentition is unknown, age at eruption for the maxillary (4.1 years) and mandibular (3.9 years) canines differs by very little (Phillips‐Conroy & Jolly, 1988), so it stands to reason that apical closure of the maxillary canine is also not more than a few months later than that of the mandibular canine.

Table 5.

Female baboon canine growth

Age (years)	n	Length (SD)	Age (years)	n	Length (SD)
3	2	2.05 (2.9)	10	4	9.65 (1.9)
4	21	8.30 (2.4)	11	5	11.94 (1.4)
5	28	10.37 (1.5)	12	2	9.10 (1.3)
6	17	10.25 (0.7)	16	2	8.80 (1.4)
7	8	10.73 (0.6)	19	2	9.60 (0.1)
8	5	11.65 (1.2)	20	2	12.70 (0.7)
9	3	10.72 (1.6)

Maximum buccal length (mm) of the maxillary canines as measured from canine tip to gum line on anesthetized female baboons, from Leigh (2009).

Size and shape change of the CFS of young adult female baboons also cannot be attributed to the appearance of the last tooth to erupt, the M3, which occurs at ages 6.2 (mandibular) and 7.1 (maxillary) years in captive females (Phillips‐Conroy & Jolly, 1988), because eruption proceeds in a very rapid manner. In captive males, the mandibular M3 erupts at 6.3 years and apical closure is achieved by 6.5 years (Swindler & Meekins, 1991). Although this M3 completion time is only for mandibular teeth in males, there is no evidence to suggest that estimates of the length of time between eruption and apical closure in maxillary teeth differs from that of mandibular teeth. Additionally, evidence from humans, macaques, and baboons demonstrates that the general trend is for teeth to develop slightly earlier in males but more rapidly in females, such that female dental development is actually completed earlier than that of males (e.g. Garn et al. 1958; Swindler et al. 1982; Swindler & Meekins, 1991). Finally, although Swindler & Meekins (1991) did not have enough individuals to provide estimates of variation in age at apical closure for the mandibular M3, the range of standard deviation estimates for the other tooth types is 0.13–0.54 years. Therefore, all forces that may be exerted by dental eruption on the CFS cease before individuals reach ~ 7.5 years old. Yet, the female baboon CFS increases in size by 2.4% and changes shape between the ages of 7 and 16.

The fact that the greatest differences are between the YA group and all older baboons lends indirect support to the idea that growth in the CFS may not cease upon achieving sexual maturity. It is interesting to note that the late‐stage ontogenetic patterns in Pan cranial development identified by Mitteroecker et al. (2004b) (to include a superior shift of the nuchal plane, an elongation of the anterior midface, and an enlargement of the torus supraorbitalis) appear remarkably similar to the differences between YA and older baboons that we have observed. Furthermore, Balolia et al. (2013) demonstrate continued growth of the CFS in young adults of multiple species of non‐human primates, which they interpret to be an extension of ontogenetic trajectories into adulthood. However, without a complementary ontogenetic sample in which to measure growth vectors to determine whether the shape changes prior to achieving sexual maturity parallel those following it, we cannot definitively determine whether the patterns we have observed are the result of ontogeny extending into ‘adulthood’.

On the other hand, if future research demonstrates that the developmental trajectories of ontogenetic and adult samples of baboons are divergent, it is less likely that changes in the adult CFS are the result of either intrinsic growth mechanisms or the influence of adaptive growth centers that accommodate change among adjacent cranial components during ontogeny. Rather, it is more likely that factors that respond to extrinsic environmental cues, such as ambient temperature, stress and strain from muscular activity, and idiosyncratic behaviors, may play more pivotal roles in structuring adult craniofacial variation. It is also possible that, as has been shown in human research (e.g. Crothers & Sandham, 1993; Johansson et al. 1993), both dental wear and intra‐population variation in dental wear could contribute to observable differences in the CFS of adult baboons. However, it may be difficult to disentangle the effects related to dental wear and those related to increased usage of masticatory musculature, which also has been shown to pattern CFS shape and is often the cause of excessive dental wear (e.g. Krogstad & Dahl, 1985; Waltimo et al. 1994; Kiliaridis, 1995; Kiliaridis et al. 1995).

