Photo-Skull and Video Superimposition

Buck Ruxton was a Lancaster general practitioner born Buktyar Rustomji Ratanji Hakim in Bombay (now Mumbai). He had lived in England since the 1920s and practised medicine in Lancaster under the anglicised name. On the night of 14-15 September 1935, he murdered his wife Isabella Ruxton and their nursemaid Mary Jane Rogerson. He then dismembered both bodies, wrapped the remains in parcels of old newspaper, transported them north, and scattered them along the Gardenholme Linn.

The police investigation identified two reconstruction challenges that determined whether any prosecution was possible: could the two sets of remains be established as definitively human and definitively female (ruling out an animal disposal cover story), and could they be linked specifically to Isabella Ruxton and Mary Rogerson? The second question was the one that brought anatomist John Glaister Jr of the University of Glasgow and his colleague James Couper Brash, Professor of Anatomy at the University of Edinburgh, into the investigation.

The remains showed that both victims had been deliberately mutilated to remove identifying features: fingertips had been removed, noses had been cut away, and lips and ears had been excised. Despite this, the skulls were largely intact. Glaister and Brash obtained portrait photographs of both Isabella Ruxton and Mary Rogerson. They then made life-size photographic enlargements of the portraits and photographically transparent prints of the skulls photographed from the same angle and at the same scale. By overlaying the skull transparencies onto the portrait photographs and aligning them on the orbit margin, the nose bridge, and the chin, they could show that the underlying bone structure was compatible with the soft-tissue features in the portraits.

The superimposition was presented in court in March 1936. Ruxton was convicted and hanged at Strangeways Prison, Manchester on 12 May 1936. The Glaister-Brash report, published in 1937 in Glaister and Brash's book "Medico-Legal Aspects of the Ruxton Case," became the foundational methodological text for photo-skull superimposition worldwide. The case is known in the UK as the "Ruxton case" or the "Lancashire murder"; in Scotland and internationally it is often called the "Buck Ruxton case" to distinguish it from other Ruxton family members.

The Ruxton case showed what the method could achieve when portrait photographs existed and skulls were well-preserved. It did not, in 1936, produce the level of methodological rigour that would satisfy modern evidentiary standards, nor did it articulate the comparison as a probabilistic or categorical opinion. Those refinements came in the following decades, particularly after the method was challenged in courts that demanded a clearer epistemological framework.

The transition from photographic to video superimposition occurred gradually through the 1970s and 1980s, with the video workflow becoming the primary operational method in forensic anthropology laboratories in Europe and North America by the mid-1990s. The video method offers several advantages over the original photographic transparency technique: real-time adjustment of skull orientation, variable mix of the overlay from full skull to full portrait, and the ability to capture the superimposition as a video record that can be reviewed in court.

The workflow begins with the portrait photograph. The photograph must show the face from a known angle (ideally a near-frontal view for an ANS-plane comparison, or a true lateral view for a profile comparison). Any existing photograph can be used, but the examiner must account for the lens focal length, the camera-to-subject distance, and any photographic distortion when setting up the skull. A portrait photograph taken at close range with a wide-angle lens will show a different apparent facial geometry from one taken at a distance with a telephoto lens, even at the same nominal angle.

The skull is mounted on an adjustable turntable with graduated rotation controls for all three planes of movement. A video camera is positioned at the same angular position as the inferred camera position in the portrait photograph. The skull is oriented progressively to match the head position shown in the photograph, using the Frankfurt horizontal plane (defined by the upper border of the external acoustic meatus and the lowest point of the orbital margin) as the baseline for pitch adjustment, and matching the degree of lateral rotation and vertical tilt visible in the portrait.

Once orientation is established, the live video image of the skull is mixed electronically with the portrait photograph. The mix can be adjusted from a faint skull overlay on the portrait, showing how the bone structure aligns with the facial features, to a full skull image with a ghost portrait, showing whether the facial outline extends appropriately beyond the bony landmarks. The examiner then assesses the alignment of specific anatomical landmarks.

The twelve landmarks listed above are assessed in pairs: the orbital margin against the visible position of the eye in the portrait, the zygomatic arch against the cheekbone width, the gonial angle against the jaw angle visible below the cheek, the gnathion against the chin point, and the anterior nasal spine against the base of the nose. The nasal aperture width is assessed against the nose width in the photograph, accounting for the soft-tissue overhang of the nasal alae.

At the ICTY, video superimposition was used as one method among several in the identification of skeletal remains from the Srebrenica massacre and other Bosnian sites. The ICTY's expert witnesses described the method and its limitations in testimony, and the tribunal accepted the method as supporting evidence when combined with other identification data. The same method has been used by Indian CBI forensic experts, by Argentinian EAAF anthropologists, and by several European national forensic laboratories in both DVI and criminal identification contexts.

