Substantial Evidence Review Explained: Why Courts Defer to Agency Fact-Finding — and When That Deference Has Limits

A vocational expert testifies at your Social Security disability hearing that there are thousands of jobs someone with your limitations could still perform. She says she based her estimates on private labor market surveys. Your lawyer asks to see the surveys. She refuses — they’re confidential and part of her private business. Your lawyer asks whether the identifying information could be redacted so the data could still be examined. The ALJ interrupts and rules that the expert won’t have to produce the surveys even with confidential information removed. The ALJ then denies your claim for benefits based in part on her testimony. That was Michael Biestek’s case. It reached the Supreme Court in 2019. The question wasn’t whether this particular expert’s testimony was reliable. It was whether testimony like that — where there’s no way to challenge the underlying data — could still qualify as “substantial evidence.” Six justices said yes. The majority, in an opinion by Justice Kagan, rejected a categorical rule that refusing to produce underlying data automatically makes vocational expert testimony insufficient. The sufficiency depends on the circumstances and the record as a whole. Sometimes the refusal might seriously undermine the testimony. Sometimes it doesn’t. The majority also described the substantial evidence threshold as “not high.” Justice Gorsuch dissented and raised the obvious concern — how is a claimant supposed to challenge numbers he can’t see? The answer tells you something important about what substantial evidence review actually is.

This is the third merits-side episode in Season 3’s judicial review arc. The previous two episodes established the law/fact/mixed questions framework and the interpretive toolkit courts use when they exercise independent judgment on questions of law. Substantial evidence review is what happens on the other side of that divide — when the question is factual, not legal. A court generally doesn’t decide the facts for itself. It asks whether substantial evidence supports the agency’s findings. That means a court could disagree with the agency’s conclusion and still have to affirm. Understanding why requires going back to 1951 and the foundational case.

Universal Camera Corp. v. NLRB involved an employee who testified in an NLRB proceeding against his employer and was fired afterward. The Board found he had been fired because of his testimony — a violation of federal labor law. The company said it had fired him over an unrelated workplace dispute. Everyone agreed he was fired; the dispute was over why. The administrative hearing examiner who heard the witnesses recommended dismissing the complaint. The Board disagreed and found the firing unlawful. Congress had provided that the Board’s factual findings would be conclusive “if supported by substantial evidence on the record considered as a whole.” Justice Frankfurter’s opinion gave courts two instructions. First, substantial evidence means “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” That phrase decides a great deal. The court doesn’t have to think the agency reached the best conclusion — just that a reasonable fact finder could reach the conclusion it did. A reviewing judge might believe the evidence supports one interpretation more strongly, but if a reasonable fact finder could reach the agency’s, the court generally must uphold it.

Second — and critically — “the substantiality of evidence must take into account whatever in the record fairly detracts from its weight.” The court can’t ignore the evidence against the agency’s position. If five witnesses testify about why an employee was fired and one supports the Board’s findings while four contradict it, the reviewing court can’t identify the one supporting witness and stop. It has to consider all the testimony and determine whether the record as a whole provides evidence a reasonable mind could accept as adequate. That doesn’t mean counting witnesses — one might be more reliable than four — and it doesn’t mean reweighing each piece of evidence to decide which side the judge personally believes. It means evidence that looks adequate standing alone can become inadequate when considered alongside everything that contradicts it. And the hearing examiner’s contrary findings had to be part of that assessment, because the examiner heard the witnesses firsthand. The Board could disagree with the examiner, but the reviewing court had to weigh that disagreement. Universal Camera requires meaningful scrutiny without authorizing the reviewing court to make the factual findings over again. That’s the balance substantial evidence review has tried to strike ever since.

A terminology point before going further. Substantial evidence isn’t the same as the burden of proof at the original hearing. The burden at the original hearing — in administrative adjudication, nearly always preponderance of the evidence — governs what the original decision-maker must establish before finding a fact. Substantial evidence is what the reviewing court asks about afterward. The reviewing standard doesn’t lower the original burden; the agency still has to meet whatever burden applies. But substantial evidence review may require a court to uphold a finding even if the judge would have weighed the evidence differently. Which brings the analysis back to Biestek. The refusal to produce the surveys was one factor a reviewing court could consider. So were the expert’s qualifications, her explanation of her methods, the opportunities for cross-examination, and the ALJ cutting off the lawyer’s pursuit of redacted surveys. Those details could matter to assessing how thoroughly the testimony was tested. But the refusal to produce the data didn’t automatically settle the question. The court had to consider the testimony in the context of the whole record.

Why defer at all? Three justifications overlap, but their weight varies depending on the kind of finding. First, the original fact-finder may have information the reviewing court doesn’t. When Kathleen testifies about her medical limitations, the ALJ hears her testimony as it happens. The reviewing federal judge ordinarily just sees the written record. That’s an informational difference, similar to why appellate courts defer to trial judges who hear witnesses live. But this can’t fully explain the structure of deference — Universal Camera itself required deference to the Board’s finding even though the hearing examiner, not the Board, had actually observed the witnesses. Firsthand observation matters. It isn’t the whole story.

