Introduction

Every lawsuit filed in federal court raises a question that comes before the merits, before personal jurisdiction, and before venue: does this court have the power to decide this category of dispute at all? That is the question of subject-matter jurisdiction. State courts are courts of general jurisdiction, presumptively open to any civil claim, so a plaintiff suing in state court rarely has to justify the forum's authority. Federal courts sit on the opposite presumption. They may hear only the categories of cases that Article III of the Constitution places within the federal judicial power, and then only to the extent that Congress has enacted a statute actually conferring that power on the district courts. A case can therefore fail at either layer: the Constitution may permit federal adjudication that Congress has never authorized, and a case outside both layers belongs in state court no matter how much the parties would prefer a federal judge.

The consequences of the doctrine are categorical rather than incremental. A judgment entered without subject-matter jurisdiction is an exercise of power the court never possessed, so the defect cannot be waived, forfeited, consented to, or cured by silence, and it can surface at any stage of the litigation — including on appeal, after a full trial, and even at the urging of the party who invoked the federal forum in the first place. This chapter begins with those foundational principles, then works through the two principal statutory grants of original jurisdiction — federal question jurisdiction under 28 U.S.C. § 1331 and diversity jurisdiction under § 1332 — before turning to supplemental jurisdiction over related claims under § 1367 and, finally, to removal and remand, the machinery by which a case moves from state court to federal court and back again.

Foundations of Subject-Matter Jurisdiction

A federal district court may exercise power over a case only when two authorities align. Article III, Section 2 fixes the outer boundary of the federal judicial power by listing the categories of cases to which that power may extend, and a congressional statute must then grant the district courts some portion of that constitutional maximum. Because the power is doubly limited, federal courts operate under a presumption against jurisdiction: the party invoking the federal forum — ordinarily the plaintiff, or a removing defendant — bears the burden of pleading and proving that the case fits within a statutory grant. A defendant who believes the case falls outside those grants raises the objection by motion to dismiss for lack of subject-matter jurisdiction under Rule 12(b)(1)¹. The motion comes in two forms: a facial challenge accepts the complaint's jurisdictional allegations and argues that they are insufficient on their own terms, while a factual challenge disputes the underlying jurisdictional facts — a party's actual domicile, for example — and permits the court to weigh evidence outside the pleadings.

The defining feature of the doctrine is that the objection never expires. Rule 12(h)(3) commands that "[i]f the court determines at any time that it lacks subject-matter jurisdiction, the court must dismiss the action"². The rule is written as a duty of the court rather than a right of a party, and that drafting choice reveals the mechanism at work. Subject-matter jurisdiction polices the structural allocation of judicial power between the state and federal systems rather than protecting any litigant's convenience, so no litigant owns the objection and no litigant can trade it away. Imagine two co-citizens litigating a purely state-law contract dispute who stipulate that the federal court may decide their case: the stipulation is ineffective, because the parties are attempting to confer a power that only the Constitution and Congress can confer. The same structural logic runs in the other direction. Even if neither party ever questions jurisdiction, a district judge who reads the complaint and sees only state-law claims between co-citizens must dismiss on her own motion — sua sponte — because the duty in Rule 12(h)(3) belongs to the court itself³.

That treatment stands in sharp contrast to the neighboring defenses in Rule 12. A defendant waives the defenses of personal jurisdiction, improper venue, insufficient process, and insufficient service of process by omitting them from the first responsive move — the pre-answer motion or the answer⁴. Failure to state a claim occupies a middle position: it may be raised as late as trial, but it too dies with the case⁵. Only subject-matter jurisdiction survives everything short of a final judgment that has become unreviewable. The difference in treatment follows from the difference in whom each defense protects. Personal jurisdiction and venue exist for the defendant's benefit, so the defendant may trade them away, deliberately or by neglect: a defendant who litigates for a year and then objects for the first time to personal jurisdiction has waived the defense, and the case proceeds. The same defendant who litigates for a year and then moves under Rule 12(b)(1) is timely, because the interest at stake — the boundary between two court systems — was never his to waive⁶.

