Empirical Analysis of Saw Mark Characteristics in Human Bone: Meeting Forensic Standards in Dismemberment Cases

In cases of criminal dismemberment, forensic anthropologists are tasked with providing medicolegal agencies with accurate information about the dismemberment methods and tool(s) utilized. Residual kerf variables are used to predict tool class characteristics, such as power source, tooth set, tooth shape and tooth size. This information can assist in the search for suspects and weapons and as evidence during legal proceedings. Current methods, however, rely mostly on anecdotal relationships between cut characteristics and tool parameters, with forensic bodies calling for more objective and statistically rigorous methods. Some studies have tried to address these concerns, but use animal models, de-fleshed remains, and/or bones restrained in devices, which does not accurately reflect forensic human dismemberment events. Many studies also have limited sample sizes or test for a single parameter and tend to focus on incomplete cuts. The proposed project will provide an overdue empirical assessment of the reliability of kerf variables to predict saw class characteristics, provide required error rate documentation, and facilitate standardization in forensic anthropological saw mark analysis for application in criminal dismemberment cases.
In the proposed project, twenty manually- and mechanically-powered reciprocating saws will be used to create experimental saw cuts (complete and incomplete) in a cadaveric (fleshed) human sample, mimicking true dismemberment scenarios. A second sample of dissected donors (“unfleshed”) will be used to assess the effect of soft tissue on residual cut characteristics. Two previously procured saw cut samples (bones and/or negative molds) are available for analysis. Saw mark data will be collected from these samples as well, resulting in a combined sample of 908 saw cuts from 44 saws. A subsample of 70 cut surfaces will be utilized for blinded inter- and intra-observer error analyses. Random forest models/decision tree analyses will be performed on reliable features to assess utility and accuracy in predicting tool class characteristics. Expected outcomes include development of statistically rigorous methods to estimate saw class characteristics; documentation of error rates; creation of an atlas with expanded feature definitions and numerous photographic examples to assist others in forensic research and case comparisons; and a large collection of saw mark data and molds that will be made available to others. This research will advance the field of forensic anthropology, but will also benefit medicolegal agencies, criminal justice  systems, and the general public, as any erroneous conclusions drawn from unreliable saw mark analyses can have major legal implications on investigations and convictions. CA/NCF