Daubert vs Frye comes down to which courtroom you are in. Federal courts and most states apply the Daubert standard, which asks the judge to confirm that an expert’s method is reliable and was applied properly to the case. A handful of states, including California, Illinois, Minnesota, New York, Pennsylvania and Washington, have historically held to the older Frye test, which asks whether the technique has gained “general acceptance” in its field. This guide covers US federal law and names states where they differ.

The gap between the two tests matters less than many people assume, but the way each one is policed has shifted a good deal since 2023, especially for forensic evidence.

The short version: Daubert versus Frye

Question Frye Daubert / Rule 702
Origin Frye v. United States, 293 F. 1013 (D.C. Cir. 1923) Daubert v. Merrell Dow, 509 U.S. 579 (1993)
Core question Is the technique generally accepted in the relevant scientific field? Is the testimony reliable and relevant, judged by the judge as gatekeeper?
Who decides In effect, the scientific community The trial judge
Typical focus New or novel techniques Any expert opinion, scientific or technical
Where used A minority of states Federal courts and most states

The state list comes from a general reference summary and may not be exhaustive. Court rules change, so anyone with an actual case needs to check the current rule in the specific court.

Where Frye came from

The 1923 case concerned an early lie-detection device that measured blood pressure. The appeals court refused to let the jury hear the results, writing that a technique must be “sufficiently established to have gained general acceptance in the particular field” to which it belongs. That sentence governed American courtrooms for seventy years.

Frye’s appeal was simplicity. A judge did not have to evaluate the science. The judge only had to count heads in the relevant discipline. Its weakness was the mirror image: a sound technique could be kept out simply because it was new, and a widely used but poorly tested one could slide through on habit alone.

What the Supreme Court changed in 1993

In Daubert, decided June 28, 1993, the Court held that the Federal Rules of Evidence, not Frye, control expert testimony in federal trials. General acceptance survived, but only as one factor among several. The Court listed considerations for judges while stressing that it was not setting a rigid checklist:

  • whether the theory or technique can be tested, and has been;
  • whether it has been through peer review and publication;
  • its known or potential error rate;
  • whether standards exist to control how it is operated;
  • whether it is generally accepted in the relevant community.

Two later decisions filled in the picture. In General Electric v. Joiner (December 15, 1997), the Court held that appellate courts review these rulings for abuse of discretion, and that a judge may exclude an opinion when too large a gap separates the data from the conclusion. In Kumho Tire v. Carmichael (March 23, 1999), it ruled that the same gatekeeping applies to engineers and other experience-based experts, not just laboratory scientists. You can read the original opinions through Cornell’s Legal Information Institute: Daubert, Joiner and Kumho Tire.

What Rule 702 says now

Federal Rule of Evidence 702 now carries the Daubert approach in its text. A qualified expert may testify only if the side offering the testimony shows the court that it is more likely than not that four things are true: the specialized knowledge will help the fact-finder; the testimony rests on sufficient facts or data; it is the product of reliable principles and methods; and the opinion reflects a reliable application of those methods to the facts. The full wording is on Cornell’s page for Rule 702.

The version quoted above took effect on December 1, 2023, according to a DLA Piper summary. Its point was to correct a habit in some courts of treating weak factual support or sloppy application as a matter of “weight” for the jury. The committee’s note says those questions belong to the judge, under the preponderance standard of Rule 104(a). It also says no new procedures are created, and that courts need not demand perfection.

Forensic evidence is where the test gets hard

Lab-bench science is one thing. Pattern-matching disciplines, in which a trained examiner compares marks and decides whether they come from the same source, are another. A 2009 National Research Council report, Strengthening Forensic Science in the United States: A Path Forward, found a fragmented system with uneven standards and weak research support, and made 13 recommendations. A later review by the President’s Council of Advisors on Science and Technology (PCAST), issued in 2016, looked at how well specific feature-comparison methods had been validated.

The 2023 committee note speaks to this directly. It says forensic experts using subjective methods should avoid claiming absolute or near-certain accuracy, and that judges should be given error-rate estimates where possible. Testimony must stay within what the method can support.

What courts have actually done

The National Institute of Justice keeps a database of court decisions after the PCAST report. Its summaries show a clear pattern. Courts usually admit the evidence, but often with limits on what the examiner may say.

  • Firearms and toolmarks. The most contested area. Courts frequently bar claims of 100 percent certainty or a match “to the exclusion of all others.”
  • DNA mixtures. Probabilistic genotyping software, such as STRmix and TrueAllele, is generally admitted, including in the Sixth Circuit in U.S. v. Gissantaner (2021).
  • Latent fingerprints. Mostly admitted, with criticism treated as a matter for cross-examination.
  • Bite marks. The weakest ground. A Pennsylvania appellate court ordered a Frye hearing in one case, while New Jersey’s high court affirmed admission in another.

Texas went furthest. On February 12, 2016, its Forensic Science Commission recommended a moratorium on bite-mark evidence until the technique is scientifically validated, and called for review of past convictions, as the Innocence Project reported.

A case walkthrough: Abruquah v. State

Maryland shows how this plays out. In 2020 the state moved from Frye to Daubert in Rochkind v. Stevenson. Three years later, on June 20, 2023, its Supreme Court decided Abruquah v. State, a shooting case in which a firearms examiner testified that bullets came from the defendant’s gun.

The court found the examiner’s method could support a statement that markings were consistent or inconsistent with a known firearm. It could not support an unqualified claim that the bullets came from that specific gun. The judges pointed to low false-positive rates in closed-set studies alongside high rates of “inconclusive” results in open-set studies, poor repeatability in one study, and no objective standard for separating individual marks from those shared by a batch of guns. The error was not harmless, so the conviction was reversed and sent back for a new trial.

The result was not an outright ban on firearms evidence. It was a narrower script for the witness, which is the same outcome seen across the wider database.

Is the gatekeeping stricter since 2023?

An article in the American Bar Association’s Criminal Justice magazine (Spring 2026) argues that gatekeeping has become more stringent, though it also notes inconsistent application. Its examples show the range. A federal court in Indiana wrote in December 2024 that rejecting expert testimony is the exception, not the rule. A federal court in New York, in December 2024, took the stricter view and found some experts insufficiently grounded. In Cohen v. Cohen, 125 F.4th 454 (3d Cir. 2025), the Third Circuit reversed after a trial court handled four expert challenges in a hearing lasting just over an hour. Read the ABA article for the details.

Lawyers on both sides now cite the 2023 committee note in their briefs, and judges who once waved a challenge through as a question of weight are being asked to explain why. That is a procedural change more than a scientific one. The honest summary: the text of the rule is firmer, and some judges have responded, but outcomes still vary by courtroom.

Common questions

Does Frye ever lead to a different result than Daubert?

Sometimes. Frye asks only about acceptance in the field, so a widely used method can pass even if its error rate is unknown, and a novel but sound one can fail. Commentators often note that the practical difference is smaller than the labels suggest. Pennsylvania illustrates the philosophical gap, where judges are told to leave the science to scientists.

Does the Daubert test apply to non-scientists?

Yes. Kumho Tire extended gatekeeping to technical and experience-based experts, which is why engineers, accountants and appraisers also face reliability challenges.

Can a jury still decide an expert is wrong?

Yes. Passing the gate means the testimony may be heard. Opposing lawyers can still cross-examine, call their own experts and argue that the opinion deserves little weight.

Who has to prove the expert is reliable?

The side offering the expert. Since December 1, 2023, Rule 702 states that burden and the more-likely-than-not standard in its own text.

This article is general information about US law, not legal advice.

Sources