The bigger justification is institutional responsibility. Congress assigns agencies the job of finding facts in administrative proceedings. Federal courts review the resulting decisions, but they generally aren’t supposed to conduct the entire proceeding again. If every Social Security appeal required a federal judge to hear Kathleen testify, obtain new medical evidence, question a new vocational expert, and independently decide whether she’s disabled, that would be an entirely different system — a second trial rather than judicial review. Instead, the law assigns primary fact-finding responsibility to the agency. The court’s role is to review the resulting decision. The third justification is expertise. Some agency factual findings involve complicated scientific, medical, economic, or technical questions, and a regulatory agency may have considerable experience evaluating them. But not every factual finding involves agency expertise in any deep sense. An ALJ relying on a vocational expert’s labor-market estimates isn’t exercising specialized agency knowledge — the expertise is the witness’s. So the justifications vary: sometimes firsthand observation matters most, sometimes technical expertise does, and sometimes the principal justification is just the institutional decision to make the agency rather than the court the original fact-finder. The same substantial evidence standard can govern all of those findings.

The reach of that standard can depend on the governing statute. In Urias-Orellana v. Bondi (March 2026), the Supreme Court held that the Immigration and Nationality Act’s review provision required substantial evidence review of the agency’s persecution determination — including its application of the accepted legal standard to established facts. Deference reached the step where law gets applied to what happened. That doesn’t automatically generalize into a holding that every mixed question receives substantial evidence review, though most do. The Court was interpreting the specific language Congress provided for reviewing those immigration determinations. The institutional point: the question isn’t just whether an agency is good at finding facts, it’s also what responsibility Congress has assigned and what kind of judicial review Congress has authorized.

Kathleen’s case shows what substantial evidence review looks like on the ground. A terminology shift matters first. In 2016, Social Security removed the term “credibility” from its symptom evaluation guidance and clarified that adjudicators should evaluate the claimant’s statements about symptoms — not their general character or truthfulness. The question is whether the intensity, persistence, and limiting effects of the reported symptoms are consistent with the evidence as a whole. That shift is especially important when the symptom is pain, which can’t be directly measured by a medical test. Adjudicators examine the medical evidence, the treatment history, the daily activities, and other information. Suppose Kathleen testifies she can only stand for 10 minutes before she needs to sit, and the ALJ finds her symptoms aren’t as limiting as she claims because she prepares meals every day and shops for groceries every week. That reasoning doesn’t hold up. She might prepare meals sitting down, or make sandwiches. She might go to the grocery store for 10 minutes and leave, or use an electric cart. Knowing that someone cooks and shops doesn’t tell you how long she can stand. If the ALJ relies on daily activities to discount statements about limitations, the ALJ has to explain why those activities are inconsistent with the claimed limitations. Merely naming an activity doesn’t establish the contradiction. The court can reject that reasoning without deciding for itself how much pain Kathleen is actually experiencing.

Now suppose the record says something different. Eight months before the hearing, Kathleen completed a function report stating she regularly stood for an hour while preparing meals. At the hearing, she says she can’t stand for more than 10 minutes. That’s a more obvious inconsistency, and an ALJ could reasonably consider it. But a reviewing court examining the record as a whole has to consider evidence that undermines the asserted inconsistency: whether her condition worsened during those eight months, whether “an hour” meant continuously or with breaks, whether other evidence explains the difference. The ALJ doesn’t have to accept Kathleen’s explanation, but the reviewing court can’t simply ask whether the ALJ cited something in the record. It has to ask whether the record as a whole provides substantial evidence for the finding — examining the connection between the evidence and the conclusion. If the record genuinely supports two different interpretations, the reviewing court generally has to respect the agency’s reasonable resolution, even if the judge would have reached a different conclusion. The judge might personally believe Kathleen’s explanation and still have to uphold the ALJ’s finding. One more distinction matters here. The ALJ has to adequately explain the findings — a reviewing court needs to be able to understand the basis for the decision. Giving reasons isn’t the same as having adequate evidence. A beautifully written explanation can still rest on a bad inference. If the ALJ doesn’t explain a material finding, the court may need to remand because it can’t determine whether the decision can be sustained. The court can also remand when the ALJ does explain and the reasons don’t hold up against the record. Those are different kinds of defects. Neither gives the reviewing court general authority to conduct the symptom evaluation over again.

Substantial evidence review isn’t the only deferential standard in APA § 706. Under § 706(2)(E), substantial evidence review applies to formal APA proceedings and certain other statutorily-provided on-the-record hearings. Social Security has its own substantial evidence requirement. Arbitrary and capricious review under § 706(2)(A) is generally associated with other agency actions, including informal rulemaking. Both are deferential, but they focus on different things. Substantial evidence review focuses on whether the agency’s factual findings have adequate support in the record. Arbitrary and capricious review also examines whether the agency considered the relevant factors and adequately explained its reasoning. An agency might have support in the record for what it wanted to do but still have failed to consider something it should have — which can send the action back. The two standards overlap substantially, but they arise in different procedural and statutory settings, and the standard tells the reviewing court something about what to examine, not just how much deference to give.