The most influential critical assessment of photo-skull superimposition reliability was published by M. Yasar Iscan and Richard P. Helmer in 1993, in their edited volume "Forensic Analysis of the Skull" (Wiley-Liss). Iscan and Helmer assembled contributions from multiple international groups who had been developing and using the method across Europe, the Americas, and Asia. The volume also brought together the first systematic attempts to quantify the method's error rate.

The problem that Iscan and Helmer confronted is methodologically fundamental. Superimposition compares a three-dimensional skull with a two-dimensional photograph. The photograph captures soft tissue, not bone. The relationship between soft-tissue surface and the underlying bone varies between individuals (due to differences in facial fat distribution, musculature, and skin thickness), varies across the face (the nose tip has no fixed bone anchor; the cheek soft tissue thickness varies widely), and varies with age, weight, and health status. A skull that is "consistent" with a portrait photograph may be consistent with many other individuals who happen to share similar skeletal proportions.

Helmer's own work, reported in the volume, included experiments in which examiners were shown skulls and portrait photographs from both matched (same individual) and mismatched (different individuals) pairs and asked to categorise them. The results were sobering: false-positive "consistent" conclusions, in which a skull was categorised as consistent with a portrait photograph from a different individual, were not rare. The false-positive rate depended heavily on examiner experience, on the photographic quality and angle, and on how carefully the comparison landmarks were assessed.

Stephan's 2009 critique, published in the Journal of Forensic Sciences, went further. Stephan argued that the error-rate evidence for superimposition had not been systematically collected in a way that would satisfy modern evidentiary standards, and that the method's acceptance in many jurisdictions was based on the authority of the Ruxton case and subsequent practice rather than on a validated, peer-reviewed reliability framework. He observed that the method had been exported beyond the English-speaking world (to India, Brazil, Turkey, Japan, and several other jurisdictions) without the reliability studies that would accompany a new forensic method introduced today.

The forensic anthropology community's response to this critique has been mixed. Several research groups have subsequently published controlled reliability studies with better experimental designs. The consensus that has emerged is that photo-skull superimposition is best understood as a screening or exclusion tool, not a positive identification method in isolation. It can reliably exclude a candidate (a skull that cannot be made consistent with a portrait from any orientation is not the same individual), and it can narrow a candidate list. It cannot, without additional corroborating evidence, constitute the sole basis for a positive personal identification.

The 1985 identification of Josef Mengele in Brazil is the most internationally prominent case in which photo-skull superimposition played a central role. Mengele, the Auschwitz physician responsible for brutal medical experiments on camp prisoners, had fled to South America after the Second World War and died in Brazil in 1979. His body was buried under the assumed name of Wolfgang Gerhard. Following a tip-off, Brazilian and West German authorities exhumed a skeleton from a cemetery in Embu das Artes (then Embu) near São Paulo in June 1985. A multi-national team including American forensic anthropologist Clyde Snow, Brazilian forensic experts, and specialists from West Germany and Israel conducted the identification.

Photo-skull superimposition was performed by Lowell Levine and Leslie Lukash using portraits from Mengele's wartime SS file and from photographs taken in South America. The skull was assessed as consistent with the portraits on multiple landmarks. Odontological comparison, biological profile analysis, and handwriting analysis of documents found with the remains were also performed. The conclusion was that the remains were consistent with Mengele's known characteristics. DNA analysis, requested by the German government and the Simon Wiesenthal Centre as a definitive confirmation, was eventually performed on the bone samples in 1992 by a team including Alec Jeffreys; the DNA comparison using Mengele's son Rolf as a reference confirmed the identification.

The Mengele case illustrates both the utility and the limitation of superimposition as it existed in 1985: it supported the identification and provided the best available evidence given the photographic record, but the final positive confirmation required DNA analysis. This sequence (superimposition as the working hypothesis, DNA as the confirmation) reflects the method's proper role in a multi-evidence identification framework.

The Karl Denke case of 1924 in Germany (Denke killed and consumed multiple victims in Münsterberg, now Ziębice in Poland) is an early historical example where the photographic record of victims was limited and facial comparison was used retrospectively in forensic historical analysis, though not in a courtroom superimposition protocol.

In India, the Aarushi-Hemraj double homicide of 2008 in Noida is the case most frequently cited in forensic anthropology teaching in connection with photo-skull superimposition. The CBI's use of superimposition in the investigation was reported in media coverage and formed part of the contested narrative of the case. The superimposition evidence, like much of the other forensic evidence in the case, was later the subject of significant scrutiny and criticism regarding both methodology and chain of custody. The Allahabad High Court's 2017 acquittal of Rajesh and Nupur Talwar on grounds that included weaknesses in the forensic evidence illustrates the standard to which superimposition evidence, like any forensic evidence, must be held in adversarial proceedings.

The Sheena Bora murder case of 2015 in Mumbai also involved claims of photo-skull superimposition in the CBI's identification of the skeletal remains found in a forest at Raigad. The superimposition work was carried out at the forensic science laboratory and its methodology was examined during the trial proceedings.

