From a morphological perspective, ACD can be seen as the outcome of a real-life experiment, albeit unplanned, on altering growth trajectories (Cheverud et al., 1992, Jimenez et al., 2012 and Liebermann et al., 2000). As such, it has been used to investigate patterns of craniofacial variation in humans by looking at secondary effects of vault modifications induced on other structures, namely the face and the base. Previous studies have shown that, due to the morphological integration of the skull, deforming the braincase during extended periods of growth leads to changes in the cranial base and face. However, as it has been pointed out (Antón, 1989 and Friess and Baylac, 2003), findings differ between studies, leaving some doubt as to the exact nature and patterns of some of these changes, and possibly reflecting uncertainties in the identification of deformation types. While the number and nomenclature of ACD vary according to different authors, a major distinction found in many publications is made between annular (or circumferential, C) and tabular (or anteroposterior, AP) types. This binary system is thought to reflect the two main types of deformation devices: soft wrapping vs. hard materials (tablets made from wood or stone). The use of these devices also tends to result in fundamentally different cranial shapes: C-type deformations lead to postero-superiorly protruding and mediolaterally narrow vaults, whereas the AP-type is characterized by anteroposteriorly shortened and mediolaterally widened braincases (Antón, 1989, Dembo and Imbelloni, 1938 and Falkenburger, 1938). Within AP deformations, a subdivision into erect and oblique forms, though not supported by all scholars, has been proposed on the basis of the vertical/inclined orientation of the occipital squama (e.g., Dembo and Imbelloni, 1938).

In addition to the general change in neurocranial shape, various associated modifications in the face and base have been reported, though not all studies agree on all of these. Thus, several authors found facial broadening in AP-type deformations vs. narrowing in C-type deformations (Antón, 1989, Cheverud et al., 1992, Jimenez et al., 2012 and Kohn et al., 1993). Increased facial height dimensions have also been reported in AP deformations (Antón, 1989), while some studies observed different degrees of facial shortening and flattening in AP (Cheverud et al., 1992 and Jimenez et al., 2012).

With respect to the cranial base, the effects of the various deformation types have been estimated inconsistently: several studies (Antón, 1989 and McNeill and Newton, 1965) found flattened basicrania (platybasia) in both AP- and C-type deformations, while others observed platybasia in C-, and basicranial kyphosis in AP-types (Martínez-Abadías et al., 2009 and Moss, 1958). Furthermore, support for platybasia in AP also comes from Oetteking (1924), while Schendel et al. (1980) found no effect on the cranial base angle (CBA) at all. Conversely, the overall proportions of the cranial base, when investigated, have been consistently found altered depending on the deformation type: AP deformations result in anteroposteriorly shortened and mediolaterally widened basicrania, while C deformations yield the opposite shape, i.e., an anteroposteriorly lengthened and mediolaterally narrowed base (Antón, 1989, Cheverud et al., 1992, Cocilovo et al., 2011, Kohn et al., 1993 and Martínez-Abadías et al., 2009).

Regarding the inconsistent findings with respect to CBA, previous studies have identified as potential causes differences in: (1) sample size, varying between N = 65 (Moss, 1958) and N = 1586 (Cocilovo et al., 2011); (2) classification of deformation types; (3) methodology (e.g., external vs. radiographic-based measurements, different definitions of CBA; see Antón, 1989 and Cocilovo et al., 2011). However, differential effects on the base by deformation type have also been observed in studies with comparatively large sample sizes (e.g., Baylac and Friess, 2005 and Martínez-Abadías et al., 2009) and, according to Antón (1989), differences in CBA definition do not account for these inconsistencies either. External vs. internal measurements (radiography, computed tomography) may also represent a potential cause for differences in the assessment of CBA.

The purpose of the present contribution is to compare the external and internal morphologies of deformed and undeformed skulls based on data obtained from standard medical imaging devices. This approach allowed us to address the following aspects quantitatively: 1) the effects of ACD on the skull base, specifically, the overall shape of the base as well as its angular variations; 2) the effects of ACD on the masticatory apparatus, specifically, the spatial relationship between the upper jaw and the mandibular fossa of the temporomandibular joint (TMJ).