All of this connects to finality. Kathleen didn’t simply lose before the ALJ and immediately go to federal court. She tried to challenge the decision within Social Security first — the Appeals Council could review the decision, issue its own decision, or send the case back to the ALJ for further proceedings. If the Appeals Council denies review, the ALJ’s decision becomes the final decision of the Commissioner. That’s when Kathleen can take the case to federal court. That’s one kind of finality — the agency has reached the point where its decision is ready for judicial review. The second kind is what happens afterward. When the agency’s factual findings survive the available administrative and judicial review and the opportunities for further challenge have ended, the findings become legally conclusive for that proceeding. That doesn’t mean the findings were objectively true. It means the legal system is no longer willing to reconsider them. New information might matter in a subsequent claim, but ordinary judicial review of the original decision isn’t an opportunity to start the fact-finding process over. There’s a difference between saying a factual finding is correct and saying it’s legally over. Legal finality isn’t the same as factual certainty. The system has to decide when a supported factual determination will stand, even though reasonable people might continue to disagree about it. That’s a legitimate legal value — a benefit system can’t operate if every factual determination remains open to unlimited relitigation. The difficult question is what procedures and safeguards justify giving factual findings that degree of legal force, particularly when the process that produced the finding isn’t the same as a civil trial.

Richardson v. Perales (1971) is where the Supreme Court addressed that question directly. Pedro Perales had applied for disability benefits. The agency relied on written medical reports from physicians who had examined him but didn’t appear at the hearing and couldn’t be cross-examined. Perales had opposing evidence including testimony from his own physician. The question was whether those written reports, even though hearsay, could qualify as substantial evidence. The Court held they could — but it didn’t ignore the informality of the administrative proceeding. It examined the safeguards the system provided. The physicians had personally examined Perales, and their reports contained medical findings. The hearing examiner had a responsibility to develop the record. The claimant had procedures available to seek subpoenas for the doctors, allowing him to secure their testimony and cross-examine them. Perales hadn’t requested subpoenas. He had a lawyer. The Court considered his failure to use those available procedures relevant. It concluded that the administrative procedures provided safeguards sufficient to allow those medical reports to constitute substantial evidence.

Perales raises a harder question that it didn’t fully answer. The Court noticed that Social Security hearings differ from adversarial trials — Social Security disability hearings are non-adversarial, with limited discovery and more flexible evidentiary procedures, and the ALJ is employed by the agency administering the benefits program. The Court looked at the procedural safeguards (the ALJ’s duty to develop the record, the opportunity to question witnesses, the possibility of judicial review) and concluded the process could produce evidence sufficient to support the agency’s findings. But identifying safeguards isn’t the same as establishing that they always work. An unrepresented claimant may not know when to request a subpoena or how to use one effectively. A claimant with a serious medical condition may not understand why a particular medical report or vocational estimate matters. The ALJ’s duty to develop the record is real; informality has genuine advantages for claimants who can’t navigate the formal requirements of a civil trial. But informality isn’t all advantage either. For Kathleen, the hearing before the ALJ isn’t a preliminary round before the real decision happens in federal court. It’s where the facts are generally determined. The court can correct errors. It won’t start the process over. The law gives the agency’s factual findings substantial weight, which makes the safeguards at the original hearing enormously important — and makes the question of whether those safeguards actually work as well as the legal system assumes a question that doesn’t go away.

Listen Now

Apple Podcasts | Listen on our site

What We Cover

  • Biestek v. Berryhill (2019) and the 6-3 ruling that a vocational expert’s refusal to produce underlying labor market surveys doesn’t automatically disqualify her testimony as substantial evidence
  • Why the ALJ refused even to require production of redacted surveys — and why that fact matters to a reviewing court’s analysis even though it didn’t change the Supreme Court’s holding
  • The Kagan majority’s description of substantial evidence as a “not high” threshold — and the Gorsuch dissent’s concern about how claimants can challenge numbers they can’t see
  • Universal Camera Corp. v. NLRB (1951) and the foundational framework for substantial evidence review
  • Justice Frankfurter’s two instructions: substantial evidence as what a reasonable mind might accept, and the requirement to consider what fairly detracts from the record
  • Why the court doesn’t ask whether the agency reached the best conclusion — just whether a reasonable fact-finder could reach the conclusion it did
  • Why the reviewing court can’t identify one supporting witness and stop — and why that doesn’t mean counting witnesses either
  • How the Board’s disagreement with the hearing examiner in Universal Camera gets weighed on review
  • The distinction between substantial evidence (a reviewing standard) and the preponderance burden (the original adjudicator’s standard)
  • Why the reviewing standard doesn’t lower the original burden, but may require upholding a finding the judge would have weighed differently
  • Three justifications for deference: firsthand observation, institutional responsibility, and expertise
  • Why institutional responsibility — Congress’s assignment of fact-finding to the agency — is the justification that holds up across every kind of case
  • Why expertise explains some but not all deference, especially where the ALJ is simply relying on outside experts
  • Urias-Orellana v. Bondi (March 2026) and how statutory review provisions decide how far substantial evidence review reaches into mixed questions
  • Social Security’s 2016 shift away from “credibility” determinations toward evaluating consistency of symptom statements with the evidence as a whole
  • Why pain is especially difficult to evaluate — and why daily activities cited by an ALJ need to be explained, not just named
  • The “meal preparation” example: why “she prepares meals and shops for groceries” doesn’t automatically contradict testimony about standing limitations
  • The function-report inconsistency example: a genuine inconsistency the ALJ could reasonably weigh, but one the reviewing court also has to examine in context
  • The distinction between a failure to explain and an explanation that doesn’t hold up against the record — both grounds for remand, but different defects
  • Why giving reasons isn’t the same as having adequate evidence
  • APA § 706(2)(E) substantial evidence review vs. § 706(2)(A) arbitrary and capricious review — and when each applies
  • Why the standards overlap but focus on different questions
  • Finality in two senses: the point at which an agency decision is ready for judicial review, and the point at which factual findings become legally conclusive
  • Why legal finality isn’t the same as factual certainty — and why the system still needs it
  • Richardson v. Perales (1971) and the hearsay medical reports question
  • The safeguards the Court identified in Social Security hearings — the ALJ’s duty to develop the record, available subpoena procedures, judicial review
  • Why identifying safeguards isn’t the same as establishing they always work — the unresolved question Perales opened

