Firearms and Fingerprint Error Rates: What the Black Box Studies Counted

In an FBI Laboratory study of firearms examiners, 173 volunteers made 8,640 bullet and cartridge case comparisons, and on the bullets that truly came from different guns, they reached the correct answer, an elimination, only 33.8% of the time. They called a wrong match on just 20 of those 2,842 bullet sets. Almost all the rest came back “inconclusive.” Both numbers are true, and which one a jury hears depends on who’s doing the counting.

The papers behind the error rates quoted for gun and fingerprint identifications hold more than the rates themselves. The published false-positive rates are low in some designs, and that cuts in the State’s favor. The papers also show where those rates come from, what got left out of them, and one fingerprint study where the false match rate ran past one in five.

Follow this site in Google:

The FBI Firearms Study

The study is Monson, Smith and Peters, “Accuracy of comparison decisions by forensic firearms examiners,” from the FBI Laboratory, published in the Journal of Forensic Sciences in 2023. Examiners from 41 states compared a questioned bullet or cartridge case against two known ones. It was an open set, meaning a questioned item didn’t always have a match in the packet, which the authors note avoids the underestimate of false positives that a closed set builds in. FBI examiners were excluded. The participants volunteered, and the paper calls them “a sample of convenience.”

The headline results: a false positive rate of 0.656% for bullets and 0.933% for cartridge cases, and a false negative rate of 2.87% for bullets and 1.87% for cartridge cases. The authors chose guns, steel-cased ammunition and consecutively made parts that would produce hard comparisons, and they wrote that their error rates “may provide an upper bound” for casework. That’s the strongest point for the State, and it’s a fair one. Errors also clustered: 80% of examiners made no errors on bullets, while the worst performer called 4 of 18 nonmatching bullet sets an identification.

Where the Inconclusives Went

The paper states that “Inconclusive conclusions shall not be considered errors.” So inconclusive answers on nonmatching sets weren’t counted as false identifications, and the headline false positive rates come from a statistical model that lets each examiner’s error rate differ. On the 2,842 nonmatching bullet sets, examiners gave 961 eliminations, 20 identifications and 1,861 inconclusives (answers that neither match nor exclude) across three levels. The paper’s own table puts specificity, the share of true nonmatches that examiners actually excluded, at 33.8% for bullets and 48.5% for cartridge cases.

The authors explain why they see inconclusives as a safeguard, and the paper cites others in the field who agree. Still, the numbers tell you something a jury should hear. When the bullet really came from a different gun, these examiners didn’t say so in roughly two of every three sets. In a case where the examiner calls an identification, the question is how that examiner handles the hard, close-looking nonmatches, and a one-line error rate doesn’t answer it.

Two Examiners, One Bullet

The same FBI team published a follow-up in 2023 on repeatability and reproducibility. They sent the same sets back out, either to the same examiner without telling them or to a different examiner. On nonmatching bullets, two different examiners gave the same answer only 31.62% of the time, on a scale that splits inconclusive into three levels. In another 42.68% of the pairs, one examiner eliminated while the other called it inconclusive.

The fair point for the other side: wrong identifications were rare in this round too, and none was repeated by a second examiner. The authors say that supports blind verification as a check in casework. That’s why I ask whether a second examiner checked the work without knowing the first answer.

Fingerprints and the Close Non-Match

A close non-match is a print from a different person that shares many features with the crime scene print and differs on only a few. Koehler and Liu, “Fingerprint error rate on close non-matches,” Journal of Forensic Sciences (2021), looked at a mandatory proficiency test taken by 125 fingerprint agencies that included two of these pairs. The false positive rates were 15.9% (17 of 107) and 28.1% (27 of 96). Those rates leave out the agencies that answered inconclusive; combined, it’s 44 false identifications among 203 definite answers, which the authors’ conference abstract gives as 21.7%.

That study has real limits. The agencies were in China, and the abstract says the instructions discouraged inconclusive answers, in keeping with practice there. Examiners also saw eight matching points followed by one point that didn’t match. I’ve read only the abstract and the conference abstract, not the full paper, so I don’t lean on anything past those numbers.