The sample used in this study consists of 51 specimens from the anthropological collections of the “Muséum national d’histoire naturelle” (MNHN), Paris. Based on the fusion of the spheno-occipital synchondrosis and third molar (M3) eruption stage, they are all adult except two adolescent individuals. Due to incomplete information, sex was not used in subsequent analyses. The sample was recruited from two major geographic regions: South America and Europe. The South-American sample includes 39 individuals from Bolivia and has been collected by Georges de Créqui Monfort and Eugène Sénéchal de La Grange at the beginning of the 20th century (Chervin, 1908). It is mostly composed of two ensembles, one of which is an Aymara sample from Tiwanaku (or Tiahuanaco, Lake Titicaca), the other one being largely from Asnapujio in the South of the country (Chervin, 1908 and Falkenburger, 1938). Based on the collectors’ reports (Chervin, 1908, Créqui Monfort and Sénéchal de La Grange, 1904 and Falkenburger, 1938), the Bolivian sample was generically attributable to the pre-Hispanic period. In the case of Tiwanaku, the remains are likely to date between the 4th and 11th centuries CE (Janusek, 2008), though a more refined chronometric dating has yet to be established.

The other geographic group consists of 12 French individuals, predominantly from the Southwest region: seven deformed skulls from the 19th century and five undeformed skulls from megalithic burials in Sauveterre (Manouvrier and De Mortillet, 1893).

The Bolivian skulls exhibit both anteroposterior and circumferential deformations, while the French skulls show a circumferential deformation known as the “Toulouse-type” (Broca, 1871 and Dingwall, 1931). Toulouse-type deformations result from prolonged wearing of tightly wrapped head scarves known as “bandeau”, and reports attest to the survival of this practice until the end of the 19th century (Broca, 1871). While it remains unclear to which extent head shape alteration was actually sought or merely an accepted side effect, both the resulting head shape and the mechanics used to achieve it match those of circumferential deformations known from other parts of the world, and thus warrant their inclusion in studies on the craniofacial effects of ACD. Still, because they stem from a different geographic and cultural context, for the purposes of this study they were considered as a separate group from the South-American C-type.

Undeformed skulls of both geographic regions were also included in the analyses. Identification of presence/absence and type of deformation were performed by one of us (MF). Given the difficulty to identify artificial deformations in cases of minimal alteration, specimen selection targeted clearly deformed skulls and presumably undeformed ones. Consequently, there was a higher chance for a minimally deformed skull to be misidentified as undeformed than vice versa. The assignment to C and AP-type, on the other hand, was further backed up by checking the cranial index (breadth/length), which is significantly higher in AP than in C deformations. Table 1 provides a summary of the sample composition.

All crania were scanned with a medical CT scanner using specific parameters for dry bone (Badawi-Fayad et al., 2005). The final voxel size was 0.45 mm  ×  0.45 mm  ×  0.4 mm. Each skull was segmented using a modified half-maximum height protocol (Fajardo et al., 2002 and Spoor et al., 1993), and 3D landmark coordinates were extracted using Avizo 8.1 (FEI Inc, Hilsboro, OR). Repeatability was assessed by computing mean delta and standard deviations of a series of repeated landmark measurements and found to be below 1 mm, thus complying with osteometric standards (Bräuer, 1988 and Buikstra and Ubelaker, 1994). The mandible, while highly informative for overall craniofacial and masticatory morphology, in particular, was not included in this study due to incomplete records and the uncertainty in achieving exact occlusion in the absence in several specimens of condyles and/or lower teeth.

Missing landmarks were estimated through thin-plate spline warping using the Morpho library for R (version 2.3.1.1; Schlager, 2016). The number of missing landmarks was below the recommended threshold of 20% (Couette and White, 2010) except for one case. Landmarks were then normalized via Generalized Procrustes Analysis (GPA; Rohlf, 2000) and subsequently submitted to multivariate statistics. Group-differences were tested with a between-group principal components analysis (PCA) and patterns of covariation between different cranial elements were investigated with two-block partial least-squares analysis (2B PLS; (Rohlf and Corti, 2000). Between-group PCA was based on the total set of landmarks, while 2B PLS used subsets of landmarks based on their assignment to the vault, cranial base, and masticatory apparatus.

Table 2 lists the landmarks and their assignment to cranial modules (Fig. 1). The choice of landmarks was mostly driven by our purposes, i.e., the need to capture the overall neuro- and basicranial shape, as well as the relative position of masticatory elements, such as the mandibular fossa, the alveolar plane, and the origin of the m. masseter. The use of extremal landmarks, such as euryon and opisthocranion, to capture the chief dimensions of the vault, was abandoned due to poor repeatability.

