The climate science chapter wouldn’t survive the evidentiary standards the Manual teaches judges to apply
I spend much of my professional life on the wrong end of a Daubert motion.1 As a forensic accountant who testifies as an expert witness in securities class actions and white-collar cases, I know what it means to have a judge test whether my conclusions outrun the evidence. It is uncomfortable, but it is the bedrock of reliable testimony. Which is why it is so troubling that the Reference Manual on Scientific Evidence (RMSE) for federal judges now includes a climate science chapter that would not survive those same standards.
The allegedly withdrawn chapter by the Federal Judicial Center (FJC), which remains available in the National Academies of Science’s (NAS) online edition of the RMSE, was authored by Jessica Wentz and Radley Horton, both affiliated with Columbia University’s Sabin Center for Climate Change Law.2 Their acknowledgments thanked Michael Burger, the Sabin Center’s executive director, who also serves as counsel to Sher Edling, the plaintiffs’ firm representing municipalities in climate liability lawsuits against fossil fuel companies. And Michael Burger’s Sher Edling employment was never disclosed by the FJC, the chapter’s authors, or the NAS, which published the chapter. Former Attorney General William Barr and others have ably documented these conflicts of interest.3 As former Attorney General Barr and three other former U.S. Attorneys General noted, the chapter reflects “a biased take on ‘climate science,’ conceived of and executed by interested parties.” But the conflicts, standing alone, are only the beginning of the problem. The deeper issue is structural, and it runs through the heart of evidence law.
Nobody has yet named the central irony: the Reference Manual exists to teach judges how to apply Daubert v. Merrell Dow Pharmaceuticals and Federal Rule of Evidence 702.4 Those standards require that expert testimony rest on sufficient facts and data, employ reliable methods, and (critically, after the December 2023 amendment to Rule 702) reflect opinions that do not overstate what the methodology can support. The climate chapter fails every one of these criteria, measured by the Manual’s own yardstick.
Start with the Daubert factors themselves. Testability? The chapter’s treatment of attribution science, which attempts to link individual weather events and economic harms to specific energy companies, presents contested probabilistic modeling as settled methodology. Known error rates? Attribution studies rely on counterfactual climate simulations whose uncertainty bounds the chapter systematically minimizes. Peer review? The chapter’s “peer reviewers” included Donald Wuebbles, who has served as a plaintiffs’ expert witness in climate cases, and Kathleen Hall Jamieson, co-founder of FactCheck.org, alongside some of the most liberal judges on the federal bench.5 This is not the kind of independent peer review that Daubert contemplates. It is circular validation, exactly what Daubert was designed to catch.
The 2023 amendment to Rule 702 makes the problem clearer. It targets experts who overstate the certainty of their conclusions beyond what their methodology supports.6 The Advisory Committee Notes specifically call out opinions that “go beyond what the expert’s basis and methodology may reliably support.” If this anti-overstatement principle applies to the individual expert on the witness stand (and it does, as courts are now enforcing with increasing rigor), then it should apply with at least equal force to the authoritative-sounding reference materials that frame how judges evaluate that testimony in the first place. An institutional publication that presents contested science as settled consensus is not merely unhelpful. It is the purest form of overstatement that the 2023 amendment was enacted to prevent.
Consider, too, the analytical gap that General Electric Co. v. Joiner warns against the space between data and conclusion that, when too wide, renders expert testimony unreliable.7 The chapter’s leap from general climate trends to specific, company-attributable harms is precisely the kind of inferential bridge that Joiner instructs courts to examine with care. A forensic accountant who presented damages testimony with this many unsupported inferential steps would be excluded before lunch.
The Daubert problem, though, is only half the story. The other half comes from an unexpected corner of federal law: sentencing.
The federal Sentencing Guidelines were rendered “advisory” by United States v. Booker in 2005.8 Judges are no longer bound to follow them. Yet two decades of empirical research, and the Supreme Court’s own doctrinal admissions, confirm that the Guidelines continue to function as powerful cognitive anchors.9 In Peugh v. United States, the Court held that even advisory Guidelines create a constitutionally significant risk of higher punishment, because the Guidelines remain the “starting point and initial benchmark” for sentencing.10 In Molina-Martinez v. United States, the Court treated an incorrect Guideline calculation as presumptively affecting the outcome, because judges anchor to the range even when they are free to depart from it.11
The behavioral science underlying these decisions is robust and directly applicable here. Guthrie, Rachlinski, and Wistrich tested 167 federal magistrate judges and found them just as susceptible to anchoring bias as laypeople.12 Crucially, subsequent research demonstrates that anchoring is strongest when the anchor comes from a credible, authoritative source. A reference manual published under the joint imprimatur of the Federal Judicial Center and the National Academies of Sciences is about as credible and authoritative as it gets.
