Every summer I catch myself thinking about the same room. It is always a convention center, always over-air-conditioned, and somewhere inside it a few tens of thousands of recent graduates are hunched over laptops trying to remember the elements of adverse possession while their futures hang on the answer. I sat in that room once. Most of my colleagues did too. And for the better part of a century the design of that room barely changed, which is why the arrival of the NextGen bar exam, with its first administration set for July 2026, has produced a level of professional chatter I have not seen since the Uniform Bar Exam debuted in February 2011.
I want to be careful here, because I teach professional responsibility, and the first thing I tell my students is that reform and improvement are not synonyms.
The bar exam is being rebuilt. That much is real. Whether it is being fixed is a separate question, and a harder one, and it happens to be tangled up with a quieter argument playing out in Oregon and Washington and a handful of other jurisdictions that have started to ask whether we need a written closed-universe examination at all.
a test that was already changing before it changed
Start with a little history, because the people selling you novelty rarely bother with it.
The licensing of American lawyers was, for most of the nineteenth century, an apprenticeship affair dressed up with an oral examination. Abraham Lincoln read law. He did not sit a standardized multiple-choice section. The written bar exam as we know it is a creature of the late nineteenth and early twentieth centuries, and the national infrastructure that made it portable is younger still. The National Conference of Bar Examiners introduced the Multistate Bar Examination, that two-hundred-question multiple-choice ordeal, in 1972. The Multistate Essay Examination and the Multistate Performance Test followed. The Uniform Bar Exam, which stitched those components into a single transferable score, was first administered by Missouri and North Dakota in February 2011 and then spread to more than forty jurisdictions.
So the exam has never been a fixed object. It has been a series of compromises between two things the profession has never fully reconciled: the desire to protect the public from incompetent practitioners, and the desire to keep the guild a manageable size.
Those two impulses have collided in the courts more than once, and the collisions are worth remembering. In Schware v. Board of Bar Examiners of New Mexico, 353 U.S. 232 (1957), the Supreme Court held that a state could not deny admission to a qualified applicant on the basis of his former Communist Party membership, because a bar’s gatekeeping had to bear a rational connection to fitness to practice law. On the same day it decided Konigsberg v. State Bar of California, 353 U.S. 252 (1957), pressing the point that character and fitness inquiries could not become instruments of ideological exclusion. Years later, in In re Griffiths, 413 U.S. 717 (1973), the Court struck down a citizenship requirement for admission, and in Baird v. State Bar of Arizona, 401 U.S. 1 (1971), it curbed the questions a bar could demand an applicant answer. The through-line is simple. Every barrier to entry the profession erects is supposed to measure something about the ability to serve clients. When it measures something else, courts have occasionally noticed.
Keep that standard in your pocket as we walk through the new exam, because it is the only standard that matters. Does the thing we test predict the thing we care about?
The NCBE convened a Testing Task Force in 2018 and spent roughly three years producing recommendations. Its conclusion, reduced to a sentence, was that the existing exam over-rewarded memorization and under-rewarded the lawyering that memorization is supposed to enable. That is not a radical finding. Any professor who has watched a student flawlessly recite the rule against perpetuities and then freeze when a live client asks a confused question about a will has felt the gap the Task Force was describing.
what the new exam actually asks
Here is the practical shape of the thing.
The NextGen exam collapses the old three-part structure into an integrated examination administered over roughly nine hours across a day and a half. It mixes item types. There are still multiple-choice questions, but they travel in the company of longer integrated question sets that fold factual investigation, research, analysis, and drafting into a single scenario. The candidate is meant to move the way a junior associate moves: read the file, spot what matters, and produce something a supervising partner could use.
The subject list narrowed, and the narrowing tells you a great deal about the priorities behind it.
The foundational concepts the NCBE chose to keep are the load-bearing walls of general practice: civil procedure, contract law, evidence, torts, constitutional law, criminal law and the constitutional protections around it, business associations, and real property. What fell off the tested list is the part that made faculty lounges argue. Family law is gone. Trusts and estates is gone. Conflict of laws is gone. Secured transactions is gone.
Sit with that for a moment.
The exam is being redesigned around skills, and in the same breath it drops four subjects that describe the actual legal problems ordinary human beings walk into a small firm to solve. A person getting divorced, a family probating a modest estate, a couple fighting over which state’s law governs their dispute: these are not exotic edge cases. They are the daily bread of the solo and small-firm bar that handles most of the individual legal needs in this country. The NCBE’s answer, a reasonable one, is that those subjects are too jurisdiction-specific to test on a national instrument and are better learned in practice. I understand the logic. I also notice that we keep telling new lawyers that the important things will be learned later, and then we act surprised when the profession’s competence problem never resolves.
The stated skills are broader than what the old exam claimed to measure. Legal research. Legal writing. Issue spotting and analysis. Investigation and evaluation of facts. Client counseling and advising. Negotiation and dispute resolution. Client relationship management. On paper this is a description of a lawyer rather than a description of someone who is good at exams about lawyers, and I do not want to be so jaded that I miss the improvement. Testing whether a candidate can advise a confused client is closer to Model Rule 1.4 than anything the old essay ever attempted.