The Seventh Circuit's decision in Belleville Catering Co. v. Champaign Market Place LLC shows how expensive the principle can be in practice. The parties litigated a commercial lease dispute in federal court on the assumption that diversity jurisdiction existed, tried the case to a jury, and obtained a verdict — and only on appeal did anyone confirm the citizenship of the plaintiff limited liability company. An LLC, though formed much like a corporation, is treated as an unincorporated association for diversity purposes and is a citizen of every state of which any member is a citizen, and one of the members shared citizenship with the defendant⁷. The verdict could not stand: the entire proceeding was dismissed for want of jurisdiction, and the parties were left to retry the whole dispute in state court. The lesson generalizes. Because the defect voids everything done in its absence, the jurisdictional inquiry belongs at the very beginning of every federal case, and counsel — like the court — carry an independent obligation to verify jurisdiction rather than assume it.

Federal Question Jurisdiction

The first and most heavily used grant of original jurisdiction is federal question jurisdiction under 28 U.S.C. § 1331, which opens the district courts to civil actions built on federal law. The statutory words are deceptively simple, and nearly all of the difficulty lies in two doctrines that police their boundaries: the well-pleaded complaint rule, which fixes where in the pleadings the federal issue must appear, and the substantiality screen, which fixes how weighty the asserted federal claim must be before its weakness becomes a jurisdictional problem rather than a merits problem. Taken together, the three pieces determine, for any complaint, whether the case "arises under" federal law for § 1331 purposes and why⁸.

The Statutory Grant

Section 1331 grants the district courts "original jurisdiction of all civil actions arising under the Constitution, laws, or treaties of the United States"⁹. The grant mirrors the language of Article III almost word for word, and that resemblance is the source of the section's central interpretive puzzle, because the same words do far less work in the statute than they could. Article III fixes the outer boundary of the federal judicial power; the statute determines how much of that potential jurisdiction the district courts actually exercise, and the courts have construed the statutory phrase narrowly, so that a case can raise genuine federal issues — even issues that will decide the litigation — and still fall outside § 1331¹⁰. The gap has a sensible structural function: it lets Congress calibrate the federal trial docket without touching the Constitution, it preserves state courts as the default forum for most disputes, and it makes the two doctrines taught in the next sections — which fix where the federal issue must appear and how weighty it must be — the working measure of the grant.

Two further features of the grant matter in practice. Section 1331 imposes no amount-in-controversy requirement, so a federal claim worth one dollar enters federal court as easily as one worth millions¹¹. And the grant is original rather than exclusive: unless Congress commits a particular subject to the federal courts alone, as it has with a few areas such as patent infringement, state courts remain fully competent to hear federal claims, which is why a plaintiff with a federal claim ordinarily has a choice of forum. A civil action pleading a claim created by a federal statute, or asserting rights under a treaty of the United States, sits comfortably inside the grant. A pure state-law negligence suit between co-citizens has no federal element at all and sits outside both § 1331 and diversity. The hard cases lie in between — state-law claims with federal law somewhere in the picture — and sorting them is the work of the rule that comes next.

Well-Pleaded Complaint Rule

The narrowing work is done chiefly by the rule of Louisville & Nashville Railroad v. Mottley: a case arises under federal law for § 1331 purposes only if the federal issue appears in the elements of the plaintiff's own properly pleaded claim¹². The Mottleys had settled injury claims against the railroad in exchange for lifetime passes, and after Congress banned free passes the railroad refused to renew them. Their federal-court complaint stated a state-law contract claim for specific performance, and it went on to plead two federal points — that the new statute did not reach pre-existing passes, and that if it did, it took their property without due process. Both federal points, however, belonged to the railroad's anticipated defense and to the reply to that defense, and neither was an element of the contract claim itself. After the case had been fully litigated below, and though no party had questioned jurisdiction, the Supreme Court raised the defect on its own motion and ordered the case dismissed¹³.