This is the sentencing parallel that should alarm anyone who cares about fair adjudication. The Sentencing Guidelines are formally advisory but functionally determinative for a large share of federal sentences. The Reference Manual occupies a similar position: formally educational but functionally quasi-authoritative, cited over 1,300 times by federal and state courts, invoked routinely in Daubert hearings, and treated by judges and clerks as the definitive resource on scientific evidence. When that resource presents one side’s scientific framing as neutral background, it does not merely inform judicial reasoning. It anchors it. The timing makes everything worse. The Supreme Court’s 2024 decision in Loper Bright Enterprises v. Raimondo eliminated Chevron deference, requiring federal judges to independently evaluate scientific and technical questions that agencies previously resolved for them.13 Judges now bear a heavier burden of scientific interpretation than at any point in modern history.14 The standard defense of the Manual, that judges need help understanding complex science, is actually correct. That is precisely why a biased chapter is so dangerous. The need is real; the remedy was corrupted.
Professor Daniel Faigman of UC Hastings recently argued in the Bulletin of the Atomic Scientists that the Manual is essential for judges navigating post-Loper Bright responsibilities.15 He is right about the need but mistaken about the remedy. A reference guide authored by individuals connected to one side of active litigation, reviewed by like-minded contributors, and presented as neutral is not a tool for independent judicial reasoning. It is a thumb on the scale cloaked in institutional authority.
I want to be clear: I am not arguing that climate science is junk science, or that judges should be ignorant of it. I am arguing that the federal judiciary’s most important scientific reference should meet the same evidentiary standards it demands. If a forensic accountant submitted a damages report with conflicts of interest, peer review, and overstatement, it would be excluded under Rule 702 before reaching a jury. The Reference Manual should not get a free pass simply because it carries the FJC’s seal.
The Federal Judicial Center was right to withdraw the chapter. The National Academies must follow suit. And the next edition of the Manual should be produced with the kind of adversarial transparency we demand of every expert who walks into a federal courtroom: balanced authorship, disclosed conflicts, genuine independence in peer review. If the science is strong enough to guide judges, it is strong enough to survive the scrutiny we apply to every other form of expert evidence. Anything less is an invitation to the very institutional bias that Daubert was created to prevent.
Endnotes
1. Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), https://tile.loc.gov/storage-services/service/ll/usrep/usrep509/usrep509579/usrep509579.pdf.
2. Reference Manual on Scientific Evidence, Fourth Edition, Reference Guide on Climate Science (Dec. 31, 2025). The chapter remains available on the National Academies website at https://www.nationalacademies.org/read/26919.
3. William P. Barr, “Federal Judges Rely on This Guide. It Just Took a Sharp Left Turn,” Wash. Post (Mar. 12, 2026), https://www.washingtonpost.com/opinions/2026/03/12/bill-barr-judges-climate-manual-bias/.
4. Fed. R. Evid. 702, as amended effective Dec. 1, 2023, https://www.law.cornell.edu/rules/fre/rule_702.
5. Letter from 27 Republican State Attorneys General to Judge Robin L. Rosenberg, Director, Federal Judicial Center (Jan. 29, 2026), https://ago.wv.gov/sites/default/files/2026-01/2026.01.29%20–%20AG%20Climate%20Science%20Manual%20Letter.pdf.
6. Advisory Committee Notes to 2023 Amendment, Fed. R. Evid. 702. See also Behrens & Trask, “Federal Rule of Evidence 702: A History and Guide to the 2023 Amendments,” 12 Tex. A&M L. Rev. 43 (2024), https://scholarship.law.tamu.edu/lawreview/vol12/iss1/6/.
7. Gen. Elec. Co. v. Joiner, 522 U.S. 136 (1997), https://tile.loc.gov/storage-services/service/ll/usrep/usrep522/usrep522136/usrep522136.pdf.
8. United States v. Booker, 543 U.S. 220 (2005), https://www.law.cornell.edu/supct/html/04-104.ZO1.html.
9. U.S. Sentencing Comm’n, 2024 Annual Report and Sourcebook of Federal Sentencing Statistics, https://www.ussc.gov/sites/default/files/pdf/research-and-publications/annual-reports-and-sourcebooks/2024/2024_Sourcebook.pdf.
10. Peugh v. United States, 569 U.S. 530 (2013), https://tile.loc.gov/storage-services/service/ll/usrep/usrep569/usrep569530/usrep569530.pdf.
11. Molina-Martinez v. United States, 578 U.S. 189 (2016), https://www.law.cornell.edu/supct/pdf/14-8913.pdf.
12. Chris Guthrie, Jeffrey J. Rachlinski & Andrew J. Wistrich, “Inside the Judicial Mind,” 86 Cornell L. Rev. 777 (2001), https://scholarship.law.cornell.edu/facpub/814/.
13. Loper Bright Enterprises v. Raimondo, 603 U.S. 369 (2024), https://www.supremecourt.gov/opinions/23pdf/22-451_7m58.pdf.
14. Sapna Kumar, “Scientific and Technical Expertise After Loper Bright,” 74 Duke L.J. 1749 (2025), https://scholarship.law.duke.edu/dlj/vol74/iss8/2/.
15. Daniel Faigman, Erwin Chemerinsky, Ben Santer & Daniel Nadel, “‘I’m Not a Scientist’: Why Judges Need the Climate Chapter,” Bulletin of the Atomic Scientists (Mar. 11, 2026), https://thebulletin.org/2026/03/im-not-a-scientist-why-judges-need-the-climate-chapter-in-the-reference-manual-on-scientific-evidence/.
Disclaimer: The opinions and views expressed in this article are those of the author and not necessarily those of The National Law Review.
