A man stepped into a rural South Carolina bank a few days before Christmas in 2001, aimed a gun at tellers and stole $7,800 from the drawers. Witnesses couldn’t identify the robber. The surveillance video was too grainy to help investigators.

More than three years later, FBI agents narrowed the investigation to a suspect. They believed John Henry Stroman robbed the bank. But during questioning, Stroman told them the security footage instead showed his brother, Roger. How could investigators prove one brother was the robber and not the other? Agents shipped the video and pictures of both Stromans to the FBI Laboratory in July 2005.

The package went to Richard Vorder Bruegge, one of the bureau’s image examiners.

In his report, Vorder Bruegge wrote that John Henry Stroman and the robber had similar “overall shape of the face, nose, mouth, chin, and ears.” But Vorder Bruegge stopped short of declaring a match, saying the video and pictures were too low resolution for that.

Nevertheless, prosecutors said in court filings that if Vorder Bruegge took the stand, he would testify that “the photograph is of sufficient resolution to definitively state that the robber is John Henry Stroman.” The judge said the testimony would be admitted if the case went to trial. A week later, Stroman accepted a plea deal.

It wasn’t the first time, nor the last, Vorder Bruegge’s lab results said one thing and the courts were told something different. Court records and FBI Lab files show statements by prosecutors or Vorder Bruegge veered from his original conclusions in at least three cases.

Vorder Bruegge, who earned a doctorate in geology 28 years ago, came to the FBI after abandoning his hopes of becoming an astronaut. He had no crime laboratory experience, but he quickly became a force in the forensic sciences.

Now 55, he is the most prominent member of the Forensic Audio, Video and Image Analysis Unit at the FBI Lab in Quantico, Virginia. The unit’s comparisons can advance investigations by sharpening pictures and narrowing the list of suspects. But most of the image examiners’ lab work has no scientific basis proving their methods are reliable and findings are correct.

A ProPublica investigation, published in January, found that image examiners’ methods have never had a strong scientific foundation. The bureau’s use of the unit’s findings as trial evidence troubles many experts and raises anew questions about the role of the FBI Lab as a standard-setter in forensic science.

Such shortcomings could have led judges to block image analysis from criminal trials. But Vorder Bruegge single-handedly built a body of case law that has kept the FBI unit’s testimony admissible in the courts. His 22-year-old comparison of bluejeans is the legal foundation for most photo comparison methods.

The FBI Lab’s image unit had routinely used unproven techniques since the 1960s, but Vorder Bruegge embraced and expanded them, according to court records and his published articles. At times, he has given jurors baseless statistics to say the risk of error was almost zero. Studies on several methods in the past decade have found them unreliable.

Today, Vorder Bruegge is one of the nation’s most influential crime lab scientists. He serves on the Forensic Science Standards Board, which sets rules for every field, from DNA to fingerprints. He’s a co-chair organizing the American Academy of Forensic Sciences meeting this week in Baltimore, a gathering of thousands of crime lab professionals, researchers, lawyers and judges.

Vorder Bruegge has testified for the federal government so often, and so successfully, that a 2013 law review article referred to him as “perhaps the most ubiquitous of the United States expert witnesses.”

ProPublica requested an interview with Vorder Bruegge and sent him written questions. The FBI declined the requests and did not respond to the questions.

Vorder Bruegge has produced an extensive public record detailing and defending his practices during his 24-year career at the FBI. Image analysis typically involves scrutinizing pictures from crime scenes to determine if suspects’ faces, hands, clothes or cars match, according to court documents and published articles. Examiners base their identifications on the pattern they see along a shirt seam, the shape of an ear or a cluster of freckles.

At a conference in Seattle last year, Vorder Bruegge recounted the most common criticism he hears from defense attorneys: he’s just looking at pictures, no different than anyone else with eyesight.

Vorder Bruegge has a ready response.

“Yes, I’m just looking at this pair of images,” Vorder Bruegge said, “the same way a radiologist looks at an X-ray. Anyone in this room can look at an X-ray, just look at it. But who do you want deciding what type of treatment you are going to get as a result of examination of that radiograph? Do you want anyone in this room to determine if you have cancer, or if you just have an artifact in your image?”

Radiologists and FBI image examiners both work with pictures. The similarities end there. Radiology is exhaustively researched and its methods continually tested to make certain they are reliable. Radiologists must graduate from medical school and complete four-year residency programs before they diagnose patients.

Image examiners rarely have advanced degrees. New examiners learn how to analyze pictures by doing casework with lab veterans. Their methods remain unproven, at best.

Vorder Bruegge, however, has not only a doctorate in geology but an ease with the language and standards of science. At public events, he sounds like a progressive voice urging crime labs to improve, said Hal Stern, a University of California, Irvine, statistics professor who researches forensic science methods.

Despite that public image, Vorder Bruegge has used unproven science throughout his career.