The effects of mastication on the aging baboon cranium

Oyen et al. (1979) developed a model for Papio brow ridge formation positing that cyclical deposition of supraorbital bone as a response to recurrently increased masticatory stress produces the distinctive torus seen in adult baboons, particularly males. With the eruption of each new permanent post‐canine tooth, the maxilla is repeatedly displaced anteriorly (Oyen, 1984). This lengthens the load arm of the TMJ, thus theoretically decreasing the biomechanical efficiency of the entire masticatory apparatus. However, subsequent research demonstrates that very little strain is actually experienced in the circumorbital region (e.g. Hylander et al. 1991; Ross & Metzger, 2004), although Dechow et al. (2010) did observe thinner cortical bone with reduced directional orientation across the cranium and including the brow ridge of edentulous individuals, suggesting that masticatory forces do play a role in maintaining CFS bone mass. Dechow & Carlson (1990) directly tested Oyen's model by conducting biomechanical and comparative anatomic analyses in immature and adult macaques, phylogenetic cousins of baboons. They were able to demonstrate that the mechanical advantage of the masticatory muscles considered (masseter and temporalis) actually increased 7–14% for molar biting between juvenile and sexually mature individuals. As their muscle force data suggest that adult macaques actually have more efficiently placed masseters, they argue that the biomechanial model of Oyen (1984) is invalid. Dechow & Carlson (1990) suggest that their unexpected results are due either to a disproportionate growth in the anterior portion of the muscle or to a mechanically advantageous shift in how the teeth are positioned relative to muscle attachment sites.

We observed both an anterior displacement of the teeth relative to the zygomaxillary suture (the anterior‐most extent of the superficial masseter origin) and a posterior displacement of the zygomatic bone itself with little relative movement of the glenoid fossa (see Fig. 5F). According to the biomechanical model presented by Dechow & Carlson (1990, Fig. 2), these changes in facial architecture as adults age should have an effect on masticatory efficiency. The result would be an increased molar biting moment (load) arm and a decreased masseter muscle moment (lever) arm. As mechanical advantage of a muscle is defined as the ratio between lever and load arms, the change in facial configuration from the YA to older age categories should have the effect of decreasing superficial masseter efficiency, as suggested by Oyen (1984). An additional issue to be considered is that increasing tooth wear and the potential for changes in mandibular ramus height as individuals age can alter masseter orientation, thus affecting its action and potentially its efficiency. Although anatomic arrangement of macaque and baboon masticatory musculature is incredibly similar (J. L. Joganic, pers. obs.; Schwartz & Huelke, 1963), the baboon facial skeleton is much more prognathic than that of macaques, which may explain the discrepancy between the results of Oyen (1984) and those of Dechow & Carlson (1990).

Our morphological data suggest that, even though the age‐effects we have observed in adult female baboons are small in magnitude, the associated CFS size and shape changes occur in regions with the potential to impact masticatory behavior. However, biomechanical analyses on the baboon masticatory system need to be conducted to test whether muscular advantage changes over the course of both adolescent growth and post‐maturational size and shape change. If future research is able to demonstrate that even subtle morphological change in adult craniofacial structure, such as the shape variation we have identified in our data, potentially alters masticatory efficiency, then the resulting modification to biomechanical loading regimes could induce additional changes in the CFS, as Oyen et al. (1979) suggest (see also Kiliaridis, 1995). In fact, the subtle decrease in orbital circumference that is visible along the superior orbital margin may be a manifestation of changes in the masticatory complex. Alternatively, it could be related to the fact that the frontal sinus enlarges with age, which may influence age‐related changes in the shape of the orbits (Fatu et al. 2006). Much more work needs to be done to determine whether such intraindividual variation has the potential to affect evolutionary fitness over the course of an individual's lifetime.

