Sex Estimation from the Skull and Mandible

Human cranial sexual dimorphism is expressed primarily in two clusters of traits. The first is overall robusticity: male skulls are on average larger, thicker, and more heavily muscled than female skulls, with more prominent surface features at the sites of major muscle attachments. The second is shape: males tend toward more prominent brow ridges, larger mastoid processes, more pronounced nuchal cresting, and a more squared and robust mandible.

The robusticity differences reflect average body-size differences between sexes, which in humans are moderate but not extreme compared with other primates. Male humans are on average 5 to 8 per cent taller than females, and skeletal robusticity scales with body size plus additional hormonal effects. The shape differences at specific anatomical sites reflect the larger temporalis and masseter muscles in males (driving greater temporal lines, larger zygomatic arches, and a more flared gonion), larger sternocleidomastoid and trapezius muscles (driving larger mastoid process and more prominent nuchal crest), and the differential hormonal effect on the supra-orbital region at puberty.

The diagnostic problem is that these differences overlap substantially between sexes, especially in the intermediate range. A female with a relatively robust skull and a male with a relatively gracile skull may both score in the ambiguous zone on any scoring system. Population differences in average skeletal robusticity add a further complication: a female from a population with generally robust cranial morphology may score higher on male-associated traits than a male from a population with gracile cranial morphology. This is not a failure of the underlying biology; it reflects the fact that cranial sex estimation, unlike pelvic sex estimation, is doing more work in a smaller signal space.

The practical consequence is that the osteologist should report skull-based sex as "most consistent with male/female, with moderate confidence" when the full suite of traits points consistently in one direction, and as "indeterminate sex, with morphology slightly trending toward male/female" when the traits are mixed. Overclaiming certainty for cranial sex estimation is a well-documented problem in published forensic casework reviews, and it is specifically flagged in the SWGANTH (now OSAC Forensic Anthropology Subcommittee) best-practice documents as a category of methodological error.

Philip Walker's 2008 paper in the American Journal of Physical Anthropology described an ordinal scoring system for five cranial traits, validated on four skeletal collections: the Terry Collection (US, Black and White Americans), the Bass Donated Collection (US), the Spitalfields Collection (UK, 18th-century English), and a Portuguese collection. The multi-population design aimed to produce a method robust across more than one reference population, though the geographic range remains predominantly Western.

The five traits and their scoring are as follows.

Mastoid process. Scored 1 to 5, where 1 is very small (most female) and 5 is very large (most male). The mastoid process is the conical bony projection inferior to the external auditory meatus on the temporal bone. It is the attachment site for the sternocleidomastoid, splenius capitis, and longissimus capitis muscles, all of which are larger in males. The mastoid process is visible on most recovered skull fragments that include the petrous or squamous temporal bone.

Nuchal crest. Scored 1 to 5, where 1 is smooth (most female) and 5 is pronounced (most male). The nuchal crest (also called the superior nuchal line or the external occipital protuberance complex depending on which part is being scored) is the roughened area on the posterior occiput at the attachment of the nuchal muscles (trapezius, semispinalis capitis, splenius capitis). Males with heavier neck musculature develop more prominent nuchal cresting.

Supra-orbital margin. Scored 1 to 5, where 1 is sharp and thin (most female) and 5 is rounded and thick (most male). The supra-orbital margin is the bony rim forming the superior border of the orbit. Female supra-orbital margins are typically thin and relatively sharp to the touch; male supra-orbital margins are thicker, more rounded, and merging into the frontal bone with a more pronounced brow shelf. This trait is visible on virtually any frontal bone fragment that includes the upper orbital rim.

Glabella. Scored 1 to 5, where 1 is flat (most female) and 5 is pronounced (most male). The glabella is the midline prominence on the frontal bone between the two superciliary arches. In males with heavy brow ridges, the glabella is markedly prominent; in females and gracile males, it is flat or barely raised. This trait is correlated with the supra-orbital margin score.

Mental eminence. Scored 1 to 5, where 1 is slight (most female) and 5 is pronounced (most male). The mental eminence is the bony chin prominence of the mandible. Human males characteristically develop a more squared and prominent mental eminence, while female mandibles tend toward a more pointed or triangular chin profile. This is one of the traits used on the mandible alone when the cranial vault is absent.

Walker reported accuracy of 84 per cent for females and 88 per cent for males when all five traits were combined using a discriminant function across his pooled multi-population sample. Within individual population subsamples, accuracy ranged from approximately 80 to 91 per cent. He also noted that the traits with the highest individual predictive power were the mastoid process and glabella, followed by the supra-orbital margin, with the nuchal crest and mental eminence contributing less individual information.

William Krogman's foundational 1962 text 'The Human Skeleton in Forensic Medicine' and its 1986 revision co-authored with Mehmet Iscan established the descriptive language for cranial sex estimation that shaped forensic anthropology practice for a generation. The framework evaluated the same anatomical regions as Walker but used verbal descriptions rather than ordinal scales and did not report accuracy statistics from a formal validation sample.

