Sex Estimation from Long Bones and the Sub-Adult Sex Problem

Postcranial metric sex estimation relies on the body-size sexual dimorphism of the human skeleton. Males, on average, have larger long bones, wider joint surfaces, and heavier cortical bone than females of the same population. These differences are expressed as discriminant functions: linear combinations of measurements that maximally separate the two sexes in the reference sample. Applied to an unknown individual's measurements, the function places the individual on the discriminant axis and assigns a sex based on whether the score falls above or below the population-specific sectioning point.

The logic of the discriminant function requires that several conditions are met. First, the reference sample must be sex-documented (known sex from records, not estimated from the skeleton). Second, the reference sample must be demographically representative of the population the unknown individual is most likely from. Third, the measurements must be taken in a standardised way using a calibrated instrument. Fourth, the sectioning point, which defines the threshold between the male and female zones, must be derived from the same reference sample used to develop the function, or cross-validated on an independent sample from the same population.

When these conditions are met, published discriminant functions for long bones achieve approximately 80 to 90 per cent correct classification in the reference population. When the case population differs from the reference population, accuracy drops, sometimes substantially. The Stewart (1979) femoral head discriminant function, for example, was derived from US military personnel and applied to a US casework context. Applications of the same cut-offs to Indian or East African samples consistently produce poorer results, because the average femoral head diameter in these populations is smaller than in the US reference, and the same cut-off incorrectly classifies more individuals.

The population-specific calibration principle applies equally to long-bone metrics as to cranial and pelvic morphological methods. It is not a minor qualification but the precondition without which the method's accuracy figure cannot be interpreted.

The femoral head vertical diameter (also called the maximum diameter or superoinferior diameter of the femoral head) is the most widely used and best-validated single long-bone measurement for sex estimation. It reflects both the joint surface size required to support body mass and the overall lower limb length, both of which are sexually dimorphic.

T.D. Stewart (1979) provided the foundational US cut-offs in the classic text 'Essentials of Forensic Anthropology':

• Femoral head diameter greater than 47.5 mm: classified as male.
• Femoral head diameter less than 43.5 mm: classified as female.
• Femoral head diameter between 43.5 and 47.5 mm: indeterminate (the ambiguous zone).

These cut-offs were derived from a predominantly White American US military and dissection room sample. Accuracy in the US reference population is approximately 85 to 90 per cent for individuals who fall outside the indeterminate zone, and the indeterminate zone contains approximately 20 to 25 per cent of individuals in the US sample.

For Indian populations, published studies have consistently found smaller femoral head diameters than the US reference, with correspondingly lower sectioning points. Mukherjee (1955) and Khanpetch et al. (2012), in a study of Thai and South-East Asian femora that is often cited alongside Indian data because of shared gracile body build characteristics, found that optimal sectioning points for South and South-East Asian populations were approximately 2 to 3 mm lower than the Stewart cut-offs. Pan's (1924) early study of Indian femora from the Calcutta medical school collection provided initial Indian reference means. More recent data from Indian forensic medicine departments, including studies from AIIMS and the Maulana Azad Medical College in Delhi, confirm male means of approximately 43 to 45 mm and female means of approximately 37 to 40 mm in Indian samples, compared with US male means of approximately 48 to 50 mm and female means of approximately 42 to 44 mm.

In South Africa, Iscan and Miller-Shaivitz (1984) and later Patriquin et al. (2005) derived discriminant functions from the Pretoria Reference Collection. Their optimal sex sectioning point for the femoral head was slightly higher than the Indian data but slightly lower than the Stewart US figure, reflecting an intermediate body-size profile in the South African Black sample.

Humeral head vertical diameter. The humeral head vertical diameter (superoinferior diameter of the humeral head) is the second most commonly used long-bone metric for sex estimation. Published cut-offs for US samples (Iscan 1983, France 1998) suggest that values above 47 mm are most consistent with male sex, values below 43 mm with female sex, and values in between are indeterminate. Accuracy in the US reference is approximately 82 to 87 per cent. As with the femoral head, Indian and South-East Asian humeral heads are on average smaller than US reference values, and population-specific cut-offs should be applied. Indian reference data from the Khanpetch et al. (2012) South-East Asian study and from Indian forensic medicine studies suggest optimal cut-offs approximately 2 mm below the US values.