Finally, modifications in craniofacial configuration are not only observed in relation to morphological features associated with mastication. The other main feature that distinguishes YA and older crania is the increasing prominence of the inionic region as individuals age (see Fig. 5A). Interestingly, despite having absolutely larger muscles, large individuals do not have an appreciably more developed inion than their smaller conspecifics. This indicates that allometric shape differences may help larger animals withstand the increased force generated by larger muscles, but not necessarily their repetitive actions. As an indirect method of testing this hypothesis, we compared the mean difference in shape between five each of the smallest YA and older individuals and between five each of the largest YA and older individuals. Shape was quantified as an individual's regression score from the multivariate regression of PSCs on age. Although the larger individuals are located in a different area of the morphospace (average regression scores for small: ${\overline{X}}_{\mathit{YA}}=-0.0248,{\overline{X}}_{\mathrm{pooled}}=-0.0065$; large: ${\overline{X}}_{\mathit{YA}}=-0.0066,{\overline{X}}_{\mathrm{pooled}}=0.0130$; Fig. S5), the amount of shape difference between the younger and older individuals (small: 0.0183, large: 0.0196) is essentially the same in both size categories.

These results, coupled with the observed pattern of shape change in the adult CFS, suggest that bone remodeling in response to cyclical bone strain from repetitive muscle use over the course of an individual's life is potentially a major factor in patterning age‐related changes in craniofacial form in adult female baboons, regardless of the size of either the individual or their masticatory muscles. Given that age‐related shape change appears to be concentrated at muscle attachment sites, including the nuchal region, we propose that future research should aim to further our knowledge of cranial‐postcranial integration. Vertebral anatomy and its relationship to basicranial anatomy and posture in primates is the subject of some research (e.g. Russo & Kirk, 2013; Nalley & Grider‐Potter, 2015; Villamil, 2018), but our results demonstrate that its effect on cranial vault anatomy via the link with the nuchal musculature should be assessed as well.

Conclusion

We have demonstrated that the size and shape of the CFS in female baboons change moderately past the age of sexual maturity and into early adulthood. Contrary to expectations based on previous knowledge about the effects of tooth loss and accompanying alveolar resorption, it is actually the YA category that differs most from the other age groups. Younger and older adults differ most in regions related to muscle attachment sites, suggesting a possible role for repetitive masticatory stress in patterning bone remodeling in the adult CFS. The facial skeleton experiences the greatest amount of age‐ and size‐related change. A complementary and detailed geometric morphometric study of post‐maturational modifications in the human CFS could provide insight into mechanisms of aging that are shared or differ across primate taxa.

Conflict of interest

Neither author has a conflict of interest to declare.

Author contributions

Jessica Joganic: project conception and design, data acquisition and analysis, results interpretation, manuscript drafting and revision. Yann Heuzé: project conception and design, results interpretation, manuscript revision.

Supporting information

Table S1. Sample composition.

Table S2. Region‐specific Procrustes distance matrices.

Fig. S1. Primary shape regression.

Fig. S2. Between‐group principal components analysis.

Fig. S3. Effects of tooth loss on CFS.

Fig. S4. Regional variation in centroid size increase.

Fig. S5. Age‐related trends in small vs. large individuals.

Videos S1–S8. Eight videos that demonstrate the regional shape changes associated with allometric and age‐related variation in the CFS. The video begins with the surface mesh of either a small (Videos S1–S4) or a young (Videos S5–S8) individual and by the end of the video it is morphed into a large or old individual. There are right lateral (Videos S1 and S5), vertex (Videos S2 and S6), basilar (Videos S3 and S7), and frontal (Videos S4 and S8) views provided for each set of surface morphs.