The Krogman-Iscan descriptors remain influential because they are comprehensive, covering traits beyond Walker's five: the zygomatic arch (more robust in males), the mastoid size, the post-orbital constriction (more pronounced in males), the occipital condyle size (larger in males), the palate size (larger in males), the cranial capacity (larger in males by approximately 10 per cent in most samples), and the frontal bone profile (more sloping in males). The text provided the field with a memorable summary: "the male skull is large, rugged, massive; the female skull is small, gracile, rounded."

The limitation of the Krogman-Iscan framework is that it was entirely qualitative and lacked documented accuracy statistics. Inter-observer reliability studies conducted in the 1990s and 2000s showed that when observers used only verbal descriptions without systematic scoring scales, inter-observer agreement on individual traits varied from approximately 70 to 85 per cent, and overall sex determination agreement was approximately 75 to 80 per cent. The Walker 2008 ordinal system was designed in part to improve inter-observer reliability by replacing verbal descriptions with numbered scales, which at least constrain the observer's response even if they do not eliminate subjectivity.

In current practice, the Krogman-Iscan descriptors serve as a checklist for initial triage (is this skull clearly male, clearly female, or ambiguous?), and the Walker ordinal scoring is applied as the reportable method. Many experienced forensic anthropologists use both: the Krogman-Iscan descriptors to orient the assessment and the Walker scores to generate the reported figure.

The mandible is the most robust facial bone and often survives as an isolated element in fragmented or scattered remains. Its sex-discriminating traits include the mental eminence (shared with the cranial scoring above), the gonial angle, the overall size and robusticity of the body, and the chin shape.

The gonial angle. The gonion is the posterior inferior angle of the mandible, where the ramus meets the body. In males, the gonion is typically everted outward (flared laterally) and the angle between the ramus and the body is relatively acute, producing a more squared mandibular angle when viewed posteriorly. In females, the gonion is less everted, and the angle is more obtuse, producing a more rounded appearance at the posterior mandible. Metric measurement of the gonial angle can be performed with a protractor or callipers on a standardised orientation. In published samples, the male gonial angle means are approximately 110 to 120 degrees and female means approximately 120 to 130 degrees, but there is substantial overlap and population variation.

Chin shape. The shape of the mental eminence viewed from the front is broadly sexual in most human populations: males tend toward a squared, broad chin with the mental eminence occupying a wide area; females tend toward a more pointed or triangular chin with a narrower mental eminence. This distinction is visible on a dry mandible and is often noted even by non-specialists. It carries approximately 70 to 75 per cent accuracy as a standalone indicator in the US and European reference samples.

Mandibular body robusticity. The corpus height (measurement from the inferior border to the alveolar margin at the mental foramen level) and the corpus breadth are both sexually dimorphic. Mandibular metric discriminant functions for sex estimation, using corpus height and breadth, have been published for US samples (Giles 1964, corrected sex-specific means), South African samples (Steyn and Iscan 1998), and Indian samples (Introna et al. 1993 with Indian sub-populations; Saini et al. 2011 for North Indian mandibles). The Indian studies show that mandibular corpus height and breadth in North Indian samples follow the same direction of sexual dimorphism as US samples (males larger) but with somewhat smaller absolute dimensions on average, which shifts the optimal discriminant cut-off.

The mandible alone achieves approximately 70 to 80 per cent correct sex classification using morphological methods and approximately 78 to 85 per cent using metric discriminant functions in well-validated samples. When the mandible is recovered with the rest of the skull, it adds incremental information to the cranial scoring. When the mandible is the only element available, sex estimation should be reported with a wide uncertainty range and explicit reference to the population-specific reference used. The same cranial measurements used for mandibular sex estimation feed into FORDISC-based ancestry assessment as part of the broader biological profile.

The FORDISC discriminant function software, developed by Stephen Ousley and Richard Jantz at the University of Tennessee Knoxville, is the most widely used craniometric sex and ancestry estimation tool in English-speaking forensic anthropology. Its current version (FORDISC 3.1) applies discriminant functions derived from the Forensic Databank, a large collection of documented US skeletal cases assembled from medical examiner offices across the United States, combined with elements of the Howells craniometric database (a global reference set of 2,524 crania from 28 worldwide population groups).

For sex estimation, FORDISC uses a subset of craniometric measurements standardised by the Osteometric Data Collection Form protocols. The discriminant function for sex, derived from the Forensic Databank, has reported accuracy of approximately 88 to 92 per cent in cross-validation on the reference sample. The software outputs a posterior probability of male or female for the submitted measurement set, together with typicality probabilities that indicate whether the skull is morphologically similar to the reference population or whether it is an outlier.

