Some common forensics techniques aren’t as reliable as courts presume: professor

A police dog sniffs out narcotics concealed in the depths of a nondescript car. An officer links a bullet found at a crime scene to a specific gun — and a specific suspect. An expert details the sequence of an attack based on the pattern of bloodstains in the area.

Such scenarios have been mainstays in crime shows and court cases for years, but new research suggests these forensic and investigative techniques are far less reliable than many presume.

Sara Gordon, an associate professor of law at the University of British Columbia, examined how Canadian courts deal with evidence related to sniffer dogs, ballistics analysis and blood spatter patterns in a series of papers, one recently issued and two slated for publication.

“With all three of them I saw this sort of tendency of courts and attorneys generally to just sort of assume that all of these different techniques are reliable,” Gordon said in a recent interview.

“We’ve all seen ‘CSI.’ People just sort of automatically believe that if a police officer is testifying that a bullet was fired from a particular gun, they must be correct. And that’s just not the case with all of these different techniques.”

While they have the “veneer of science,” the vast majority of forensic techniques weren’t developed in a lab or university but in police departments, Gordon said.

It’s only in the last 10 to 20 years or so that their validity has been put to the test empirically, and “for many of them, the evidence isn’t supportive of the types of conclusions examiners tend to reach,” she said.

For one thing, each of these techniques is vulnerable to some form of bias, she noted.

“If the police officer is pretty sure that a person has narcotics, they might unintentionally signal that to the dog. They can do that, just something as simple as kind of slowing down when they walk past the person and that can, unconsciously on the human’s part … cue the dog to alert maybe even if there’s not drugs present,” Gordon said.

In ballistics and bloodstain pattern analyses, there’s a risk of contextual bias, since the majority of examiners are police officers, she said.

“By virtue of their position, they have access to information about the crime scene that can be biasing,” she said.

“For instance, if a bloodstain examiner has information about the crime scene, you know, maybe who was involved, the relationship between the victim and the accused … they might go into that crime scene with a preconceived expectation about what they might find, and that can bias their interpretation of what they see there at the crime scene.”

A couple of major reports in the United States flagged the lack of scientific support for several forensics techniques used throughout North America about a decade ago, setting off a flurry of research, Gordon said.

Courts, however, have largely not caught up to the science, said Gordon, adding she aims to provide that information with her papers so that judges and attorneys can make informed decisions about the admissibility of evidence.

The Supreme Court of Canada last looked at the constitutionality of sniffer dog searches in 2013, finding that police could conduct such searches with less cause than for others because they are minimally intrusive and accurate.

Since then, a number of studies across jurisdictions have shown sniffer dogs to be far less reliable than believed, Gordon said.

Successful dog alerts, meaning those that lead to the discovery of illicit drugs, “are often no better than chance,” according to her paper in the Alberta Law Review. An ombudsman’s review out of New South Wales in Australia examined more than 10,000 alerts by 17 drug detection dogs and found that dogs falsely alerted to the presence of narcotics 74 per cent of the time, the paper said.

“There are no national standards for the training, testing, and deployment of police sniffer dogs in Canada, nor any meaningful transparency about how these functions are carried out by the dozens of canine units around the country,” Gordon wrote in the paper.

Dog searches are also far from minimally intrusive, she said.

“It can be embarrassing, it can be humiliating, and it can really scary for people. And that’s especially the case for racialized people or people who are sort of often subject to enhanced police scrutiny,” she said in the interview.

As such, the same standard of reasonable and probable grounds should be applied to sniffer dog searches to make sure they comply with the Charter of Rights and Freedoms, she wrote in the paper.

Canadian courts have admitted ballistics evidence for nearly a century based primarily on examiners’ testimony that it’s reliable, “without subjecting those claims to meaningful scrutiny or limiting examiner conclusions to those that are empirically supported,” according to Gordon’s upcoming paper in the Dalhousie Law Journal.

While courts in the United States have started to push back, Gordon said she could only find one case in Canada where a judge refused to admit forensic firearms testimony due to concerns over the scientific validity of the technique.

The case, heard in Ontario five years ago, included testimony from a firearms examiner with the province’s Centre of Forensic Sciences, who said shell casings found at the scene were fired from a handgun found at the home of one of the accused, she wrote in the forthcoming paper.

Under cross-examination, the examiner said there was no specific number of “agreements,” or tiny corresponding markings on the two cartridge cases, required to conclude there was a match, the paper said.

That was one of the concerns raised by the judge in rejecting her testimony, the paper said. The judge also noted the examiner had not undergone proficiency testing, and that she had acknowledged the absence of a discipline-wide error rate for ballistics examiners, it said. He pointed to a landmark 2016 report in the United States that found no scientific support for the validity or reliability of firearms analysis.

Gordon’s paper said the validation studies ballistics examiners typically invoke to back up their expertise “suffer from significant methodological flaws,” including unrepresentative samples, testing conditions that don’t align with actual working conditions and the categorization of inconclusive results as correct. Meanwhile, examiners routinely present their conclusions in “highly certain terms” without disclosing the limitations of the technique, it said.

Ballistics examiners “will regularly testify in court that the error rate for that forensic technique is one per cent, one-and-a-half per cent,” the researcher said in the interview.

“It is absolutely not true that the error rates are that low. And forensic examiners just shouldn’t be permitted to testify to that.”

Recent studies on bloodstain pattern analysis have also shown very high levels of error, up to 32 per cent, Gordon wrote in another upcoming paper, this one for the Canadian Bar Review.

The technique does not appear to have received the scrutiny courts are required to apply to novel or contested scientific evidence, the paper said. Rather, courts have focused on the training and experience of the examiners in deciding whether to admit the evidence, “taking it for granted that the field is sufficiently reliable,” the paper said.

The principle underlying blood spatter analysis is said to be fluid dynamics, an “incredibly complicated” physics principle, Gordon said in the interview. While police officers receive training, some courses are as short as a weekend, and the vast majority of officers just don’t have the scientific training to apply those physics principles, she said.

“It’s a scientific language that dresses up this forensic technique, which is largely just the examiner … looking at a particular pattern and saying, oh, that matches up with this other pattern that I’ve read about or that I studied,” she said.

“The problem is that they often try to go far beyond that and recreate crime scenes in really extreme detail,” she said. The science doesn’t support those kinds of conclusions, which are also particularly vulnerable to contextual bias, she said.

As a result, crime scene reconstruction based on bloodstain pattern analysis should generally be inadmissible, she wrote in the paper.

“A forensic technique with error rates exceeding 30 per cent should not be relied upon to establish guilt beyond a reasonable doubt,” the paper said.

This report by The Canadian Press was first published Nov. 10, 2026.

Paola Loriggio, The Canadian Press