Full Rough Transcript

Gwen: Welcome to Administrative Remedies, because you can't fix what you don't understand. Brought to you in part by the University of Tulsa College of Law's Mabee Legal Information Center. I'm Gwendolyn Savitz, an associate professor here at TU.

Marc: And I'm Marc Roark, a professor at the College of Law. We'll be breaking down complex doctrines with real-life analogies and examples to demystify the world of administrative law for everyone trying to understand how government actually works.

Gwen: Agencies are the main way the federal government gets things done. It's not through Congress for reasons we'll be addressing over the course of the series. Marc, suppose the government denies your application for disability benefits because it concludes that there are thousands of jobs you could still perform in the national economy.

Marc: OK, we've seen that play out a couple of times this year.

Gwen: And suppose the government reaches that conclusion based on testimony from a vocational expert. And the expert says she relied on labor market surveys to determine how many jobs were available. And your lawyer asked to see those surveys and the expert refuses.

Marc: That seems a little unfair. I mean, if they're basing judgment on some mysterious survey out there, you should be able to see that.

Gwen: She refuses because she says they're confidential and they're part of her private business.

Marc: But how is the lawyer supposed to challenge her numbers?

Gwen: That is the problem. And that's the problem that reached the Supreme Court in 2019 in Biestek v. Berryhill. Michael Biestek had applied for Social Security disability benefits. And at his hearing, a vocational expert testified that there are jobs someone with his limitations could still perform. And she relied in part on private labor market surveys to produce her estimates. But when Biestek's lawyer asked for the underlying data, she declined to provide the data.

Marc: Couldn't they just remove whatever confidential information was in the surveys and still provide it to the petitioner?

Gwen: That's what Biestek's lawyer asked for. He asked whether the identifying information could be redacted, but the ALJ interrupted and ruled that the expert wouldn't have to produce the surveys, even with confidential information removed.

Marc: So the lawyer actually proposed a way to protect the confidential information, but the ALJ still wouldn't require the expert to produce the evidence.

Gwen: Yes. The ALJ relied on the expert's testimony to deny benefits for an earlier period that Biestek had claimed. He was awarded benefits as of his 50th birthday. If you remember from the Social Security episode, there's a chart and at different ages you qualify for increasing benefits. He tipped the scale at 50 there, but he didn't get it for the earlier period. So the question that reached the Supreme Court was whether testimony like that, where there's no way to challenge the underlying data, could still qualify as substantial evidence.

Marc: I would think the evidence has to be something you can examine. You know, otherwise you're just taking the witness's word for it.

Gwen: Gorsuch expressed a similar concern in the dissent, but the Supreme Court itself rejected some sort of categorical rule that refusing to produce the underlying data automatically makes the testimony insufficient. Kagan wrote for the six-justice majority, and there's an important distinction here. The court wasn't deciding whether this particular expert's testimony was reliable. It was deciding that withholding underlying data doesn't automatically make vocational expert testimony insufficient. The sufficiency of the evidence depends on the circumstances and the record as a whole. And sometimes the refusal to provide data could seriously undermine the testimony, but it doesn't necessarily do so.

Marc: So the testimony can count as substantial evidence, even when the claimant never gets to inspect the data behind it.

Gwen: Yes.

Marc: Okay, that seems like a very low bar, though.

Gwen: This is partly because the Supreme Court described the substantial evidence threshold as not high, but that doesn't mean courts can accept absolutely anything that the agency relies on. And that's what we're going to be talking about today. Last week, we talked about how courts interpret statutes after Loper Bright. Courts exercise independent judgment on questions of statutory meaning. But when a court reviews factual findings in a disability case like Kathleen's, the relationship changes. The court generally doesn't decide the facts for itself. It asks whether substantial evidence supports the agency's findings.