The rule's shape follows from what it is for. Testing jurisdiction on the face of the complaint makes the forum determinable at the moment of filing, without speculation about defenses the defendant may or may not actually raise, and it prevents a plaintiff from manufacturing a federal forum by pleading around defenses she expects. The cost is real, and Mottley itself displays it: the only genuinely disputed issues in the case were federal, yet the case did not arise under federal law, which teaches that the presence of federal issues somewhere in the litigation is never enough — position in the pleadings, and only position, controls. What the dismissal takes from a plaintiff is a trial forum rather than a claim: the contract action remains available in state court, where the defendant is free to raise the federal statute in defense, so the rule allocates where the litigation begins and leaves the federal issues to be decided in the ordinary course of that litigation.

Because the rule looks only at the plaintiff's complaint, the defendant's pleadings can never supply the missing federal question. In Holmes Group, Inc. v. Vornado Air Circulation Systems, Inc. the Court held that a counterclaim — even a compulsory counterclaim for federal patent infringement — cannot create arising-under jurisdiction¹⁴. A plaintiff who sues on a state contract claim therefore keeps the case a state case even if the defendant answers with a federal counterclaim. Excluding counterclaims preserves the plaintiff's traditional status as master of the claim and keeps the availability of a federal forum from turning on the defendant's litigation choices. On the affirmative side, the usual measure of a well-pleaded federal claim is creation: where federal law creates the cause of action, federal law necessarily appears in the elements of the plaintiff's own claim, so an employee suing her employer under Title VII, or a patent holder suing for infringement under the patent laws, satisfies the rule without difficulty¹⁵.

Substantiality of the Federal Claim

One question remains: how strong must the asserted federal claim be before it counts? In Bell v. Hood, plaintiffs sued federal agents for damages, claiming that searches and imprisonment violated the Fourth and Fifth Amendments, at a time when no damages remedy for constitutional violations had been recognized. The Court held that jurisdiction existed: when the complaint is drawn to seek recovery directly under the Constitution or laws of the United States, the district court has § 1331 jurisdiction and must decide the case on the merits, and whether the asserted federal right actually supports relief is itself one of those merits questions¹⁶. Jurisdictional dismissal is proper in only two narrow situations — where the federal claim is immaterial and pleaded solely for the purpose of obtaining jurisdiction, or where it is wholly insubstantial and frivolous¹⁷.

The doctrine separates two questions that are easy to run together: whether the court has power to hear the claim, and whether the claim wins. Making jurisdiction turn on what the plaintiff asserts rather than on whether the assertion is correct keeps the threshold inquiry cheap, and it keeps merits questions from being smuggled into jurisdictional rulings, where they would be decided without the procedural protections of a merits judgment and without preclusive effect. So a plaintiff who pleads a novel but non-frivolous theory under a federal statute has invoked the court's jurisdiction, and if the theory fails she loses under Rule 12(b)(6) — a merits ruling entered by a court with power to enter it. The escape valve is deliberately narrow, and it catches only the abuse it was built for: a plaintiff who staples a plainly foreclosed constitutional label onto what is transparently a state-law property dispute, solely to reach federal court, presents a claim that is immaterial or wholly insubstantial, and that dismissal is jurisdictional¹⁸. The single variable separating the two outcomes is whether the federal theory is colorable, and the bar is set low on purpose, because federal trial courts exist in part to define the contours of federal rights, including uncertain ones.

Diversity Jurisdiction

The second principal grant is diversity jurisdiction. Section 1332(a) gives the district courts original jurisdiction over civil actions "where the matter in controversy exceeds the sum or value of $75,000, exclusive of interest and costs," between the party alignments the statute lists¹⁹ — most importantly, citizens of different states²⁰, and citizens of a state on one side with citizens or subjects of a foreign state on the other, the alignment known as alienage jurisdiction²¹. The grant rests on a worry about the home-court advantage: an out-of-state party haled before a local court and a local jury may face bias that has nothing to do with the merits, and a federal forum supplies a neutral alternative. Because federal courts are courts of limited jurisdiction, the statute is administered through bright-line rules — about whose citizenship counts, when it is measured, and how much must be at stake — that keep the threshold inquiry cheap and predictable. The four sections that follow supply those rules in the order a court applies them: identify each party's citizenship, test for complete diversity, fix the moment of measurement, and confirm the amount in controversy.

