Pattern Recognition: A New Look At Old Forensics

At last month’s Innocence Project symposium at the American Chemical Society meeting in Philadelphia, speakers explained that many traditional forensic disciplines lack scientific validation. Several researchers are examining those traditional fields with a quantitative lens.

For instance, quantum chemist Nicholas D. K. Petraco is working on a technique that, while not yet ready to be brought into a courtroom, might someday help associate a weapon such as a Glock pistol with crime scene evidence. The marks those pistols leave on cartridge cases—the part that holds gunpowder—are typically used by forensic experts to pair a cartridge with a gun. Petraco, of John Jay College of Criminal Justice at the City University of New York, uses computational pattern recognition and confocal microscopes to build three-dimensional models of striation and impression marks on cartridge surfaces (Scanning, DOI: 10.1002/sca.20251). Others are using Raman spectroscopy to link gunshot residue to ammunition (C&EN, May 21, page 36).

At the State University of New York, Buffalo, Peter J. and Mary A. Bush are part of a team that examines bite-mark analysis. The technique assumes each person’s bite mark is unique, but the field lacks data supporting this assumption. The Bush team is using statistics and tooth scans obtained from a company that makes mouth guards to examine whether individual tooth arrangements are unique (J. Forensic Sci., DOI: 10.1111/j.1556-4029.2010.01531.x). They’ve also borrowed a technique from paleontology and evolutionary biology that allows them to measure the extent of variation in the pattern a bite leaves behind (Forensic Sci. Int., DOI: 10.1016/j.forsciint.2011.03.028).

The team is using these indirect approaches for several reasons—including regulations from human subject research review boards. “You can’t just go out and bite a bunch of people,” Peter Bush says. The team places bite marks on research cadavers with casts of teeth mounted in a vise grip to precisely control bite pressure. But the model isn’t fully representative of bite marks at a crime scene. The researchers cannot replicate the violent altercation that goes along with the act of biting. The cadavers they use don’t retain all the properties of living skin. “But in science you have to start somewhere,” Peter Bush says. The results thus far, which have already been introduced in a handful of hearings, indicate that bite-mark analysis should be undertaken with caution because multiple suspects could have similar teeth.