Marc: Which means a court could disagree with the agency and still have to affirm.

Gwen: Yes, exactly. All right. Back to Kathleen, the warehouse supervisor whose disability claim we followed last season. We spent a lot of time on how the agency decides whether someone is disabled, including how it evaluates medical evidence and vocational testimony. Today, we're asking a different question. Once the ALJ has made those factual findings, what can a federal judge do about them? And why would the law give those findings so much weight? Because substantial evidence review isn't just a rule about how much evidence is enough. It determines who has authority to resolve factual disputes, how that authority is reviewed, and when the resulting findings become legally conclusive. To understand the standard, we need to go back to 1951 and Universal Camera Corp. v. NLRB. The facts here are pretty straightforward. We had an employee at Universal Camera who testified in a National Labor Relations Board proceeding involving his employer. And the company later fired him. The board found that he had been fired because of his testimony, which would violate federal labor law. And the company insisted that it had fired him because of a separate workplace dispute.

Marc: The underlying event isn't in dispute. They agree he was fired. They just disagree about why.

Gwen: Right. And here's what makes it interesting. An administrative hearing examiner listened to the evidence and recommended dismissing the complaint. The board disagreed and found the firing was unlawful.

Marc: So the examiner who actually heard the witnesses reached one conclusion and the board reached a different one.

Gwen: Then a federal court had to decide how to weigh the board's factual findings. Congress had provided that the board's factual findings would be conclusive if supported by substantial evidence on the record considered as a whole. What does that mean? Well, Frankfurter's opinion gave two instructions. First, substantial evidence means such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.

Marc: The court doesn't have to think the agency reached the best conclusion. It just has to find that a reasonable person could reach the conclusion it did reach.

Gwen: Yes, that's a crucial distinction. A reviewing judge might think the evidence supports one interpretation more strongly than another, but if a reasonable fact finder could reach the agency's conclusion, the court generally must uphold it. Now, there's another part of Universal Camera that prevents this from becoming a search for just any evidence supporting the agency. Because Frankfurter also said that the substantiality of evidence must take into account whatever in the record fairly detracts from its weight.

Marc: In other words, the court can't ignore the evidence against the agency's position.

Gwen: Yeah. Suppose we have five witnesses and they're testifying about why an employee was fired. One supports the board's findings, four contradict it. The reviewing court can't simply identify the one supporting witness and stop. It has to consider all the testimony and determine whether the record as a whole provides evidence that a reasonable mind could accept as adequate.

Marc: But it also can't just count the witnesses. Maybe the one witness is more reliable than the other four.

Gwen: Sure. This isn't a mathematical formula. The court wasn't telling judges to reweigh every piece of evidence and decide which side they personally believed. It was telling them that evidence can appear adequate when considered by itself, but become inadequate when considered alongside everything that contradicts it. And the hearing examiner's findings were part of that assessment.

Marc: Because the examiner heard the witnesses firsthand.

Gwen: Yeah, the board could disagree with the examiner, but the reviewing court had to consider that disagreement when evaluating the whole record.

Marc: Okay, so Universal Camera requires meaningful scrutiny without giving the reviewing court authority to make the factual findings all over again.

Gwen: Yes, that's the balance. And there's another distinction. Substantial evidence isn't the burden of proof at the original hearing.

Marc: Meaning it isn't the same thing as preponderance of the evidence or beyond a reasonable doubt.

Gwen: Yes. When we're talking administrative law, we're nearly always talking about a preponderance. That governs what the original decision maker must establish before finding a fact. Substantial evidence review is what happens afterwards. It tells a reviewing court when it can disturb the finding that someone already made.

Marc: The agency still has to satisfy whatever burden applies to the original proceeding.

Gwen: Yes. The reviewing standard doesn't lower that burden, but it does mean that a court may have to uphold a resulting finding, even if the judge would have weighed the evidence differently. Let's take that back to Biestek. The refusal to produce surveys was one factor a reviewing court could consider. They could also look at the expert's qualifications, the explanation of her methods, and the opportunities for cross-examination. Those could also be important.

Marc: Okay, including the fact that the ALJ stopped the lawyer from pursuing the redacted surveys.

Gwen: Yes, and that could matter to an assessment of how thoroughly the expert's testimony was tested. But the refusal to produce the surveys doesn't automatically settle the question. The court had to consider the testimony in the context of the whole record. And that brings us to what we're really looking at. Why should a court defer at all? There are a few reasons why we would want deferential factual review. They overlap, but their importance can vary depending on the kind of finding we're talking about. The first is that the original fact finder may have information the reviewing court doesn't.

Marc: Because an ALJ was in the hearing room and the judge simply wasn't.

Gwen: Yes. When Kathleen testifies about her medical limitations, the ALJ heard her testimony as it happened. The reviewing federal judge ordinarily just sees the written record. That's an informational difference.