The assignment to different cranial modules largely followed Bastir and Rosas (2005), Bookstein et al. (2003), Martínez-Abadías et al. (2009), and von Cramon-Taubadel (2011), and aimed at obtaining a comparable number of landmarks per module. However, in some cases, assignments vary by author, as it is the case for landmark selection. Notably, von Cramon-Taubadel (2011) defines the ‘chondrocranium’ as a module, rather than the base, and most landmarks assigned here to the base and included by von Cramon-Taubadel, such as opisthion, are part of her ‘chondrocranium’; conversely, porion and mastoideale are here assigned to the vault, following von Cramon-Taubadel (2011), whereas Martínez-Abadías et al. (2009) assign both to the base.

We then performed analyses of covariation between the vault/base and vault/masticatory apparatus using a single GPA for the whole sample (Klingenberg and Marugán-Lobón, 2013). We also performed covariation analyses for each a priori group separately (AP, C, T, undeformed) and compared the angle of the first PLS vectors in order to assess whether patterns of within-group covariation were similar or different between groups (Klingenberg and Marugán-Lobón, 2013 and Kligenberg and Zimmermann, 1992).

Shape changes associated with PLS axes were visualized by regressing Procrustes residuals onto axis scores and predicting coordinates for extreme values and group means. Shape differences between groups were visualized using mean group configurations. The CBA was defined as the angle formed by foramen caecum, sella and basion, and alternatively by nasion, sella, and basion. All statistics were computed in R (version 3.30, R Development Core Team, 2016) using dedicated libraries (Adams and Otarola-Castillo, 2013, Dryden, 2015 and Schlager, 2016), except for angular comparisons of PLS axes, which were performed in MorphoJ 1.06 (Klingenberg, 2011).

This study used 3D geometric morphometrics to investigate patterns of variation and covariation in artificially deformed skulls, particularly regarding the effects of the deformations on the base, including its masticatory elements (mandibular fossa). Previous work had mainly focused on examining how the major cranial elements develop in an integrated way under “normal” conditions (e.g., Ackermann, 2005, Bastir and Rosas, 2004, Mitteroecker and Bookstein, 2008 and von Cramon-Taubadel, 2011), while the specific context of cranial deformation has largely produced 2D studies based on radiographs (but see Khonsari et al., 2013). The use of medical imaging and an explicit analysis of covariance in the present study allowed us to gain insight into the plasticity in the craniofacial region when submitted to altered growth trajectories. While the sample size is still limited, making our conclusions to some degree preliminary until the findings are replicated on a larger sample, the results shed nonetheless further light on some secondary effects of cranial deformations. They confirm previous findings, according to which ACD are associated with changes in the skull base architecture, namely causing platybasia (Antón, 1989, McNeill and Newton, 1965 and Oetteking, 1924), but also inducing altered proportions of the base and face (Antón, 1989, Cheverud et al., 1992, Cocilovo et al., 2011, Kohn et al., 1993 and Martínez-Abadías et al., 2009). However, the notion that all deformed skulls exhibit basicranial flattening regardless of the deformation type is not corroborated by the present study. Rather, we find that AP deformed skulls, when grouped together, show a trend towards basicranial kyphosis, a finding first reported by Moss (1958; see also et al., 2009Martínez-Abadías). We also found evidence for a more complex pattern when a possible subdivision of AP skulls into so-called “erect” and “oblique” forms is taken into consideration. Thus, our findings support the idea that AP oblique skulls conform to the general pattern of platybasia, while AP erect skulls are relatively more kyphotic, and in this respect resemble more undeformed skulls. While requiring further corroboration on a larger sample, this finding is consistent with similar results obtained on external measurements (Baylac and Friess, 2005).

The use of 3D medical imaging allowed us to also suggest specific effects of artificial deformation not really appreciable in plain X-ray films or by 3D surface data. Specifically, while narrowing of the crania base in C deformations, or shortening in AP deformations, had previously been reported (e.g., Cheverud et al., 1992 and Kohn et al., 1993), our results show to which extent this affects the anterior portion. The same applies to the forward-migration of the mandibular fossa in relation to the anterior cranial base (Ferros et al., 2016), regardless of deformation type.

