What are the scientific facts about skeletal sexual dimorphism and how definitive are they?
Executive summary
Adult human skeletons show clear, measurable sexual dimorphism: the pelvis provides near‑100% accuracy for sex estimation, cranial traits give ~73–77% accuracy, and long bones (femur) yield high but population‑specific accuracy when pelvis/cranium are absent [1] [2] [3]. Dimorphism is driven by hormones, growth factors and mechanical loading and varies by age, population and preservation, so methods work well with the right reference data but are far from universally definitive [4] [5] [6].
1. Pelvis wins; cranium and long bones are useful but fallible
For practical sex estimation the pelvis is the most dimorphic skeletal region, producing near‑perfect discrimination in many studies; the cranium is the next most informative, with mapping studies reporting ~73% accuracy for individual cranial regions rising to 77% when combined (glabellar/supraciliary, mastoid, nasal) [1]. When pelvis and cranium are absent or fragmented, osteometric analysis of long bones such as the femur can still produce high accuracy, but only when models are built on appropriate, population‑specific samples [2] [3].
2. Population and temporal variation change the signal
Sexual dimorphism in the skeleton is not constant across human groups or through time. Research reviews stress “considerable population variation in sexual dimorphism” driven by differences in growth patterns, body size and environmental factors; forensic and archaeological sexing must therefore use regional reference data or risk systematic bias [3] [6]. Multiple authors recommend developing and validating discriminant equations for specific populations—recent Thai femur work illustrates that segment‑specific models can retain high accuracy when tuned to the local sample [2].
3. Age and developmental stage limit reliability—subadults are especially problematic
Adult bones express pronounced dimorphism; subadult skeletons do not. Reviews emphasize that juvenile and subadult remains show overlapping morphology and subtle secondary sexual traits, so sex estimation in young individuals is unreliable and methods often fail or are discouraged [7]. Ageing also shifts morphology—bone loss and muscle decline alter size and form—so age composition of reference samples affects model performance [2] [4].
4. Mechanisms: hormones, growth factors and mechanical loading
Skeletal dimorphism arises from multiple biological mechanisms. Sex steroids (androgens, estrogens), growth hormone–IGF1 axes, and sex differences in bone response to mechanical loading together shape bone mass, size and robusticity; studies conclude dimorphism “is not just the end result of differences in sex steroid secretion” but reflects these interacting systems [5] [4]. That complexity explains why simple rules (e.g., “big bone = male”) can mislead and why sex‑specific physiology matters.
5. Morphology versus soft‑tissue assumptions: a challenge for interpretation
Some traditional morphological sexing methods assume larger muscles create larger bony attachments (entheses). New testing questions that causal link: enthesis size may relate to muscle but not closely enough to serve as a robust sex marker, challenging assumptions behind certain non‑metric traits used in forensic anthropology [8]. This undermines confidence in some qualitative indicators, reinforcing the need for validated metric approaches.
6. New tools improve but do not eliminate limits
Machine‑learning and deep‑learning approaches applied to coxal bones and crania show improved, automated sex prediction and faster workflows; however, gains depend on balanced, representative data and do not remove the core issues of population variation, age effects, and fragmentary preservation [9]. Automated methods may raise average accuracy but inherit biases present in training sets [9] [3].
7. Practical takeaways for forensic and archaeological work
In applied cases, use the pelvis when available; otherwise combine multiple cranial and postcranial measurements and apply population‑specific discriminant functions or validated ML models [1] [2] [9]. Report probabilities or classification accuracies rather than categorical certainty, and explicitly state the reference population, age range and condition of remains because these factors determine how definitive any sex estimate is [3] [7].
8. Limits of current reporting and open questions
Available sources document population, age, hormonal and mechanical contributions to skeletal dimorphism and report performance figures for bones and algorithms, but they do not resolve how best to harmonize modern ML methods with long‑standing population biases, nor do they settle the exact causal pathways linking muscle, enthesis and bone robusticity across all human groups [8] [9] [5]. Further cross‑population, age‑stratified validation remains necessary [6] [3].