In Europe, the German Karl Denke case aside, the method has been used in the identification of victims of the former Yugoslav conflicts. The International Commission on Missing Persons (ICMP), working in Bosnia and Herzegovina, used superimposition as a triage tool in cases where no DNA reference sample was available from family members, to prioritise which candidate identifications merited the expensive DNA matching work.

Digital photo-skull superimposition, which became the operational norm in research-grade forensic anthropology laboratories by the 2000s, uses photogrammetry or structured-light scanning to create a three-dimensional model of the skull and places a two-dimensional portrait image onto the model surface, allowing orientation in three dimensions and comparison with anatomical landmark placement tools. The digital workflow produces a reproducible record of the comparison: every orientation parameter is logged numerically, every landmark placement is stored as a coordinate, and the final overlay image can be independently reproduced from the stored data.

The FaceLab software system (developed by Caroline Wilkinson's group, then at the University of Dundee, subsequently at Liverpool John Moores University) includes a superimposition module that operates in conjunction with three-dimensional skull scanning. The ReFace software, developed through a collaboration involving the US Air Force and FBI, provides a combined workflow for facial approximation and superimposition from three-dimensional skull data. Both systems require the same input: a portrait photograph with sufficient resolution and a known-or-estimated camera geometry, and a skull in adequate condition to provide measurable landmarks.

The limitation that Iscan and Helmer identified in 1993 is not resolved by digitalisation. The fundamental relationship between soft-tissue surface and underlying bone is still subject to the same individual variation, and the digital landmark placement on the portrait is still a subjective judgment about where the soft-tissue feature corresponds to the skeletal feature. A digital superimposition is more precisely reproducible than a video superimposition, but its accuracy (the correspondence between its conclusions and the true biological facts) is bounded by the same biological uncertainties.

Several research groups have tested digital superimposition against ground-truth datasets, where the correct answer (matched or not matched) is known. Wirtz et al. (2010) published a controlled study of 57 skulls against 57 portraits (some matched, some mismatched) and found that experienced examiners using a structured landmark protocol achieved sensitivity (correct identification of true matches as "consistent") of approximately 72 per cent and specificity (correct identification of non-matches as "inconsistent") of approximately 88 per cent. These figures, while better than chance, indicate a non-trivial false-positive rate: roughly 12 per cent of mismatched skull-portrait pairs were rated as "consistent" by experienced examiners.

This performance profile makes digital superimposition most useful as an exclusion tool and a candidate narrower, rather than as a positive identification method. A negative superimposition result (conclusive inconsistency between skull and portrait) is informative: it removes a candidate from the identification list. A positive "consistent" result places the skull in the same candidate pool as the portrait, but does not close the identification without corroborating evidence.

The admissibility and evidential weight of photo-skull superimposition evidence vary across jurisdictions, and the variation reflects differing legal philosophies about how forensic opinion evidence is regulated rather than differing scientific assessments of the method.

In the United States, photo-skull superimposition evidence is submitted as expert opinion under Federal Rule of Evidence 702 (the Daubert standard). Post-Daubert, a federal trial judge applies a gatekeeping function that assesses whether the expert's methodology is reliable (tested, peer-reviewed, with a known error rate, and generally accepted in the relevant scientific community). Superimposition satisfies the "generally accepted" criterion but the error-rate evidence, particularly after Stephan's 2009 critique, has made the "known error rate" criterion more complex to establish. Several US federal courts have limited superimposition evidence to corroborating rather than sole-identification status.

In the United Kingdom, photo-skull superimposition evidence is evaluated under the Criminal Practice Directions (CrimPD) expert witness framework and the Law Commission's 2011 guidance. The reliability requirement established by subsequent case law requires the expert to explain the scientific basis for the comparison, the error rate to the extent known, and the meaning of the categorical conclusion in terms that a jury can assess. UK Crown Court practice since the mid-2000s has consistently treated superimposition as supporting evidence rather than primary identification evidence.

In India, superimposition evidence is submitted under Section 45 of the Indian Evidence Act 1872 (now Section 39 of the Bharatiya Sakshya Adhiniyam 2023), the same expert-opinion provision that applies to DNA evidence. The standard is qualification of the expert and explanation of the methodology; there is no formal reliability hearing equivalent to Daubert. The CBI's use of superimposition in major cases has been challenged in Indian courts, and the Allahabad High Court's treatment of the forensic evidence in the Aarushi-Hemraj acquittal implicitly required a higher standard of methodological transparency than many earlier Indian forensic expert reports provided.

At the ICTY, the reporting language framework of "consistent / inconsistent / cannot determine" was developed specifically to prevent the systematic overstating of superimposition conclusions. ICTY expert witnesses were instructed that "consistent" in the context of superimposition means only that the morphological features examined do not exclude the skull as deriving from the same individual as the portrait; it does not mean that no other individual could share those features. This formulation has been adopted in many subsequent DVI and judicial contexts as the appropriate categorical framework.