Clifford Speigelman and Bill Tobin

In a series of columns in the Austin American Statesman, as well as in our scientific and legal writings, we have indicated that the efficacies of various forensic practices have been greatly overstated in judicial proceedings for many years. Some methodologies are outright scientifically flawed. We have articulated examples of questionable, discredited, and discontinued techniques in forensic practices such as the now-defunct comparative bullet lead analysis (CBLA), firearms/toolmarks identification (aka “ballistics”), fire/arson investigations, and bitemarks, used as indicia of purported guilt. Based on our research and consequent papers, along with those of many other respected scholars of science, law, and forensic practice, courts are beginning to recognize limitations, and outright flaws, of the various forensic practices. For example, in December 2018, in the matter of Steven Mark Chaney v. State of Texas, the Texas Criminal Court ofAppeals(TCCA)[DocketNo.WR-84,091-01] vacated the 1987 wrongful conviction o fChaney primarily because the bitemark evidence introduced against him has been discredited by scientific and legal scholars subsequent to his trial.

During the trial, two of the State’s expert witnesses (odontologists) had individualized an alleged bitemark found on one of the victim’s arms to Chaney. The term ‘alleged’ is used because there was some debate as to (a) whether or not it was even a bitemark and, if it was, (b) whether it was from a human or animal, and (c) when it was sustained relative to the homicide.One expert testified “to a reasonable degree of dental certainty” (a scientific-sounding but meaningless assertion) that Chaney was the biter. The second expert’s testimony was even more egregiously inflammatory: he testified not only that the bitemark on the victimwasa“ perfect match” to Chaney’s exemplar, and worse, but also that there was a “[o]ne to a million” chance that someone other than Chaney caused the bitemark. The TCCA sided with a lengthy list of highly respected scientific and legal scholar signatories to an amicicuriae(‘friends of the court’) brief and concluded that there was no foundational validity for such claims as testimonial evidence. The TCCA set aside Chaney’s conviction and ruled Chaney actually innocent. Chaney lost nearly 30 irrecoverable years of his life wrongfully incarcerated. Used by prosecutors in criminal trials since approximately 1692, the State of Texas eventually disallowed its use as evidence of guilt since the Chaneyruling.

This is but one example of forensic practices considered venerable evidence used for decades in criminal trials that were eventually discredited when exposed to true scientific scrutiny, practices developed not by true scientists but rather by law enforcement personnel staffing crime laboratories.

Some forensic practices have acceptable scientific bases for the underlying measurement phase of the particular practices, but misleading or objectionable basis for extrapolation of measurement results for use in assessment of guilt or innocence (known as ‘probative value’)in criminal trials. This column will focus on one such forensic procedure called gunshot residue analysis, known in law-enforcement circles by the acronymGSR.

GSR characteristic particles originate from the discharge of a firearm and are primarily comprised of unburnt and partially burned ammunition propellant and primer. The particles are spherical and typically comprised of 3 compositional elements: barium, lead, and antimony.

In evaluating GSR, one should consider an adage guiding statisticians and other scientists: correlation does not imply causation. Occurrences of two events or experimental outcomes that trend together do not necessarily cause each other. For example, the numbers of Nobel prizes and per capita chocolate consumption for different countries are highly correlated. If chocolate consumption actually caused Nobel prizes, science students would consume much more chocolate than they currently do. If two variables are highly correlated, only when no third, possibly unknown, influential variable can explain the relationship can they be said to be causally related.

Unrelated correlations are known in science as ‘spurious relationships.’ For years, many spurious relationships abounded in popular beliefs: that chicken soup cured colds and that bulls become angry at the sight of red objects, are examples. There are hidden conflating influential variables subtly imbued in those erroneous beliefs of misleading correlations. The existence of additional hidden conflating influential variables in the interpretation of GSR is the central thesis of this column.

The basis for forensic GSR practice is the belief that shooters, individuals in proximity to shootings, and/or those exposed to shooters or items handled by shooters, are more likely to have GSR on their hands than others not in those categories. That belief is logically valid. Thus, the presence of GSR particles and proximity to a shooting and/or contact with a shooter are logically correlated.

However, as with Nobel Prizes and chocolate consumption, correlation is not dispositive of causal relationship. For forensic application, it should be obvious that, if there is no causal relationship between GSR presence and proximity to the discharge of a firearm during a specific criminal event under investigation, the use of GSR evidence is, at best, useless (has no probative value) and, at worst, is misleading and prejudicial to a defendant. According to the Israeli police, it is for this reason that they limit the prospective sample pool by “consider[ing] only the situation in which the suspect declares that he was neither near the crime scene nor in physical contact with a weapon elsewhere.  This is motivated by the policy of our [Israeli] lab to exempt a suspect fromGSRtestingifhehastiedhimselftothecrimesceneorifhebelongstoapopulationlikely to be contaminated by GSR particles (for example hunters or soldiers).” (Damary, N., et al., “Calculation of likelihood ratios for gunshot residue evidence – – statistical aspects”, doi:10.1093/lpr/mgw001.

The Israeli police policy is justified by available GSR prevalence studies, limited as they are. In one international prevalence study, over 35% of gun owners were found to have GSR  on their hands. In the USA generally, there are no similar corresponding exclusions in U.S. law enforcement practice. In Harris County, Texas, 3 particles are arbitrarily considered incriminating; in Bexar County, Texas, even a single particle is considered incriminating, regardless of gun owning status, occupation orhobbies.

The 95% confidence intervals (95 percent plausible range of values) from recent prevalence studies range from 0 at the low end, to 2 to 6 percent at the high end, for the general (non-gun owning) population expected to have GSR on their hands. Thus, for approximately 2 or more percent of the general population, GSR analyses are not at all correlated to a particular crime. Given that approximately 2% of the US are Vietnam era veterans, it would be similar to introducing forensic evidence that is inadvertently incriminating to such veterans.

When one considers both the percentage of US households with guns, and that approximately 35% of gun owners have GSR on their hands at any point in time (according to GSR prevalence studies), then approximately 8% of people with GSR on their hands are in the non-gun owning general population. Thus, at any point in time, a randomly selected citizen might have incriminating evidence on their person that prosecutors could allege to be connected to a homicide by firearm. That situation is not unlike vulnerability to random arrest for using or dealing drugs if you have U.S. currency in your wallet or purse.       In some cities, 100% of the U.S. currency samples were found with detectable traces of cocaine on it. (For the 90% study, reference Available online  at

Given the lack of correlation for approximately 2 percent of the population, gun owners, and perhaps others, we believe are evaluation of the use of GSR in criminal trials is appropriate. The evaluation should encompass groups to be excluded from potentially incriminating GSR testimony, the numbers of GSR particles required for such testimony, and interpretational cautions required to be presented at trial when such testimony isgiven.

Cliff Spiegelman is distinguished professor of statistics at Texas A&M University and statistics advisor to the Texas Forensic Science Commission and the Houston Forensic Science Center.

Bill Tobin retired as head of forensic metallurgical operations for the FBI Laboratory and is now in private practice with Forensic Engineering International. He examines evidence in criminal, civil, and non-litigious matters, including transport and mining disasters, and has testified in 45 states and in Congress as an expert witness.