Previous studies of endocasts of deformed skulls have revealed modifications of craniovascular vessels, which have been interpreted as indicative of intracranial pressure changes (Dean, 1995 and O’Loughlin, 1996). CT scanning could prove useful for quantifying such alterations and improve our understanding of potential clinical consequences of artificial deformation, as they cannot be observed on live subjects (Tiesler, 2014).

Interestingly, devices resembling those used to obtain ACD are used in clinical practice in order to correct posterior positional plagiocephaly, a flattening of the posterior aspect of the skull secondary to prolonged supine sleep position (Xia et al., 2008). These helmets exert lateral pressure on the skull vault, as they are designed to fight against excessive anteroposterior flattening. The assessment of the influence, if any, of helmet therapy on basicranial flattening would be of tremendous interest in the discussion of the differential influence of AP and C deformations on craniofacial structures, as it would exhibit the effects of forces perpendicular to both usual ACD modalities.

Several other patterns of cranial alteration add further weight to the hypothesis that differences in deformation techniques lead to different responses of the growing skull: (1) the anterosuperior shift of the mandibular fossa, part of the temporomandibular joint; (2) the sagittal reorientation of the petrous region in AP skulls, particularly in the erect subtype; and (3) the more horizontally inclined foramen magnum in C deformations, in conjunction with the significant differences in the angles of the first PLS axes between groups, illustrate that the vault, base and masticatory portions of the skull function as integrated structures. These results stress the plasticity with which the human cranium reacts to mechanical constraints during growth along different trajectories and in direct response to the nature of the deformation device.

The modification in the position of the mandibular fossa in different ACDs is of particular importance for the understanding of induced growth alterations of the craniofacial region, as an anterior shift of the mandibular fossa is a known factor in the development of dento-skeletal anomalies where the lower jaw is shifted forward. These anomalies can lead to class III malocclusion, depending on the maxillary and dental compensations for the abnormal mandibular position (Delaire et al., 1981).

Present results thus indicate that abnormal constraints exerted on the skull vault can have an influence on the position on the mandible, and thus on occlusion, via modifications in the position of the mandibular fossa of the skull base. Nevertheless, several studies devoted to the structure of the mandible in ACDs indicate that vault deformations do not induce occlusal anomalies, despite significant modifications in the shape of the mandible (Cheverud and Midkiff, 1992, Ferros et al., 2016 and Jimenez et al., 2012). The absence of malocclusion in ACD most probably corresponds to compensatory mechanisms that could be better understood by 3D studies of the shape of the lower jaw, as currently available studies on the mandible have been performed using 2D data. Furthermore, studying occlusion in dry skulls relies on an approximate placement of the condylar head into the mandibular fossa, without taking into account the soft tissues forming the temporomandibular joint.

In sum, current data on the mandible in ACDs (Ferros et al., 2016) suggest that potential occlusal anomalies (class III malocclusion) – that could be predicted by the anterior displacement of the mandibular fossa – are compensated by mechanisms most probably involving 3D mandibular shape modifications. A simple way to progress in the understanding of these compensatory mechanisms would be to assess the 3D angulation of the occlusal plane, but this again relies on an appropriate positioning of the condylar heads relative to the skull base, and therefore renders the inclusion of the mandible a not so trivial issue when working with dry skulls.

As a whole, we provided support for differential effects of different types of ACD on craniofacial growth, and our results may be seen as further indication of the existence of multiple cranial modules, such as the petroso-mandibular unit (Bastir and Rosas, 2004, Bastir and Rosas, 2005 and Bastir et al., 2004), though testing modularity hypotheses is beyond the scope of this study and should be addressed in future research on ACD.

Our findings also lend support to the idea that the basic dichotomy of AP vs. C deformations does not well represent the shape variation (e.g., Manríquez et al., 2006), at least with respect to internal structures. Future work should focus on classifying ACD in a manner consistent with both ethnographic data on devices, as well as with the resulting internal and external shape (co)variation. Improved classifications may benefit from the use of semilandmarks (e.g., Perez, 2007). As indicated above, the cranial vault is not extensively covered by standard landmarks, so that it is conceivable that minute shape variations that may help in the classification process are not well accounted for here. After all, the tremendous amount of shape variation seen in deformed crania around the world is a direct reflection of the diversity of cultural practices responsible for the morphological variability, and any classification system is prone to reductionist fallacies.