Matthew Spradley and Richard Jantz (2011) compared sex classification accuracy of cranial and postcranial skeletal elements using the Forensic Databank, finding that most postcranial elements outperform the cranium for sex estimation, with multivariate postcranial models reaching up to 94% accuracy while the best cranial models did not exceed 90%. However, multiple independent studies have documented that FORDISC's sex discriminant function, trained predominantly on US populations, shows reduced accuracy when applied to skulls from South Asian, East Asian, and sub-Saharan African populations that are morphologically more distant from the reference sample.

Bhasin (2011) applied FORDISC sex estimation to a sample of 120 Indian crania (60 male, 60 female) from AIIMS and found correct classification of approximately 79 per cent, a reduction of approximately 10 percentage points from the reported US accuracy. Hennessy and Stringer (2002) found similar accuracy reduction for East African Homo sapiens skulls. These findings do not invalidate FORDISC as a tool; they establish that its outputs must be interpreted with explicit knowledge of the reference population composition and the likely morphological distance between the skull under analysis and the training sample.

South Asian reference data. Multiple Indian studies have published population-specific data for cranial sex estimation. Nath (2006) examined morphological and metric cranial variables in a Delhi sample, finding that the mastoid process and supra-orbital margin showed the greatest discriminatory power and that discriminant functions derived specifically from the Indian sample outperformed FORDISC on the same specimens. Saini et al. (2011) measured mandibular ramus dimensions in 116 North Indian individuals and derived sex-specific discriminant functions for mandibular sex estimation. The AIIMS Department of Forensic Medicine has published multiple studies on cranial morphometrics for Indian populations over the 1990s to 2010s, representing the most consistent Indian reference database for this purpose.

The Aarushi-Hemraj double homicide (Noida, 2008) involved skeletal assessment of recovered remains, including cranial evaluation alongside other identification methods. The case has been cited in Indian forensic medicine literature as an example of multi-element biological profile methodology in a criminal investigation.

East African reference data. Hennessy and Stringer (2002) examined craniometric sex differences in sub-Saharan African samples and found that the general cranial sexual dimorphism pattern held across the populations they studied, but that population-specific discriminant coefficients were necessary for reliable classification. The Pretoria Reference Collection and the Raymond Dart Collection at Witwatersrand continue to serve as the primary southern African skeletal references for cranial sex estimation in South African casework. The South African Police Service Forensic Science Laboratory (SAPS FSL) uses population-specific reference data from these collections for cranial biological profile assessments.

East Asian and South-East Asian reference data. Japanese, Chinese, Korean, and Thai skeletal reference data exist in the published literature (Iscan and Shihai 1995 for Chinese, Kanchan et al. 2008 for South-East Asian immigrant populations), and the Forensic Anthropology Society of Europe (FASE) has published a compendium of European population-specific skeletal reference data. These studies confirm that the direction of cranial sexual dimorphism is consistent across populations (males larger and more robust), but that population-specific discriminant coefficients improve accuracy significantly compared with direct application of US-trained FORDISC functions.

Practical reporting implication. When an osteologist applies cranial sex estimation in a casework context, the forensic report must state: (1) the method used and its publication reference; (2) the reference population from which the accuracy figure is derived; (3) whether the population of the unknown individual is likely similar to or different from the reference population; and (4) the resulting uncertainty adjustment to the reported accuracy, if any. This is the standard required by the ABFA, SWGANTH/OSAC, and the ENFSI Forensic Anthropology Working Group's best-practice guidelines.

1. Cranial inventory and condition
   Document which cranial elements are present and their preservation state. Note which of the five Walker traits are assessable. Record any postmortem damage or taphonomic alteration to the scored regions (erosion of nuchal crest, damage to supraorbital margin, missing temporal for mastoid assessment).
2. Walker 2008 five-trait ordinal scoring
   Score each of the five traits (mastoid process, nuchal crest, supra-orbital margin, glabella, mental eminence) on the 1-5 scale. Score only the traits for which the morphology is clearly visible. Apply the discriminant function or consult published posterior probability tables. Record the population reference used and its documented accuracy.
3. Mandibular assessment (if available)
   Score mental eminence (already in Walker five), gonial angle, chin shape. Measure corpus height and breadth at the mental foramen if intact. Apply population-specific discriminant function for mandibular metrics if Indian, South African, or other non-US population suspected.
4. Craniometric measurements (if precision tools available)
   Collect the standardised FORDISC measurement suite. Run FORDISC 3.1. Note the posterior probability and typicality probability. If typicality is low (p < 0.05), the skull is an outlier from the reference population, and the sex posterior probability is unreliable. Apply population-specific adjustment if available.
5. Population-specific calibration check
   Identify the most likely population affinity of the remains from context, associated material, and preliminary morphological assessment. Compare this with the reference population of each method applied. If there is a known or likely mismatch, note the expected accuracy reduction and adjust confidence level.
6. Composite reporting
   Combine results from all applied methods. When results are consistent: report with the appropriately calibrated confidence for the matched population. When results conflict: report each method's result, explain the source of conflict, and give a composite conclusion with wider uncertainty range. Never report cranial sex estimation at higher confidence than the best-validated method warrants for the relevant population.