Crime Scene Reconstruction: From Traces to Events

Reconstruction as a whole has not been subjected to a black-box study with known event histories and measured error rates. Chisum and Turvey's Crime Reconstruction says that little scientific research has examined reconstruction as a process, that no universal principles have been articulated and that it is among the least understood forensic procedures. It describes Locard's exchange principle as an axiom accepted without proof.

Some reconstruction claims rest mainly on practitioner experience. Hazelwood and Napier's 2004 account of crime-scene staging used a telephone survey of twenty law-enforcement officers known to the authors, who estimated cases from memory. The authors said that documented staging statistics were unavailable and that the results should not be generalised. Those limits should accompany any reliance on the study.

Black-box studies also require scrutiny. Khan and Carriquiry (2023) note self-selected participants and non-response on difficult comparisons. In their analysis, a reported error rate "as low as 0.4%" became at least 8.4% when inconclusive responses were treated as correct and above 28% when treated as missing. The existence of a study therefore does not resolve design and counting questions. For much of reconstruction, no comparable study exists.

Intermediate targets also deflect bullets. Nishshanka et al. found in 2021 that rifle-round deflection through sheet metal increased as incidence angle fell, reaching several degrees below 40°. Below about 20°, fragmentation produced secondary marks. Mattijssen's 2018 particle-board study found peak deflection near a critical angle; a .32 round arriving at 18° left 12° off line. Haag showed that a tumbling bullet may be nose-forward during part of its cycle and leave a round entrance, so wound shape alone does not prove a direct path. Nishshanka later observed shallow-angle concrete ricochets producing nose-forward impacts resembling orthogonal hits.

Blood droplets are affected by gravity and drag. Straight-line stringing may place the origin too high. Attinger et al. reported in 2019 that the possible origin region grows approximately with the fifth power of distance, from about 0.1 litre near the wall to 95 litres for a slow impact just over one metre away. Schyma's 2020 high-speed study of 102 shots found backspatter jets during collapse of the temporary wound cavity, not penetration, with stochastic deviation around 13° despite the same firearm and ammunition.

Walters and Liscio tested the 2D ellipse method on 220 impacts with 31 analysts in 2020. It performed well at shallow angles but poorly near perpendicular, where a 1 mm marking error could produce a 17° angle error. Keldson and Liscio found errors from 1° to 13.9° when estimating shooter position from car-door bullet holes, with calibre-specific error curves. Riva et al. reconstructed shooting geometry from body scans in 2021. Standing postures produced errors of a few degrees, but assuming the person was leaning when he was in fact sitting upright produced mean error of 15.8° and could reverse the inferred direction.

Base rates also matter. Keppel and Weis reviewed 5,224 Washington State murders from 1981 to 2000 and found six staged scenes, about 0.1%, and fourteen posed bodies. Most investigators may never encounter a staged homicide. Krywanczyk and Shapiro found statistical differences between homicide and suicide blade wounds but no wound characteristic that was definitive; every feature favouring homicide had also appeared in a documented suicide.

Van den Eeden and colleagues gave experienced scene investigators an ambiguous mock scene after telling them it involved suicide, murder or no stated hypothesis. The prior affected first impressions and the traces collected, although similar traces were physically available. A 2019 comparison found no evidence that experts outperformed novices. Lidén and Almazrouei reported two similarly experienced investigators collecting 15 and 38 samples from the same scene, with case theory influencing search decisions.

Edmond et al. call the downstream effect the biasing snowball. An opinion influenced by the case theory may later be treated as independent corroboration of that same theory. Case management and sequenced information exposure can reduce this circularity by documenting physical observations before broader theories are introduced. Awareness alone is not an adequate control.

Martire et al. found a weak-evidence effect in how lay participants interpreted evaluative opinions. When an expert intended an LR of 495,000, the median participant updated as though it were 1.5. The lowest verbal category, "weak support," led many to treat prosecution evidence as favouring the defence. De Keijser and Elffers found that 92.4% of judges, 80.9% of defence lawyers and 58.2% of forensic experts endorsed a transposed conditional, while 85% of experts believed they understood the reports almost perfectly.

The 2025 OSAC terminology standard for shooting reconstruction defines 89 terms concerning angle, ricochet, trajectory and directionality, but does not provide a method for expressing uncertainty, probability or evidential strength. Any use of "consistent with" therefore needs an explanation of the alternatives and degree of support.

John Morgan's 2023 Journal of Forensic Sciences study coded 732 wrongful convictions associated with false or misleading forensic evidence, covering 1,391 examinations. Testimony errors appeared in 376 cases and errors by officers of the court in 519. Morgan estimated that improved technology, testimony standards or practice standards available at trial could have prevented about half of the convictions. Incorrect source attribution was less common than misleading expression and use of the evidence.

The New South Wales Court of Criminal Appeal quashed Gordon Wood's murder conviction in 2012. A physicist had reconstructed a "spear throw" from a cliff using daylight experiments with cooperative volunteers on safe ground. The court found opinions outside his expertise, unsophisticated experiments, a clearly wrong assumption and active participation in making evidence. It reiterated that speculation, conjecture and suspicion cannot establish proof beyond reasonable doubt. A defensible reconstruction ties conclusions to documented physical evidence, states assumptions and uncertainty, uses undetermined when support is insufficient, and remains within the examiner's expertise.