“It’s a little disturbing, to be sure,” Stern said.

From the Cosmos to Forensics

Vorder Bruegge moved to Providence, Rhode Island, in the fall of 1981 for his freshman year at Brown University, a couple of hours from his family’s home in Connecticut. He majored in electrical engineering and spent summers working for a data processing company.

His focus turned sharply during a planetary science course on the solar system, taught by geology professor James Head III. Over the years, Head’s lectures have inspired many Brown undergraduates to study space, “including Rich,” said Scott Murchie, who met Vorder Bruegge while both were graduate students.

Vorder Bruegge completed his bachelor’s degree in engineering in 1985, then secured a place on Head’s research team. He would study Venus’ mountain belts while earning a master’s degree and doctorate in planetary geology. He met his future wife, a fellow Venus researcher, and aspired to join the NASA space program.

Duane Bindschadler worked alongside Vorder Bruegge examining Soviet radar images of Venus’ surface. Vorder Bruegge was innovative from the start, Bindschadler said, “trying to come up with new interpretations or extract new information from them.”

The research required complex image analysis, said Murchie, now a planetary scientist at the Johns Hopkins University Applied Physics Laboratory. Vorder Bruegge was one of several impressive students working with Head in the late 1980s. (Ellen Stofan, director of the Smithsonian’s National Air and Space Museum, was another.)

“I have nothing but wonderful things to say about Rich,” Murchie said. “He was a young researcher with a great deal of integrity.”

Vorder Bruegge finished writing his doctoral thesis in October 1990 and went to work that same month for a NASA contractor in Washington. He was providing technical support for space missions, but he intended to make it a stepping stone. “The person whose job I took left to become an astronaut and that was actually something I was trying to do, so I thought it would be a good career move,” Vorder Bruegge said during a 2008 court hearing in which he was asked about his credentials and training.

NASA chose new astronaut classes in 1992 and 1994; Vorder Bruegge didn’t make it to the interview stage of the intensely competitive process either time.

In January 1995, he again veered onto an entirely different course, taking a position at the FBI Lab on what was then called the Special Photographic Unit. He’d examine security video rather than spacecraft images.

Vorder Bruegge’s move to the FBI surprised some of his colleagues, said Bindschadler, now a systems engineer at the NASA Jet Propulsion Laboratory. “Scientists with kind of an academic bent, that isn’t the first place you think that people are going,” he said. “Especially if they’re in the physical sciences. I doubt the FBI employs more than one geologist.”

Vorder Bruegge’s resume shows that, even with his Ivy League degrees and image analysis experience, he started with a two-year apprenticeship under the unit’s veterans, same as any other photo examiner. But he proved his value shortly thereafter.

The U.S. Supreme Court had recently raised the standard for scientific evidence to require proof that methods are reliable. No one had tested any of the FBI Lab’s long-standing photo comparison techniques, let alone proven them trustworthy. Defense lawyers might be able to block image examiners’ testimony from trials outright.

The high court’s opinion lists several ways a method can meet the new requirement, including “peer review” — scrutiny by outside experts — and publication in a scientific journal.

In a 1996 bombing and bank robbery case in Washington state, Vorder Bruegge identified bluejeans in surveillance footage as the pair seized from a suspect. He used one of the unit’s established techniques: matching the series of light and dark spots along the seams.

Vorder Bruegge’s testimony helped convict defendants in the bombings. Then he used the case to secure something vital for his team: publication in a scientific journal. The new image examiner wrote an article describing the unit’s method of comparing bluejeans’ seams in pictures and submitted it to the Journal of Forensic Sciences.

In the article, Vorder Bruegge used technical terms — “ridge-and-valley pattern” and “planar surface” — that echoed his doctoral thesis about mountains on Venus. He included pictures showing his results, zoomed-in images of bluejeans with arrows pointing where the seams and hemlines allegedly matched.

The journal accepted Vorder Bruegge’s article and published it in spring 1999. The article repeatedly acknowledged that the technique had not been validated. Nonetheless, court records show Vorder Bruegge referenced the article in at least a dozen trials and hearings as proof that the image unit’s methods were reliable evidence.

At ProPublica’s request, several forensic scientists, statisticians and clothes manufacturing experts reviewed Vorder Bruegge’s article. They said the FBI examiner’s central claims were misleading or wrong.

Building a Legal Foundation

But in the years after the article on bluejeans identification was published, Vorder Bruegge won acclaim for his work. Newspapers ran short articles characterizing the method as a forensic science breakthrough. In interviews, Vorder Bruegge gave credit to his predecessors at the FBI . “I’m really standing on their shoulders,” he told the Chicago Tribune in June 1999, adding, “It’s exciting to find ways to show that everything around us is unique.”

The television documentary series “Forensic Files” aired an episode about the Washington state case a couple of years later. It featured Vorder Bruegge extensively, even showing him outfitted in a full-length lab coat to take pictures of bluejeans.

