Facial Reconstruction: How Experts Rebuild a Face From a Skull, and How Accurate It Is

Every reconstruction starts with the skull itself. Before any clay goes on, the practitioner assesses the person's likely sex and age, the ancestry indicators forensic specialists look for, and the proportions of the face. They study where the muscles attached to the bone, the shape of the nasal aperture (the opening for the nose), the eye sockets, which specialists call orbits, and the cheekbones, jaw and chin.

Each of those features constrains the face. According to Caroline Wilkinson, the orbits help set the position of the eyes, and the nasal aperture helps estimate the general width and projection of the nose. The jaw, chin, cheekbones and forehead set the overall shape of the face. A skull can also carry personal details: a healed injury, signs of disease, lost teeth or an unusual bone feature can give a reconstruction something individual.

If you are curious about the bones involved, our page of human skull facts covers the parts of the skull, and what bones tell archaeologists explains how age and sex are read from a skeleton.

There are three main approaches, and modern work often combines them.

The tissue-depth method. Small markers are fixed at anatomical landmarks on a cast or digital copy of the skull. Each marker is cut to the average thickness of soft tissue at that point. Muscle, connective tissue and skin are then built up until they meet the tips of the markers. The average values change with age, sex, population and body mass, so the data have to suit the person being rebuilt.

The Manchester method. This approach grew out of the work of Richard Neave and Caroline Wilkinson. It combines the skull's anatomy, the muscle attachments, tissue-depth markers and anatomical standards that predict features such as the nose and eyes. Where the bone gives no firm answer, the practitioner uses artistic judgement. In her paper "Facial Reconstruction: Anatomical Art or Artistic Anatomy?", Wilkinson describes it as a combination of anatomical interpretation, tissue-depth data and artistic modelling.

Digital sculpting. CT or laser scans turn the skull into a three-dimensional computer model. Software then fits tissue-depth data, anatomical landmarks or statistical face models to it. Claes and colleagues described such a statistical method in 2006 and validated it with a database of 118 individuals. A digital face can be rendered as an image or printed as a physical model.

In every method, the same sequence holds: read the bone, add tissue to known depths, shape the features the bone supports, and make careful choices about the rest.

A skull does not hold an exact portrait. It cannot directly reveal the precise shape of the lips, the ears or the eyelids, a person's expression, their hairstyle, how much fat they carried, or many of the finer details of the nose.

Facial fullness is especially hard. Bone does not show how much soft tissue or fat covered the face. In a 2015 study, Lindsay notes that facial approximation normally assumes an average body weight, while body mass index changes how thick the soft tissue of the face really is.

Colour is another limit. Skin, hair and eye colour cannot normally be read from the shape of a skull. Ancient DNA can sometimes give probabilities for pigmentation, but that depends on how well the DNA has survived and which genes can be analysed. Other evidence can help, such as the archaeological setting, a mummified body, portraits, preserved hair or historical records. Unless one of those directly preserves the detail, it stays an estimate.

People often ask whether the ears and nose are just guesswork. The nose is partly guided by the bone, because the nasal aperture gives its general width and projection, but its finer shape is not. The ears are among the least reliable parts of any reconstruction. Ancient mummies are a special case. When soft tissue survives, the artist has much more to work from, as the Ötzi reconstruction below shows.

Reconstructions have been put to the test, more than once, and the results are mixed. Broad structure is reproduced fairly well. Recognizing a specific person from a reconstruction alone is harder.

In Wilkinson and colleagues' evaluation of the Manchester method, 67 percent of the reconstructed facial surface was within 2 millimetres of the real face, in tests involving two adults. The nose, eyes, jawline, forehead and chin matched more reliably than the mouth and ears. Wilkinson also reports a blind recognition study in which Manchester reconstructions scored more than 50 percent above chance.

Other tests are less encouraging. In a double-blind study published in 1989, Helmer and colleagues reconstructed 12 skulls. Most results showed an approximate to strong general likeness to the real person, and the authors noted that information about hair colour, hair length and hairstyle made the faces more useful. A 2017 study of computerized approximations, by Richard and Monson for the FBI, found that people who knew the subjects picked the correct reconstruction in 10.5 percent of 220 tests, against a random-chance rate of 9 percent.

Scientists disagree about what these numbers mean for identifying people. Some laboratory tests produce recognizable likenesses, while others show that a reconstruction on its own performs poorly. An early forensic evaluation by Snow and colleagues warned that a restored face should not establish identity by itself. In practice, a reconstruction is used to produce leads and public recognition, and other evidence confirms who someone was.

Some reconstructions are famous in their own right. Each shows a different mix of bone, scans, DNA and preserved evidence.

Facial reconstruction also has a quieter, more personal use. Forensic teams turn to it when someone's remains have been recovered but there is no usable photograph, dental record or DNA match to name them. The skull is analysed, scanned or cast, and a possible face is shared with investigators, entered in missing-person databases or released in a public appeal.

Its value is in starting a conversation. A face may prompt a relative or witness to recognize someone, suggest a possible identity, or help rule out people who are not a match. It is not normally enough on its own. Confirmation comes from stronger evidence, such as DNA, dental comparison, fingerprints, medical records or a distinctive feature of the skeleton.

Caroline Wilkinson notes that reconstruction can help when remains are badly decomposed or otherwise hard to recognize. Success depends on the condition of the skull, the population data used, the skill of the practitioner, the information available about hair and colouring, and how the final image is presented to the public. Each of these faces belongs to someone whose family may still be waiting for a name. If you have come across bones yourself, our page on finding human bones explains what to do first.

When you next stand in front of a reconstruction, it helps to read it in layers. Trust the overall shape of the head and face, the set of the jaw and the position of the eyes, because those come from the bone. Treat the nose as well guided in outline and estimated in detail. Read the lips, ears, the fullness of the cheeks, the hairstyle and the expression as the artist's best informed choices. Check where the colour came from: in Richard III's case it came from DNA, and in Ötzi's case from a preserved body, while in many ancient faces it is an estimate.

Seen that way, a reconstruction is a meeting point between a real person's bones and the careful judgement of the person rebuilding them.

The questions people ask most about rebuilding faces from skulls, answered briefly.

Every face in this story belongs to someone with a history of their own. Read about Richard III's skeleton and how it was identified, meet Ötzi the Iceman and his 61 tattoos, and see how Neolithic people at Jericho gave their dead new faces in plaster in our story on the Jericho plastered skulls. For older faces still, read about the Neanderthal skull and the oldest human skull.