But I teach the rules for a living, so let me be precise about what a written exam can and cannot capture. Competence under Model Rule 1.1 is defined as the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. Comment 8 to that rule, added by the ABA in 2012, extended competence to include keeping abreast of the benefits and risks of relevant technology. You cannot examine thoroughness in nine hours. You cannot examine the discipline it takes to keep abreast of anything. You can examine whether someone can perform competence on a Tuesday, which is not nothing, and is also not the same thing.
The NCBE has said it will continue to offer the current examination during a transition window running into early 2028, so jurisdictions are not being forced off the old instrument overnight. And the Multistate Professional Responsibility Examination, the separate ethics test that has been a fixture since the late 1970s, continues on its own track. My students still have to pass it. I confess a certain institutional affection for the MPRE, though I have never met a practicing lawyer who believed a two-hour multiple-choice test on the Model Rules was what kept him honest.
oregon decided the exam was the wrong question
While the NCBE was rebuilding the test, Oregon quietly decided the test itself might be the problem.
In November 2023 the Oregon Supreme Court approved the Supervised Practice Portfolio Examination, a pathway to licensure that does not run through a bar exam at all. It grew out of a task force the court had charged with examining alternatives, and it became available to graduates in 2024. The structure is disarmingly old-fashioned, which is to say it looks like the apprenticeship the profession abandoned a century ago, cleaned up and given a rubric. A graduate completes a defined block of supervised practice, on the order of six hundred and seventy-five hours, working under a qualified supervising attorney, and assembles a portfolio of real legal work that bar examiners then evaluate.
Think about what that pathway measures. It measures thoroughness, because thoroughness reveals itself over months. It measures the client relationship, because there is an actual client. It measures whether the candidate produced work a supervisor was willing to put a name behind. Those are precisely the qualities the closed-universe exam cannot reach, and Oregon is now the first state to have said so out loud and built a door for it.
Oregon was not writing on a blank slate. Wisconsin has admitted graduates of its two in-state law schools by diploma privilege for generations, without any bar exam, on the theory that a supervised course of study is itself the credential. New Hampshire’s Daniel Webster Scholar Honors Program at the University of New Hampshire, running since 2005, licenses a cohort of students through a two-year performance-based curriculum assessed by bar examiners in lieu of the July exam. These are small programs. They have also produced lawyers for two decades without any evidence that the public suffered for the absence of a July convention hall.
Washington moved next. In March 2024 the Washington Supreme Court approved recommendations from its Bar Licensure Task Force to create alternative pathways to admission, including apprenticeship and experiential routes attached to law school. The details are still being built out, but the direction is unmistakable, and it is the same direction Oregon took: license people by watching them do the work.
I find this movement more interesting than the exam redesign, and I will tell you why in the language of my own field.
The entire structure of professional discipline assumes supervision. Read the cases. In Spaulding v. Zimmerman, 116 N.W.2d 704 (Minn. 1962), a young lawyer’s failure to disclose a life-threatening aneurysm discovered in an opposing party turned on judgment that no exam question could have trained. In Nix v. Whiteside, 475 U.S. 157 (1986), the Court blessed a lawyer’s refusal to help his client commit perjury, a decision made in the pressure of a real trial, not in the calm of a testing center. The ethical lawyer is forged in the supervised handling of actual matters, watched by someone senior enough to say no. A portfolio pathway builds that supervision into licensure itself. A bar exam bolts it on afterward and hopes for the best.
I am not a romantic about apprenticeship. It has a documented history of gatekeeping by connection, of favoring the graduate who already knows a lawyer willing to supervise her, of reproducing the profession’s existing distribution of privilege. A student without a family friend in practice cannot manufacture a supervisor out of thin air. That is a genuine equity problem, and the exam, whatever its faults, at least treats the stranger and the senator’s daughter to the same three questions about future interests. Oregon will have to prove that its supervised pathway does not quietly become a pathway for the already-connected. The early design tried to account for this. Whether it succeeds is an empirical question we will not be able to answer for years.
california reminded everyone what can go wrong
If you want a cautionary tale about the machinery of bar examination, you do not need a hypothetical. California handed everyone one in early 2025.
Facing a budget shortfall, the State Bar of California moved away from the NCBE’s materials and contracted with an outside vendor to deliver a hybrid exam, including a remote testing option, for the February 2025 administration. The administration failed in ways that were not close calls. Candidates could not log in. The software crashed and froze. People lost work. Some reported being unable to save answers or complete sections. The stories that came out of that weekend were the kind that end careers before they start, because a bar exam you cannot take is functionally a bar exam you failed, through no fault of your own.
Then it got worse. In the weeks after, it emerged that some of the multiple-choice questions had been developed with the assistance of artificial intelligence, a disclosure that landed badly with a bar already furious about the technical collapse. The California Supreme Court intervened, scores and remedies were litigated and negotiated, and the whole episode became a national object lesson in what happens when the delivery of a licensing exam is treated as a procurement problem rather than a public trust.