Over the following decade, Vorder Bruegge went on a legal winning streak. He convinced judges across the country that unproven methods were sound science.

ProPublica searched court databases and found more than a dozen criminal cases involving Vorder Bruegge’s lab work since 2000. In those cases, judges overruled each request from defense lawyers to block his testimony. The FBI did not respond to questions about Vorder Bruegge’s casework.

Courts have historically permitted evidence from the FBI Lab, sometimes without considering its accuracy. “Jurors think that if you’re a big FBI examiner you know it all,” said Alicia Carriquiry, an Iowa State University statistics professor and director of the Center for Statistics and Applications in Forensic Evidence.

Vorder Bruegge’s statements contradicted his written lab results in at least three cases, court and FBI Lab records show. His testimony in several other trials indicate he improvised techniques.

In a 2002 trial highlighted in ProPublica’s investigation of image analysis, Vorder Bruegge testified that he had identified a defendant’s plaid shirt as the shirt a robber wore to seven banks during a spree in South Florida. He told jurors only 1 in 650 billion plaid shirts would randomly match as precisely as the defendant’s shirt.

None of Vorder Bruegge’s lab reports included calculations to support the statistics he gave in court. In fact, the reports said nothing about how he reached his conclusions. And for one of the robberies, Vorder Bruegge wrote he could not conclusively match the defendant’s shirt to the robber’s. He said the opposite on the stand, according to trial transcripts.

At the time, Vorder Bruegge led a group that wrote most of the guidelines for law enforcement image analysis. It compiled a list of criminal cases in which judges ruled that examiners’ testimony was scientific evidence. Those provided the field with a kind of legal foundation, giving judges a clearer path to admitting photo comparison evidence.

The plaid shirt case, U.S. v. McKreith, was the first to win clearance for image analysis. Vorder Bruegge’s facial comparisons in the South Carolina bank robbery case, U.S. v. Stroman, is another.

In child pornography cases, prosecutors must often provide evidence that video and pictures show actual children. Such “authentication” has become part of FBI image unit’s regular caseload. During a 2008 federal court hearing in Boston, the transcript shows Vorder Bruegge estimated a victim’s age in a picture based solely on the size of her breasts and pubic hair.

The image analysis group lists that case, U.S. v. Frabizio, as another piece of the field’s legal foundation.

In his presentations and articles,Vorder Bruegge hasn’t mentioned perhaps the most remarkable legal victory of his career. To bolster a conviction that was being challenged, Vorder Bruegge took the stand in 2010 to assail his field’s most common method and dispute his own lab results.

A jury had convicted 19-year-old Brian Avery in 1995 of participating in two armed robberies at Milwaukee convenience stores. Prosecutors built their case on witness identification and Avery’s confession during police interrogation, court records show. (Avery recanted his confession and declared his innocence thereafter.) The trial judge sentenced him to more than 20 years in prison.

Lawyers at the Wisconsin Innocence Project took up Avery’s case in 2007. They hired a video analyst to measure the robber’s height in surveillance footage. The images had been too fuzzy to use at trial. But in intervening years, software engineers had developed programs to filter and sharpen pictures and others to measure the distance between pixels.

FBI examiners have calculated criminals’ heights in pictures for decades using a collection of techniques called “photogrammetry.” Computers increasingly allowed the bureau and others to analyze low-quality images.

Avery stood 6 feet, 3 inches tall when police booked him into jail.

The innocence project’s video expert, unaware of Avery’s actual height, determined the robber was under 6 feet tall, according to court records. The defense lawyers filed a motion for a new trial.

Local prosecutors asked the FBI Lab to see if its own analysis could include Avery. Vorder Bruegge determined the robber could not have been taller than 6 feet, 2 inches, court records show. Therefore, the Lab’s own results also found Avery was at least an inch taller than the robber, which the defense team argued exonerated him.

However, Vorder Bruegge’s testimony at an appellate hearing did otherwise. He said the photogrammetry method is too imprecise to reliably rule out a suspect. “I am saying you can’t determine absolutely that it can’t be this person,” he said of Avery. “That is the bottom line of my examination.”

The bureau’s examiners have presented height measurements in court in scores of criminal cases.

Under cross examination, Vorder Bruegge acknowledged he knew Avery’s height before starting his analysis. That height — 6 feet, 3 inches — was what prosecutors hoped Vorder Bruegge would calculate for the robber. Such information can influence how an examiner performs lab work and reaches conclusions.

More than a dozen scientists and law professors filed a brief with the Wisconsin Supreme Court urging the judges to disregard the FBI examiner’s testimony as severely biased by trial records, especially details on Avery’s height.

The state Supreme Court did not disregard Vorder Bruegge’s testimony. Rather, the judges accepted his argument that height measurements aren’t scientifically reliable enough and denied Avery a new trial.