Marc: Okay. And similar to why appellate courts defer to trial judges who actually hear witnesses.

Gwen: Yes, but Universal Camera shows that this explanation isn't sufficient by itself. The hearing examiner in Universal Camera had actually observed the witnesses, and the board didn't. Yet it was the board's factual finding that the reviewing court was examining, with all the deference it entailed. The Supreme Court required consideration of the examiner's contrary findings, but it didn't make the examiner's view controlling. So firsthand observation matters, but it certainly can't fully explain the structure of deference.

Marc: Okay, what's the second reason then?

Gwen: The big one is institutional responsibility. That Congress assigns agencies the job of finding facts in administrative proceedings. Federal courts review these agency decisions, but they generally aren't supposed to conduct the entire proceeding again.

Marc: Okay, otherwise judicial review would become a second trial.

Gwen: Yeah, imagine if every Social Security appeal required a federal judge to hear Kathleen testify and obtain new medical evidence and question a vocational expert and independently decide whether she's disabled. That's a totally different system. Instead, the law assigns primary fact-finding responsibility to the agency. The court's role is to review the resulting decision.

Marc: Okay, and then the third reason is expertise?

Gwen: Yeah, specialized expertise can matter. Some agency factual findings involve complicated scientific, medical, economic, or technical questions. And a regulatory agency may have considerable experience evaluating those questions. But we shouldn't assume that every factual finding receives deference because the agency possesses superior expertise. The ALJ is relying on vocational evidence from that expert. So justifications vary. Sometimes firsthand observations matter most. Sometimes technical expertise does. And sometimes the principal justification is just the institutional decision to make the agency, rather than the court, the original fact finder.

Marc: But the same substantial evidence standard can govern all of those findings.

Gwen: Yes. And the scope of that deference can also depend on the statute. We previously talked about the Supreme Court case from March 2026, Urias-Orellana v. Bondi. This involved asylum and the question of whether particular facts amount to persecution.

Marc: And even if the underlying events aren't disputed, somebody still has to decide whether they satisfy the legal standard for persecution.

Gwen: Yes, that's a mixed question. It involves the application of law to facts. The court held that the Immigration and Nationality Act's particular review provision required substantial evidence review of the agency's persecution determination, including its application of the accepted legal standard to the facts.

Marc: So deference reaches the step where the law is applied to what happened.

Gwen: Yes, under that statutory provision it does. That doesn't automatically mean that it generalizes to a holding that every mixed question receives substantial evidence review, although most do. The court was interpreting the specific language that Congress provided for reviewing those immigration determinations. And that reinforces the institutional point. The question isn't just whether an agency is good at finding facts. It's also what responsibility Congress has assigned to the agency and what kind of judicial review Congress has authorized. All right, let's go back to Kathleen. Last season, we talked about how disability adjudicators evaluate medical conditions and a claimant's description of symptoms. We also talked about Sarah and Michael, whose similar medical problems led to different outcomes because their limitations were assessed differently. Today, we're looking at what happens when those assessments reach federal court.

Marc: Because the judge isn't going to conduct a new disability hearing.

Gwen: Yes. And I need to make a terminology point first. So in 2016, Social Security removed the term credibility from its symptom evaluation guidance and clarified that adjudicators should evaluate the claimant's statements about symptoms, not their general character or truthfulness. The question is whether the intensity, persistence, and limiting effects of the reported symptoms are consistent with the evidence as a whole.

Marc: Which is especially difficult when the symptom is pain.

Gwen: Yeah, pain isn't something that can be measured directly by a medical test. So adjudicators examine the medical evidence, the treatment history, the daily activities, and other information. But let's suppose Kathleen testifies that she can only stand for 10 minutes before she needs to sit. And the ALJ finds that her symptoms aren't as limiting as she claims. And the explanation just says that she prepares meals every day and shops for groceries every week.

Marc: Okay, why does that contradict her testimony?

Gwen: What do you mean?

Marc: Well, she might prepare meals sitting down. She might make sandwiches. She might go to the grocery store for 10 minutes and then leave. Or she could use an electric cart to get around. I mean, knowing that someone cooks and shops doesn't tell you how long she can stand.

Gwen: Yes, and that's a useful example of what judicial review is supposed to catch. If the ALJ relies on Kathleen's daily activities to discount her statements about her limitations, the ALJ also needs to explain why those activities are inconsistent with the claimed limitations. Merely naming an activity doesn't establish the contradiction.

Marc: The court can reject the ALJ's reasoning without deciding for itself how much pain Kathleen is actually experiencing.

Gwen: Yeah. Now, let's suppose the record says something different. Eight months before the hearing, she completed a function report stating she regularly stood for an hour while preparing meals. And at the hearing, she says she can't stand for more than 10 minutes.

Marc: Okay, that's at least a more obvious inconsistency.

Gwen: It is, and an ALJ could reasonably consider it. But what else would you want to know?

Marc: Well, I'd want to know whether her condition got worse during those eight months.

