Forensic anthropologists and physical anthropologists distinguish male from female skulls by examining a cluster of physical features that tend to differ between the sexes, including the brow ridge, jawline, chin shape, and the bony bump behind the ear called the mastoid process. No single feature is reliable on its own. It is the combination of traits, sometimes scored on a numerical scale and sometimes fed into statistical models, that allows a trained observer to estimate sex with accuracy rates that typically fall between about 77% and 90%, depending on the method and the population the skull comes from.
The Major Features That Differ
Male skulls are, on average, larger and more robust than female skulls. But raw size alone is a weak indicator because a small man and a large woman can overlap considerably. What matters more is the shape and prominence of specific bony landmarks. In women, the forehead tends to be more vertical and smooth, with a rounder, less pronounced brow ridge (the glabella) and a subtler rim above the eye sockets. In men, the brow ridge juts forward more, the forehead slopes back at a sharper angle, and the cheekbones are more pronounced. The male mandible is typically larger, more angular, and more square-cut, giving the lower face a broader, boxier look, while female faces tend toward a rounder or more heart-shaped outline.1PubMed Central. Aging and Sexual Differences of the Human Skull
Behind the ear, the mastoid process is one of the most commonly assessed features. In males it is generally larger and more projecting. The triangle of bone formed around it (the mastoid triangle) has been studied extensively, and all its measurements tend to run higher in males.2PubMed Central. Morphometric Study of the Mastoid Triangle for Sexual Dimorphism of Dry Skulls in the North Indian Population At the back of the skull, the nuchal crest, a ridge of bone where neck muscles attach, is typically more prominent in males.3PubMed Central. A new metric method for sex estimation using three-dimensional imaging of the nuchal crest The chin also tells a story: male chins tend to be squarer and more prominent, while female chins are often more pointed or narrow.
A large-scale mapping study confirmed that the glabella and brow ridge area, the mastoid process, and the nasal region are the most sexually dimorphic parts of the skull, achieving about 73% accuracy individually and 77% when evaluated together. Interestingly, the same study found that the bump at the base of the back of the skull (the external occipital protuberance) was not a reliable sex indicator at all; its size was more related to overall body size than to sex.4Scientific Reports. Mapping sexual dimorphism signal in the human cranium
How Experts Actually Score a Skull
When a forensic anthropologist examines a skull, they do not simply eyeball it and declare “male” or “female.” Most follow standardized scoring systems. The most widely used protocol, originally published by Buikstra and Ubelaker in 1994, asks the examiner to score five cranial traits on a scale from 1 (most feminine) to 5 (most masculine). These traits include the nuchal crest, the mastoid process, the sharpness of the upper eye-socket margin, the prominence of the brow ridge, and the shape of the chin.5PubMed Central. Evaluating macroscopic sex estimation methods using genetically sexed archaeological material Some European protocols use a similar scale running from hyperfeminine to hypermasculine.
The scores are then typically plugged into a statistical formula, often a discriminant function, that outputs a probability of the skull being male or female. A logistic regression model using all five standard cranial trait scores correctly classified about 88% of modern skulls in one large validation study, with almost no bias toward guessing one sex over the other.6PubMed. Sexing skulls using discriminant function analysis of visually assessed traits Other researchers have supplemented these visual scores with actual measurements using calipers or digital tools, pushing accuracy for some populations above 90%.
This visual scoring approach has its strengths and its limits. It is quick, requires no specialized equipment beyond the skull itself, and works reasonably well when applied to the population it was developed for. But subjectivity is baked in. Two trained observers looking at the same brow ridge might give it a 3 versus a 4 on the scale. That disagreement usually washes out when multiple traits are scored together, but it remains an acknowledged weakness of the method.7PubMed Central. Sexual Dimorphism of Cranial Morphological Traits in an Italian Sample
The Measurement Approach
Instead of, or alongside, visual scoring, some researchers take direct measurements of the skull with calipers or from CT scans and use those numbers to build population-specific equations. A study on 200 Thai skulls found that a formula using six measurements, including maximum cranial length, cheekbone width, ear-to-ear width, nasal height, eye-socket width, and mastoid length, correctly identified sex about 91% of the time.8PubMed Central. Craniometric study for sex determination in a Thai population A Brazilian study using a similar approach reported accuracy in the range of 82–90%.9Research, Society and Development. Sex estimation of Brazilian skulls using discriminant analysis of cranial measurements
These equations are impressive, but there is a catch: they are tuned to a specific population. A formula built on Thai skulls will not necessarily work well on skulls from East Africa or Northern Europe, because the degree of sex difference in skull shape varies across populations. This is one of the central complications in the field.