Scapular glenoid breadth. The glenoid cavity on the scapula (the shoulder socket) is sexually dimorphic in its anteroposterior and superoinferior diameters. Dittrick and Suchey (1986) published discriminant functions for the glenoid using a California sample. More recent validation (Iordanidis 1961; Steele 1976 updated by Giles and Elliot) has confirmed that glenoid breadth achieves approximately 75 to 80 per cent sex classification accuracy, lower than the femoral and humeral head. It is most useful when only the proximal upper limb elements survive.

Tibial nutrient foramen width and proximal tibial dimensions. The proximal tibia and its articular surface show sexual dimorphism that has been exploited in sex estimation. The mediolateral width of the proximal tibial plateau (the plateau width) is significantly larger in males. Iscan and Miller-Shaivitz (1984) and France (1998) published tibial discriminant functions with accuracy of approximately 80 to 83 per cent in the US reference sample. The tibial nutrient foramen position, by contrast, is not a sex indicator; it is cited in some older texts as a side-determination landmark but it does not carry sex information.

Maximum length of long bones. While femoral and humeral maximum length are used primarily for stature estimation (the Trotter-Gleser equations for the US, Mukherjee 1955 and Khanpetch 2012 for India and South-East Asia), they also carry sex information through their overlap with joint surface dimensions. Males are on average taller, with longer femora and humeri. However, the overlap in length distributions between male and female is substantial enough that length alone is rarely used for sex estimation; joint surface dimensions provide better separation with less overlap.

Multiple postcranial measurements can be combined in a multivariate discriminant function to improve sex classification accuracy beyond what any single measurement achieves. The standard approach is to use a linear discriminant function of the form: D = b1(x1) + b2(x2) + ... + bn(xn) + c, where b1...bn are the discriminant coefficients derived from the reference sample, x1...xn are the measurements on the unknown individual, and c is the constant. The sectioning point (typically the midpoint between the two group centroids) divides the discriminant axis into the male zone and the female zone.

Mehmet Iscan and colleagues published a suite of postcranial discriminant functions through the 1980s and 1990s, covering femur, humerus, tibia, and radius. Steele's (1976) equations for the femur were among the earliest systematic multivariate approaches. France (1998) published a comprehensive set covering all four major long bones and the scapula, using a documented US reference sample. These functions, together with the FORDISC software interface (which allows input of standardised postcranial measurements alongside the cranial set), constitute the current US-derived toolkit for metric postcranial sex estimation.

The multivariate functions achieve accuracy of 85 to 90 per cent in US reference samples when multiple measurements from a single element are combined. Combining measurements across elements (femoral head plus humeral head, for example) provides incremental improvement and can push accuracy toward 90 per cent in ideal cases. The practical limitation is that each additional measurement requires the corresponding element to be present, measurable, and undistorted by taphonomy. In fragmented forensic cases, the full multivariate battery is rarely achievable; more commonly the osteologist applies the best available single or dual measurement discriminant function for whatever elements survive.

Indian-calibrated postcranial discriminant functions are less comprehensively published than the US set. The available studies (Khanpetch 2012 for South-East Asia, individual Indian forensic medicine department studies) confirm the direction of the dimorphism and provide adjusted means and cut-offs but rarely derive full multivariate discriminant functions with cross-validated accuracy statistics. This represents a gap in the Indian forensic anthropology evidence base that limits the precision of postcranial sex estimation in Indian casework compared with US or European casework.

Sexual dimorphism in the human skeleton is largely a post-pubertal phenomenon. Before puberty, the male and female skeleton are morphologically very similar because the hormonal differences between sexes, primarily testosterone and oestrogen and their differential effects on periosteal bone formation, endosteal resorption, and growth plate activity, have not yet acted on the growing skeleton to produce the adult pattern.

This means that the traits used for sex estimation in adults (the ventral arc, the sciatic notch width, the mastoid process, the femoral head diameter) are either absent, poorly expressed, or not yet dimorphic in sub-adult remains. A child's pelvis shows a sciatic notch, but it is not yet shaped by the adult female obstetric flare or the adult male narrowing. A child's skull has a mastoid process, but it has not yet been remodelled by differential muscle loading. A child's femoral head is measurable, but its diameter falls below even the female adult range, and the male-female separation has not yet emerged.