Gwen: Exactly. And whether she really meant she stood continuously for an hour, whether she did breaks, whether other evidence could explain the difference. The ALJ doesn't necessarily have to accept Kathleen's explanation, but a reviewing court examining the record as a whole has to consider evidence that undermines the asserted inconsistency.

Marc: Which takes us back to Universal Camera.

Gwen: Yes. The reviewing court doesn't simply ask whether the ALJ cited something in the record. It asks whether the record as a whole provides substantial evidence for the finding. That means examining the connection between the evidence and the conclusion.

Marc: But suppose the record genuinely supports two different interpretations.

Gwen: Then the reviewing court generally has to respect the agency's reasonable resolution of that factual dispute, even if the judge would have reached a different conclusion.

Marc: The judge might personally believe Kathleen's explanation, but still have to uphold the ALJ's finding.

Gwen: Yes, and there's another important distinction here. The ALJ has to adequately explain the findings. A reviewing court needs to be able to understand the basis for the decision. But giving reasons isn't the same thing as having adequate evidence.

Marc: Okay, a beautifully written explanation can still rest on a bad inference.

Gwen: Exactly. If the ALJ doesn't explain a material finding, the court may need to remand because it can't determine whether that decision can be sustained. But the court could also remand when the ALJ does explain the findings and the reasons don't hold up against the record. Those are different kinds of defects. And neither gives the reviewing court general authority to conduct the symptom evaluation over again.

Marc: Which makes the original hearing pretty important.

Gwen: Extremely important. The agency's opportunity to hear the testimony matters, but so does the quality of its reasoning and the evidence supporting its findings.

Marc: We've been talking about substantial evidence, but we've also spent time this season discussing arbitrary and capricious review. So how are those two things different?

Gwen: The APA identifies them as distinct grounds for setting aside agency action. Under 706(2)(E), substantial evidence review applies to formal APA proceedings and certain others statutorily provided on-the-record hearings. Social Security, like we said earlier, has its own substantial evidence requirement. And arbitrary and capricious review under 706(2)(A) is generally associated with other agency actions, including informal rulemaking.

Marc: But if substantial evidence and arbitrary and capricious review are both deferential, is there any real meaningful difference?

Gwen: There can be. Substantial evidence review focuses on whether the agency's factual findings have adequate support in the record. Arbitrary and capricious review also examines whether the agency considered the relevant factors and adequately explained its reasoning. An agency might have support in the record for what it wanted to do, but have failed to consider something it should have so it can be sent back to the agency. We will talk about it more next week. Although the two standards can overlap substantially, they arise in different kinds of procedural and statutory settings.

Marc: Okay, the standard tells us something about what the reviewing court is supposed to examine, not just how much deference it owes the agency.

Gwen: Yes, and that brings us back to something we've been talking about throughout this episode, the relationship between judicial deference and the process that produces the agency's findings. Some administrative hearings are highly formal and adversarial. Others, including Social Security disability hearings, are non-adversarial and are using more flexible evidentiary procedures. But courts still apply the same substantial evidence formulation to findings generated by these very different processes.

Marc: Which brings us back to the justification. If the original proceeding wasn't as formal as a federal trial, why should its factual findings receive such substantial deference? The standard doesn't tell us how rigorous the original hearing was.

Gwen: No, and that matters because the judicial review standard helps determine when those factual findings become final. So let's quickly review what finality means. Kathleen didn't simply lose before the ALJ and immediately go to federal court. She tried to challenge the decision within Social Security first.

Marc: Right. She asked the Appeals Council to review what the ALJ had decided.

Gwen: And that's an important step. The Appeals Council can review the decision, can issue its own decision, or send the case back to the ALJ for further proceedings. If the Appeals Council denies review, then the ALJ's decision becomes the final decision of the commissioner.

Marc: That's when Kathleen can take the case to federal court.

Gwen: Yes. And that's what we mean by a final agency decision in this context. The agency has reached the point where its decision is ready for judicial review. But that's different from saying its factual findings can never be reconsidered.

Marc: Because the federal court can still find a problem with them.

Gwen: Right. The court could determine that they lack substantial evidence and send it back to the agency. But it generally won't make new factual findings itself. That brings us to the second kind of finality. When the agency's factual findings survive the available administrative and judicial review, and the opportunities for further challenge have ended, then they become legally conclusive for that proceeding. That doesn't mean the findings were objectively true.

Marc: It means the legal system is no longer willing to reconsider it.

Gwen: Yes, ordinarily, subject to whatever further review or reopening procedures the law might have. Let's think about Kathleen. The ALJ finds she can perform other work, existing in significant numbers in the national economy. She appeals internally. The Appeals Council denies review. Now the agency decision is final and she can challenge it in federal court. The court examines the administrative record and decides that substantial evidence supports the agency's findings. And assuming there's no other reversible error, the court upholds the decision.

Marc: Even if the judge would have interpreted the vocational evidence differently.

Gwen: Exactly. The federal judge doesn't call a new vocational expert and hear testimony again and independently determine how many jobs exist. The original factual determination was the agency's responsibility.

Marc: Okay, what if new information later shows that the vocational expert's numbers were wrong?