Why Population Matters So Much
Research has documented considerable variation in how different populations express skeletal sex differences. Both genetic background and environmental factors like nutrition and physical activity influence how robust or gracile a person’s skeleton becomes. A male skull from a population with relatively low sexual dimorphism might look similar to a female skull from a population with high dimorphism. If you apply standards developed on one group to another, accuracy can drop sharply.10PubMed. Population variation in skeletal sexual dimorphism
This is why forensic anthropologists emphasize using population-specific reference data whenever possible. In a real forensic case, the analyst ideally has some idea of the individual’s probable ancestry or geographic origin before applying a sex estimation method. Without that context, the margin of error widens. The pelvis is actually a more reliable indicator of sex than the skull, largely because the female pelvis is shaped for childbirth, creating differences that are driven by strong reproductive pressures and are more consistent across populations.11Legal Medicine. Comparative analysis of four morphometric methods for sex estimation When a skeleton includes both the pelvis and the skull, the pelvis gets more weight in the final call.
Does Age Complicate Things?
You might expect that as people age, their skull features change enough to throw off sex estimates. Bone does remodel over a lifetime: the brow ridge can become more pronounced, the mandible angle can shift, and some features soften or coarsen with time. A recent study specifically tested whether age-related changes significantly affected the accuracy of standard cranial sex estimation methods. The findings were somewhat reassuring. Among the traits studied, only the nuchal crest showed significant score differences between women under 40 and women over 40. Overall classification accuracy shifted only slightly when age-specific standards were applied, and the effect of age was considerably smaller than the effect of population background.12PubMed Central. The impact of age-related changes in the skull on sex estimation using morphoscopic traits
The jaw angle, called the gonial angle, does change with age in ways that can differ between men and women. One study found that in the 51-to-60 age range, men had a larger gonial angle on average, but by the 61-to-70 range the relationship had flipped, with women showing the larger angle.13PubMed Central. Gonial Angle in Forensic Anthropology to Determine Age and Gender Shifts like this are one reason the gonial angle alone is not a strong sex indicator and why multi-trait approaches are preferred.
Children’s skulls are a separate problem entirely. Before puberty, many of the sex differences in the skull have not yet developed, because they are driven by hormonal changes during adolescence. Estimating sex from a child’s skull is far less reliable and is generally considered unreliable with morphological methods alone.
How 3D Imaging and AI Are Changing the Game
Traditional methods depend on a trained human looking at bone. Increasingly, researchers are using CT scans and 3D digital models to capture skull shape in fine detail, then running that data through statistical or machine-learning algorithms. A scoping review of the field concluded that 3D medical imaging has improved both the efficiency and consistency of skull-based sex estimation, partly by reducing the subjectivity that plagues visual scoring.14PubMed Central. Sex estimation techniques based on skulls in forensic anthropology: A scoping review
One study using geometric morphometrics on 3D skull models from Bosnia and Herzegovina reported classification accuracy of about 93% for males and 86% for females when both shape and size information were included.15PubMed Central. Sex Estimation Based on the Cranial Base of Three-Dimensional Skull Models from the Bosnia and Herzegovina Population Using Geometric Morphometrics Meanwhile, deep learning, a form of artificial intelligence, has shown even more striking results. In one head-to-head comparison using 200 cranial CT scans from Indonesian individuals, the best-performing deep-learning model achieved 97% accuracy, compared with 82% for a human observer evaluating the same scans.16PubMed Central. Deep learning versus human assessors: forensic sex estimation from three-dimensional computed tomography scans
Another AI prototype, tested on a Thai population using CT images focused on the eye sockets, mastoid process, and the hole at the base of the skull, hit 95% accuracy, dramatically outperforming the traditional measurement-based approach it was compared against.17Journal of Forensic and Legal Medicine. Artificial Intelligence (AI)-driven prototype for sex estimation from cranial Computed Tomography (CT) images These tools are still being validated across different populations and are not yet standard in most forensic labs, but the trajectory is clear: machines trained on thousands of scans can pick up on subtle patterns that human eyes miss.