The timing of the transition is approximately as follows:

• Before age 11 (females) or 12 (males): essentially no reliable morphological sex dimorphism in any skeletal region.
• Ages 11 to 15 (the peri-pubertal period): partial development of dimorphism; some traits beginning to show sex differences but the within-sex variance is high and the between-sex overlap is large.
• After age 16 to 18 (when most growth plates are fusing): adult dimorphism is substantially established, and the adult methods can be applied with approaching adult accuracy.

The forensic consequence is that for any case involving sub-adult skeletal remains (estimated age at death below approximately 12 to 15 years), the classical morphological and metric methods are unreliable, and any sex determination based on them should be reported with very wide uncertainty or declined entirely. This is not a hypothetical problem: Indian missing-child cases, child homicide investigations, and sub-adult skeletal remains recovered from clandestine graves frequently raise the question of sex, and an osteologist who reports sex based on a child's sciatic notch morphology is reporting an error. Sub-adult age at death in these cases is most reliably established from dental eruption stages and epiphyseal fusion.

Geometric morphometric methods capture biological shape by recording the three-dimensional coordinates of anatomical landmarks on a skeletal element and then analysing the landmark configurations statistically after removing differences due to position, orientation, and scale (Procrustes superimposition). The residual shape variation can be analysed by principal component analysis or discriminant function analysis to search for sex-associated shape differences.

The appeal of geometric morphometrics for sub-adult sex estimation is that it can detect subtle shape differences that exist before the full adult dimorphism develops, differences that are too small or too continuous to be captured by the discrete scoring or simple linear measurements used in classical methods.

Wilson et al. (2011) applied three-dimensional geometric morphometrics to dental root shape, using landmark configurations on mandibular molar roots to distinguish male from female sub-adults. Their result, on a documented US sub-adult sample, was approximately 70 to 75 per cent accuracy in peri-pubertal individuals (ages 8 to 14), rising toward 80 per cent in older sub-adults. The discriminating shape differences appear to reflect differential timing and pattern of root formation related to earlier female dental maturation. The method requires computed tomography (CT) scanning to image dental root shape accurately without sectioning the tooth, which is a resource requirement that limits its routine application in lower-resource casework contexts.

Pelvic geometric morphometrics has been applied to sub-adult ischia and ilia in several studies (Vlak et al. 2008; Gonzalez 2009). The pelvic approach attempts to capture the early differentiation of pelvic shape that precedes the full adult female flare. Results have generally been in the range of 65 to 75 per cent accuracy for individuals under 12, improving in older sub-adults. The method requires well-preserved, un-fragmented pelvic elements, which are rarely the forensic reality in sub-adult remains.

The current state of geometric morphometric sex estimation in sub-adults is that it offers accuracy in the 65 to 80 per cent range in research settings, well below the standard required for a courtroom sex determination. The methods are not yet in routine forensic casework use in most jurisdictions. The ENFSI and OSAC documentation on forensic anthropology methods does not yet endorse any geometric morphometric sub-adult sex method at the level of a validated standard. The practical forensic tool for sub-adult sex determination remains DNA amelogenin typing.

The amelogenin gene encodes the principal protein component of dental enamel and is located on both the X chromosome (AMELX) and the Y chromosome (AMELY). The two alleles differ in size: the Y-chromosome version has a 6-base pair deletion in intron 1 relative to the X-chromosome version, producing amplicons of different length when the flanking region is amplified by PCR. The commercially available forensic STR multiplex kits, including AmpFlSTR Identifiler Plus (Applied Biosystems), GlobalFiler (Applied Biosystems), PowerPlex 16 (Promega), and the CODIS-16 compliant kits used by the US FBI, all include amelogenin as a standard locus. It is co-amplified with the STR loci in a single reaction, adding no additional laboratory workload.

Interpretation is straightforward: two bands (both X-length and Y-length) indicate a 46,XY genetic male; a single band (X-length only) indicates a 46,XX genetic female. The method determines genetic sex, which is the same as biological sex in the vast majority of individuals; rare chromosomal conditions (45,X Turner syndrome; 47,XXY Klinefelter syndrome; 46,XY complete androgen insensitivity) may produce discordant results, but these conditions are rare enough (combined prevalence below 0.2 per cent of the population) that they are not operationally significant for forensic sex determination at the population level.

