Power Divided

A deportation stopped by one chamber of Congress

Jagdish Rai Chadha was born in Kenya, held a British passport, and entered the United States in 1966 on a student visa. The visa ran out in 1972. He conceded that he was deportable, and an immigration judge suspended the deportation under a statute that let the Attorney General do exactly that. Then, in December 1975, the chairman of a House subcommittee introduced a resolution opposing permanent residence for six people, Chadha among them. The House passed it four days later, without printing it for the other members. The Senate never saw it, and neither did the President.

The Supreme Court struck that one-house veto down in INS v. Chadha (opens in a new tab). Congress could have reached the same result. Article I simply required it to legislate, which means passage by both chambers and presentment to the President. So the case turned on procedure rather than on whether Chadha should stay, and that is the pattern this reading is about: who holds a power, and what happens when someone else uses it. The activity has you run six landmark cases through an AI and check its work, which goes faster once you can see which structural question each case was answering.

Power split two ways

The Constitution divides power along two axes, and nearly every structural argument is a fight about one of them.

Federalism is the vertical split, between the national government and the states. Separation of powers is the horizontal split, among Congress, the President, and the courts. Sitting across both is checks and balances, the tools each branch holds over the others: a veto, a confirmation vote, an impeachment, a court order voiding a statute. Division decides who may act. Checking decides who can stop them.

The horizontal split was borrowed. Montesquieu argued in 1748 that liberty disappears when the legislative and executive powers land in the same hands, and that it disappears again when the judicial power joins either one. His chapter on the English constitution (opens in a new tab) is the source the framers worked from. What they built is not a clean three-way partition. Congress sets the courts’ jurisdiction, the President appoints the judges, the Senate confirms them, and the courts read the statutes Congress writes. The branches were made to overlap so that each would have a grip on the others.

Hold that in view when you read the constitutional text (opens in a new tab), because the text rarely announces which principle it serves. Article I, Section 8 reads like a list of grants. It is also a limit, since a power left off the list was never given. Justice Scalia’s foreword (opens in a new tab) to a 2008 symposium makes the case for structure at length, and it is the framing essay for this activity.

How far the national government reaches

Congress has no general power to legislate. It has the powers Article I lists, plus whatever the Necessary and Proper Clause lets it use to carry those out. Maryland tested that in 1819 by taxing the Bank of the United States, pointing out that the Constitution says nothing about a bank. In McCulloch v. Maryland (opens in a new tab) Chief Justice Marshall answered that if the end is legitimate, means “plainly adapted to that end” are constitutional too. Implied powers enter American law there, and so does the argument about how far they run.

The Commerce Clause is where that argument mostly happens. Roscoe Filburn had a wheat allotment of 11.1 acres on his Ohio farm in 1941. He sowed 23, and the extra 239 bushels went to his own livestock and his own table. In Wickard v. Filburn (opens in a new tab) the Court held that Congress could reach even that, because wheat a farmer grows is wheat he does not buy, and enough of it moves the national price. If growing food for yourself counts, it is fair to ask what does not.

For decades after Wickard, almost nothing did. Then a twelfth grader carried a concealed .38 into Edison High School in San Antonio, and the federal charge against him ran into United States v. Lopez (opens in a new tab). The Gun-Free School Zones Act regulated no commercial activity and required no connection to interstate commerce, so the Court held it exceeded the commerce power. NFIB v. Sebelius (opens in a new tab) supplied the other half of the limit in 2012. The clause “is not a general license to regulate an individual from cradle to grave,” so Congress may regulate commercial activity without being able to command it.

Two more doctrines mark the vertical line. The Supremacy Clause makes federal law override conflicting state law. That is how Arizona v. United States (opens in a new tab) knocked out three provisions of that state’s immigration statute in 2012, while leaving the status check during a lawful stop in place for the moment. Anti-commandeering runs the other way. In Printz v. United States (opens in a new tab) two county sheriffs challenged the Brady Act’s demand that they run background checks, and the Court held that Congress may not conscript state officers to administer a federal program. Federal law wins a conflict. It still cannot hand the states the bill for enforcing it.

Which branch may do what

Presidential power is the hardest of these questions, because Article II says a great deal less than Article I does. The framework everyone uses did not come from a majority opinion. It is Justice Jackson’s concurrence in Youngstown Sheet & Tube Co. v. Sawyer (opens in a new tab), the steel seizure case. Hours before a nationwide strike in April 1952, President Truman issued an order directing the Secretary of Commerce to take possession of most of the country’s steel mills and keep them running.

Jackson sorted presidential action into three situations. Acting with authorization from Congress, the President’s power is at its maximum, his own plus whatever Congress can delegate. Acting where Congress has said nothing, he relies on his own powers alone, in what Jackson called a zone of twilight. Acting against the expressed or implied will of Congress, his power is “at its lowest ebb.” The question is never only what the President did. It is what Congress had already said about it.

Congress runs into its own limit going the other direction. It may delegate, and it has to supply a standard when it does. A.L.A. Schechter Poultry Corp. v. United States (opens in a new tab) struck down a New Deal code for the live poultry trade in 1935 as a delegation without one. Mistretta v. United States (opens in a new tab) upheld the Sentencing Guidelines in 1989, because the statute laid down an “intelligible principle” for the commission to follow. That test is easy to state and hard to fail.

The removal power shows doctrine moving in real time. Myers v. United States (opens in a new tab) held in 1926 that a President may remove an executive officer without the Senate’s consent. Nine years later Humphrey’s Executor (opens in a new tab) carved out an exception for commissioners of independent agencies, who could be protected from removal except for cause. Seila Law (opens in a new tab) narrowed that exception in 2020, holding that a single agency director shielded by for-cause protection violates the separation of powers. In 2026 the Court overruled Humphrey’s Executor outright in Trump v. Slaughter. A case can be canonical for ninety years and still stop being law.

Congress faces the same discipline about its own shortcuts, which is where Chadha started. The convenience of a one-chamber veto was real. It was not one of the procedures Article I allows.

The check that makes the limits enforceable

None of this is self-executing. A division of power is a rule, and a rule needs someone to apply it to a case. Judicial review is that someone. In Marbury v. Madison (opens in a new tab) Chief Justice Marshall wrote that it is “the province and duty of the judicial department to say what the law is.” A court handed a statute that conflicts with the Constitution has to choose between them.

That power comes with a limit built into Article III. Courts decide cases, not questions, so someone with a concrete injury has to bring one. Lujan v. Defenders of Wildlife sets out what that takes: an injury in fact, caused by the conduct complained of, and redressable by a court. Structural violations can go unchallenged for years because nobody who can show that kind of injury wants to sue.

Now to the activity. Part 1 has you pick two cases from each of the three groupings and run them through the Case Analyzer, then open the opinion at the Google Scholar link and check the output against it. Use the groupings as a diagnostic while you read. A case about whether Congress had the power at all is federalism. A case about whether the right branch used it is separation of powers. A case about who gets to stop it is checks and balances. Several of these cases sit in two categories at once, and saying which two, and why, is most of the analysis.

One thing to watch for in the AI output. Models sound confident about whether a case is still good law, and that is exactly where they go stale. Humphrey’s Executor stood for ninety years, and a model trained before 2026 will tell you it still stands. Run the Claim Auditor first on at least two of your six, and treat every claim that a case remains good law as something to check rather than something to repeat.