Mastoid triangle measurements have also been studied with machine learning, where one comparison found that traditional statistical classification achieved about 80% accuracy while machine-learning methods reached roughly 89% on the same data.18PubMed. Sex Estimation From Measurements of the Mastoid Triangle and Volume of the Mastoid Air Cell System Using Classical and Machine Learning Methods
When DNA Disagrees With the Bones
One sobering reality check for morphological sex estimation comes from studies that compare the human observer’s call against genetic evidence. In a recent study of a 19th-century Portuguese skeletal collection, ancient DNA analysis was used to determine genetic sex for comparison against traditional anthropological assessment. The result: a 20% discordance rate. Many of the individuals that anthropologists had classified as “possible female” or “indeterminate” turned out to be genetically male. DNA analysis resolved most of the cases that human observers had been unable to call with confidence.19PubMed Central. When Bones Blur the Lines: Ancient DNA Validation of Morphological Sex Estimation Traits and the Challenges of Population-Specific Dimorphism
This does not mean morphological methods are useless. DNA is often degraded or unavailable, especially in older archaeological remains or fire-damaged forensic cases. But it does highlight that even trained experts working with well-preserved skeletons get it wrong a meaningful fraction of the time, particularly when the individual falls in the ambiguous middle range of dimorphism.
Legal Admissibility of Skull-Based Sex Estimation
In the United States, forensic anthropologists who testify in court must meet evidentiary standards for their methods. Since the 1993 Supreme Court decision in Daubert v. Merrell Dow Pharmaceuticals, judges have served as gatekeepers for scientific testimony, requiring that methods be testable, peer-reviewed, have known error rates, and be generally accepted in the relevant scientific community. After the Daubert ruling, many forensic anthropologists worried that their traditional visual scoring methods, which involve a degree of subjective judgment, might be challenged or excluded in court.20PubMed. The impact of Daubert on the admissibility of forensic anthropology expert testimony
This concern helped drive the push toward quantitative methods, population-specific discriminant functions, and computational approaches that can report their accuracy rates and error margins in concrete numbers. The newer statistical models and AI tools are partly a response to this legal pressure. A method that says “88% accuracy with a 0.1% sex bias in classification” is much easier to defend in court than one that says “I looked at the brow ridge and it seemed masculine.”21Forensic Anthropology. How to Tell If a Skull Is Male or Female?
Transgender Individuals and Skeletal Identification
A growing area of concern in forensic anthropology involves the identification of transgender and gender-diverse individuals. If someone has undergone long-term hormone therapy, some skeletal features may shift slightly from what would be expected based on their chromosomal sex. Facial feminization surgery, which directly alters the skull’s brow, jaw, and chin, can leave visible traces on bone, including surgical cuts, implant materials, or hardware. A systematic review of the literature noted that these surgical modifications could actually serve as identifying features. At the same time, in the absence of known surgical history, a transgender person’s skeleton could show a mismatch between the skull and the pelvis: the pelvis reflects the sex hormones present during growth, while the skull may have been surgically altered later in life.22PubMed. Identification of human remains in case of transgender individuals in forensic anthropology: a systematic review
This is a relatively new area of forensic science, and standardized protocols for handling such cases are still being developed. The key takeaway for practitioners is that a skull-pelvis discrepancy should not automatically be written off as observer error; it may reflect surgical or hormonal history that needs to be investigated through other avenues, such as medical records or dental work.
The Mastoid Debate
The mastoid process is one of the most frequently studied features in sex estimation, but researchers do not fully agree on how useful it is. Several studies have found its measurements to be among the best single-feature predictors of sex. In a study of North Indian dry skulls, the area of the mastoid triangle was the strongest predictor of sex among all the measurements taken.2PubMed Central. Morphometric Study of the Mastoid Triangle for Sexual Dimorphism of Dry Skulls in the North Indian Population But a separate study reached the opposite conclusion, calling the mastoid triangle a poor indicator of sex with limited value without population-specific reference data.23PubMed. Estimation of sex from mastoid triangle – a craniometric analysis
The disagreement probably reflects population differences more than a fundamental flaw in one study or the other. In some groups, the mastoid is highly dimorphic; in others, less so. This is a pattern that runs through the entire field: almost any cranial feature’s utility as a sex indicator depends heavily on which population you are working with. The safest approach, and the one most experts now recommend, is never to rely on a single feature. Multiple traits, ideally scored or measured systematically and interpreted with population context, give the most defensible and accurate result.