I raise California not to score a cheap point against a struggling bar. I raise it because the NextGen exam is, at bottom, a more complicated instrument to deliver than the one it replaces. Integrated question sets, mixed item types, longer performance components: all of that is harder to administer at scale than a bubble sheet. The NCBE is a far more experienced testing organization than the vendor California hired, and I have no particular reason to predict disaster in July 2026. But the California weekend should have cured anyone of the belief that the hard part of a bar exam is writing the questions. The hard part is the room, the software, the proctoring, and the thousand ways a system can fail a candidate who did everything right.
There is a professional responsibility dimension here that rarely gets named. A licensing system owes candidates something like a duty of competence in return. We demand that they demonstrate they can serve clients reliably. It is fair to demand that the bodies licensing them can administer a reliable examination. California, for one weekend, could not, and no amount of pedagogical sophistication in the questions matters if the delivery collapses.
the training problem the exam was never going to fix
Now let me say the contrarian thing I have been circling.
The NextGen exam is being sold, gently and by implication, as a fix for the profession’s competence complaint. New lawyers, the argument runs, are underprepared for practice; the old exam rewarded the wrong things; a skills-oriented exam will pull legal education toward the right things. There is a theory of change buried in that argument, and the theory is that the exam is the lever that moves the curriculum.
I do not believe it, and I think the history of my own field explains why.
Consider what already happened when the profession decided lawyers needed to understand technology. Comment 8 to Rule 1.1 has instructed lawyers to keep abreast of relevant technology since 2012. More than a dozen years later, we still read discovery-sanctions opinions in which counsel plainly did not understand the systems holding their client’s evidence. The rule changed. The behavior lagged, because a rule is a statement of aspiration and the behavior is produced by training, incentive, and the daily pressure of a caseload. The exam is the same kind of lever, which is to say a weak one. It signals. It does not teach.
What teaches is the first three years of practice, and those three years are governed less by any exam than by the economics of the firms and offices that employ new lawyers. When a legal aid office is drowning and a second-year associate is billing at a rate that requires her to look competent immediately, no redesigned exam reaches her. What reaches her is whether anyone senior has the time and the incentive to supervise. And supervision, real supervision, has been quietly disappearing from the profession for a generation, squeezed out by billing pressure at the top of the market and by resource starvation at the bottom.
This is why the Oregon pathway interests me more than the NCBE’s redesign, even as I worry about its equity. Oregon is at least attacking the actual mechanism by which lawyers become competent, which is supervised repetition, rather than adjusting the proxy we use to guess at competence from the outside.
The exam, old or new, has always been a proxy. It correlates, imperfectly, with something we care about. The Court in Schware demanded that gatekeeping bear a rational relationship to fitness to practice, and a skills-integrated exam probably bears a slightly better relationship than a memorization contest did. I will grant that. It is an improvement in the proxy. It is not a solution to the thing the proxy stands in for.
And there is a subtler risk, one I have watched play out in every reform I have lived through in this profession. When you make the proxy more sophisticated, you make people more confident that it captures the truth. A bubble sheet nobody believed measured lawyering. A rich, skills-based, client-counseling simulation invites everybody to believe it does, and belief is dangerous when it outruns evidence. The most honest thing the NCBE could say about NextGen is that it is a better guess. I am not sure the profession is in the mood for better guesses. It wants a fix, and it will be tempted to call this one.
what I will tell my students in the fall
Come August I will stand in front of a room of students, most of them anxious, some of them already calculating whether their jurisdiction adopts NextGen in July 2026 or waits, and a few of them wondering aloud whether they should relocate to Oregon and skip the exam entirely. I have thought about what I owe them, and it is not reassurance.
I will tell them that the exam is changing in the right direction and for defensible reasons, and that this changes almost nothing about how they will actually become good at this work. I will tell them that the subjects the new exam dropped, family law and trusts and estates and conflicts and secured transactions, are exactly the subjects that describe how most Americans encounter a lawyer, and that a credential which no longer tests them is a credential quietly admitting it never really licensed people to serve those Americans in the first place. I will tell them that if they land somewhere with a supervising attorney who takes the duty seriously, they will learn more in six months than any exam could certify, and that if they do not, no exam will save them.
I will tell them to watch California, and to remember that the profession that demands proof of their competence has not always managed to prove its own.
And I will tell them the thing I believe most firmly, which is the least comfortable to say to a room of people about to pay for a bar review course. The instrument is not the point. It was never the point. A profession decides how good its new members are by how seriously it supervises them in their first years, by whether it pays for that supervision, by whether the senior lawyer with the answer has five minutes for the junior lawyer with the question. Oregon is testing whether we can build that into the front door. The NCBE is testing whether we can measure it a little better at the threshold. One of those is a harder and more honest project than the other, and I think you can tell which one I am rooting for, even as I mark the exam booklets every July and hope the room does not fail anybody who did the work.