Gwen: Depending on the circumstances, later information might matter in some sort of subsequent claim, but there are only limited reopening procedures. It also might apply through statutory procedures governing new material evidence, but ordinary judicial review of that original decision isn't an opportunity to start the fact-finding process over.

Marc: There's a difference between saying a factual finding is correct and saying that it's legally over.

Gwen: Yes. That's one of the important distinctions in administrative law. Legal finality isn't the same thing as factual certainty. The system has to decide when a supported factual determination will stand, even though reasonable people might continue to disagree about it.

Marc: Okay. As a property lawyer, I have some sympathy for that. A lot of property depends on finality. We have recording systems. We have rules governing title, doctrines that sometimes prevent old claims from being reopened. At some point, people need to know and rely on what rights they have.

Gwen: And the same basic problem exists in administrative adjudication. We can't operate a benefit system if every factual determination remains open to unlimited relitigation. Finality is a legitimate legal value. The difficult question is what procedures and safeguards justify giving factual findings that degree of legal force.

Marc: Particularly when the process that produced the finding isn't the same as civil trial.

Gwen: Yeah. Civil litigation can involve discovery. There are formal evidence rules. There's cross-examination. We might have a jury. Administrative proceedings vary considerably. As we've discussed, Social Security disability hearings are non-adversarial. There's limited discovery and more flexible evidentiary procedures. And the ALJ is employed by the agency administering the benefits program.

Marc: Which sounds like it might disadvantage a claimant.

Gwen: Yeah, we talked about it. This can create concerns, but informality isn't necessarily all disadvantage. The ALJ also has the duty to develop the record, and that means that the claimant doesn't have to navigate all the formal requirements of a civil trial. Those features could mean that the process is more accessible.

Marc: But they also mean the safeguards aren't the same.

Gwen: Yes. This brings us to Richardson v. Perales, decided by the Supreme Court in 1971.

Marc: The court actually considered whether this kind of informal hearing could produce evidence sufficient for deferential review.

Gwen: Yes. Pedro Perales had applied for disability benefits. The agency relied on written medical reports from physicians who had examined him. Those doctors didn't appear at the hearing. They weren't able to testify under cross-examination. And Perales had opposing evidence, including testimony from his own physician.

Marc: So the agency had written reports on one side and contrary evidence on the other.

Gwen: Yes, and the question was whether those written reports, even though they were hearsay, could qualify as substantial evidence. And the Supreme Court held that they could. But the court didn't ignore the informality of the administrative proceeding. It examined the safeguards that the system provided. Those physicians had personally examined Perales, and their reports contained medical findings. The hearing examiner has the responsibility for developing the record, and the claimant has procedures available to seek subpoenas for the doctors, allowing them to seek their testimony and cross-examine them.

Marc: Okay. Did he request the subpoenas?

Gwen: No, he didn't. And he had a lawyer. So the court considered his failure to use those available procedures relevant to its analysis.

Marc: The court thought the written reports could be reliable enough despite the absence of live cross-examination.

Gwen: Yes. It concluded that the administrative procedures provided safeguards sufficient to allow those medical reports to constitute substantial evidence.

Marc: But that raises a different question for an unrepresented claimant. How would someone without a lawyer know when to request a subpoena or how to use one effectively?

Gwen: That matters. And the Supreme Court didn't fail to notice that Social Security hearings differ from adversarial trials. It did notice and it looked at the procedural safeguards and concluded that the process could produce evidence sufficient to support the agency's findings. The harder question is just whether those safeguards work in practice as effectively as the court assumed.

Marc: Especially when the claimant has a serious medical condition and may not understand why a particular medical report or vocational estimate matters.

Gwen: Yes, the ALJ's duty to develop the record, the opportunity to question witnesses, and the possibility of judicial review are important protections. But identifying those safeguards doesn't establish that they always work.

Marc: So for Kathleen, the hearing before the ALJ isn't simply a preliminary round before the real decision happens in federal court.

Gwen: Yes, it's where the facts are generally determined. The court can correct errors, but it's not going to start the process over. The law gives the agency's factual findings substantial weight, which makes the safeguards at the original hearing enormously important. And the question is whether those safeguards actually work as well as the legal system assumes. Next week, as I mentioned earlier, we will be looking at arbitrary and capricious review. Today, we've been asking whether the evidence supports the agency's factual findings. Next week, we will broaden the focus on the agency's reasoning: whether it considered the factors the law required, whether it addressed significant issues, and whether it explained how it reached its decision.

Marc: The question becomes not just whether there's evidence supporting a finding, but whether the agency has made a defensible decision based on the things it was supposed to consider.

Gwen: Yes, and we'll see how courts police that reasoning without simply replacing the agency's judgment with their own.

Marc: That's next week. So that does it for today's episode on administrative remedies. Thank you for joining us today. Please, if you enjoy this podcast, enjoy this episode, give us a like on Spotify, iTunes, or whatever platform you're listening on. And be sure to tune in next time where we'll continue to dive into the contours of administrative law, because remember, you can't fix what you don't understand.

Related Guides