A 2025 study of latent prints, the kind lifted from crime scenes, points the other way. Hicklin and colleagues, building on the 2009 FBI and Noblis black box study, gave 156 examiners prints paired with candidates returned by the FBI’s Next Generation Identification database, and wrong identifications came in at 0.2% of responses on nonmated pairs (prints from different people). One participant made the majority of those errors, and none was reproduced by a second examiner. Erroneous exclusions were 4.2% on mated pairs. I’ve read the abstract of that paper, not the full text.

What Florida Law Asks of This Testimony

Florida judges decide whether opinion testimony comes in under section 90.702, Florida Statutes. The Florida Supreme Court adopted the Daubert version of that statute in 2019, replacing the older Frye test. The statute allows the opinion only if “(1) The testimony is based upon sufficient facts or data; (2) The testimony is the product of reliable principles and methods; and (3) The witness has applied the principles and methods reliably to the facts of the case.”

Each of those three prongs maps onto the studies. A study rate is a group average, and the FBI authors wrote that their confidence intervals “should not be interpreted as bounding the error probabilities of any one examiner.” The third prong asks about this examiner and this comparison. A newer bullet study, Hicklin and colleagues in 2024, went further and recommended “against using overall decision rates” to summarize its results, because the rates swung with the gun model, the ammunition and the examiner.

What I Ask the Lab For

When a fingerprint or a firearms match ties someone to a crime, I ask for the bench notes and the images the examiner relied on, the examiner’s proficiency test history, how inconclusive calls are handled at that lab, and whether verification was blind. For a print, I ask whether the candidate came from a database search and how many candidates came back, because a large database search is where close non-matches turn up. Then I ask which study the analyst will cite for an error rate, and whether that study counted inconclusives, used an open set and looked like the comparison in my client’s case.

My fingerprint and pattern evidence page explains how I challenge a print identification, and the firearm evidence page covers toolmarks, gunshot residue and DNA on a gun. The forensic science defense section covers the rest of the lab evidence in a Florida criminal case.

Further Reading

The sources behind this post, so you can read what I read.

  1. Monson, Smith and Peters, “Accuracy of comparison decisions by forensic firearms examiners,” Journal of Forensic Sciences 68(1):86-100 (2023)
  2. Monson, Smith and Peters, “Repeatability and reproducibility of comparison decisions by firearms examiners,” Journal of Forensic Sciences 68(5):1721-1740 (2023)
  3. Koehler and Liu, “Fingerprint error rate on close non-matches,” Journal of Forensic Sciences 66(1):129-134 (2021)
  4. Hicklin et al., “Accuracy and reproducibility of latent print decisions on comparisons from searches of an automated fingerprint identification system,” Forensic Science International 370:112457 (2025)
  5. Hicklin et al., “Accuracy and reproducibility of bullet comparison decisions by forensic examiners,” Forensic Science International 365:112287 (2024)
  6. In re Amendments to the Florida Evidence Code, 278 So. 3d 551 (Fla. 2019)
  7. Section 90.702, Florida Statutes

This post is general information about the law, not legal advice, and it does not create an attorney-client relationship. Past results do not guarantee a similar outcome. The law discussed is section 90.702, Florida Statutes, and In re Amendments to the Florida Evidence Code, 278 So. 3d 551 (Fla. 2019). The studies discussed are Monson, Smith and Peters, J. Forensic Sci. 68(1):86-100 (2023); Monson, Smith and Peters, J. Forensic Sci. 68(5):1721-1740 (2023); Koehler and Liu, J. Forensic Sci. 66(1):129-134 (2021); Hicklin et al., Forensic Sci. Int. 370:112457 (2025); and Hicklin et al., Forensic Sci. Int. 365:112287 (2024).

Available 24/7 for Immediate Defense

Your first strategy session is free. One call can start protecting your future today.
Get in Touch

You’re better Safir than sorry!

Arrested for DUI? Time matters. Complete the form to schedule a free strategy session with attorney Rory Safir. Your information is confidential, and we will follow up promptly.

200+
Client Testimonials
1 of 6
Forensic Lawyer-Scientists in Florida
4.9★
Google Rating
24/7
Availability

Let’s Go Over Your Case


Email Newsletter