Amelogenin typing is applicable to sub-adults at any age because it is not dependent on the morphological expression of sex dimorphism. It can be performed on dental pulp (the richest cellular DNA source in teeth), cortical bone, or dentine, all of which survive for decades to centuries under good preservation conditions. Teeth are preferred because their enamel shell protects the underlying dentine and pulp DNA from environmental degradation. In a sub-adult case where only deciduous or developing permanent teeth are recovered, amelogenin typing from dental tissue is the standard approach.

Amelogenin typing is a PCR-based method conducted in a forensic DNA laboratory following the same contamination control protocols, quality assurance standards, and chain-of-custody requirements as STR profiling. The biology of the amelogenin system, including the detailed PCR and capillary electrophoresis steps, is covered in the companion Forensic Biotechnology topics. From the forensic anthropologist's perspective, amelogenin provides a genetic sex determination that operates independently of skeletal age and morphological state, making it the preferred first-line method for all sub-adult cases and a valuable corroborative tool for adult cases where the skeletal sex assessment is ambiguous.

In Indian casework, the FSL (Forensic Science Laboratory) DNA divisions in major states, and the CFSLs (Central Forensic Science Laboratories) in Delhi, Kolkata, Hyderabad, and Chandigarh, all routinely perform amelogenin typing as part of standard DNA profiling. The DNA Technology (Use and Application) Regulation Bill 2019, which proposed to govern forensic DNA use in India and would have covered identity-determination testing including sex determination from DNA, was passed by the Lok Sabha but never enacted into law: it was withdrawn from Parliament in July 2023 without passing the Rajya Sabha. In the UK, amelogenin is a mandatory component of the National DNA Database (NDNAD) profiling kit. In the US, it is part of all CODIS-compliant multiplex kits used by the FBI and state crime laboratories.

1. Assess age-at-death first
   Before applying any sex estimation method, establish whether the remains are adult or sub-adult. Use dental development (Moorrees, Demirjian, AlQahtani) and epiphyseal fusion (Scheuer-Black) to estimate age range. If the age estimate is below approximately 15 years, classical morphological and metric sex methods are not applicable; proceed directly to DNA amelogenin typing.
2. Pelvic assessment (if elements present)
   If any pelvic elements are recoverable and assessable, apply the [Phenice triad and Walker sciatic notch scoring](/topics/forensic-anthropology/sex-estimation-from-the-pelvis-phenice-and-sciatic-notch) first. The pelvis is the most reliable skeletal sex indicator for adults. Postcranial long-bone metrics are supplementary to the pelvis, not a substitute.
3. Cranial assessment (if skull present)
   If the skull is present and the pelvis is absent or too damaged, apply [Walker 2008 five-trait ordinal scoring](/topics/forensic-anthropology/sex-estimation-from-the-skull-and-mandible). Population-specific calibration applies here as for long-bone metrics.
4. Long-bone metrics (if pelvis and skull absent)
   Measure femoral head vertical diameter, humeral head vertical diameter, and any other measurable joint surface dimensions. Apply the population-appropriate sectioning point: Stewart 1979 US values for a North American case, Indian-calibrated adjusted values (approximately 2-4 mm lower for femoral head) for a likely Indian case. Apply multivariate discriminant function if multiple measurements from the same element are available.
5. Sub-adult: request DNA amelogenin typing
   For individuals estimated under approximately 15 years of age, prepare a tooth or cortical bone sample for DNA extraction and amelogenin typing via the forensic genetics laboratory. This should be initiated as early as possible in the investigation because DNA analysis takes time and the biological sex answer is unambiguous.
6. Geometric morphometrics (research context only)
   If the case is being processed in a research or academic context with access to CT scanning and landmark analysis software, geometric morphometric dental or pelvic analysis can be attempted as supplementary evidence for sub-adult sex. Report accuracy as approximately 65-80 per cent (peri-pubertal) and explicitly note this method is not yet validated to court standard in most jurisdictions.
7. Composite report
   Combine all applicable methods. For adults: pelvis (primary), skull (secondary), long bones (tertiary). For sub-adults: DNA amelogenin (primary); note that skeletal methods are not applicable. Always state the reference population used, its documented accuracy, and any calibration adjustments made for a non-US/non-European case.