Complete Diversity Requirement

The statutory diversity-of-citizenship requirement has long been construed to demand complete diversity: no plaintiff may share state citizenship with any defendant²². Running the test requires a citizenship input for every party, and for individuals the input is domicile — the state where the person resides with the intent to remain indefinitely. A person holds one domicile at a time and keeps it until a new one is genuinely established, so a student attending school out of state without any intent to stay remains a citizen of home. The mechanics of the test are then a mapping exercise: place every party's citizenship on its side of the caption and look for a single overlap. Plaintiffs from Ohio and Michigan may sue defendants from Illinois and Indiana, because no plaintiff shares a state with any defendant; substitute a Michigan defendant for the Indiana one and the lone Michigan-to-Michigan pairing destroys jurisdiction over the entire action, however diverse everyone else may be²³.

The all-or-nothing character of the rule tracks the grant's purpose. If any plaintiff shares citizenship with any defendant, then at least one in-state litigant sits on each side of the case, and the risk of local bias that justifies a federal forum is diluted — a local factfinder cannot systematically favor "its" side when its side appears on both ends of the caption. Confining the grant to fully diverse lineups reserves the federal forum for the cases where the outsider-protection rationale is strongest, and it gives courts a mechanical test they can apply at the front end without weighing anything. Alienage jurisdiction follows the same architecture with two refinements written into the statute itself: there is no jurisdiction between a state citizen and a foreign citizen who is a lawful permanent resident domiciled in the same state, an alignment Congress treated as functionally local²⁴, and where citizens of different states are the principal adversaries, foreign citizens may appear as additional parties²⁵.

One structural point should be kept straight. The complete-diversity requirement is a construction of § 1332 rather than a rule spelled out in its text: § 1332(a)(1) says only "citizens of different States," and the all-or-nothing reading is the settled interpretation of those words²⁶. Because the rule lives in statutory construction, it sits within Congress's control, alongside the other dials the statute already turns — the monetary floor of § 1332(a), the alienage refinements described above, and the special citizenship rules the next section takes up²⁷. A Congress persuaded that the diversity docket should be wider or narrower could therefore redraw the complete-diversity line by ordinary amendment to the statute the courts are construing.

Citizenship of Corporations and Associations

Non-human litigants need their own citizenship rules, and the statute gives corporations a special one. Under § 1332(c)(1), a corporation is deemed a citizen of every state and foreign state by which it has been incorporated and of the one state or foreign state where it has its principal place of business²⁸. In Hertz Corp. v. Friend the Court defined the principal place of business as the corporation's "nerve center" — the place where its high-level officers direct, control, and coordinate its activities, which will typically be the headquarters²⁹. The Court chose that test over a rival approach that tallied the bulk of business activities state by state, and it chose it for administrability: a single, easily identified place keeps jurisdictional litigation short and predictable, while an activities census invites a mini-trial at the threshold of every case. The headquarters must be the actual center of direction rather than a mail drop, so the label cannot be gamed³⁰. Dual citizenship itself has a purpose: a company that actually operates from a state cannot claim outsider status there merely because its incorporation papers were filed elsewhere, so the statute closes both routes to a manufactured federal forum.

Unincorporated entities follow an entirely different regime. In Carden v. Arkoma Associates the Court held that a limited partnership is a citizen of every state of which any partner — general or limited — is a citizen, and the same member-citizenship rule extends to LLCs, LLPs, and other unincorporated associations³¹. The entity label controls because Congress wrote the special dual-citizenship rule for corporations alone, and the courts have declined to decide, form by form, which other business structures deserve entity treatment, preferring the bright statutory line to case-by-case judgments. The consequence is that the choice of business form can make or break diversity. A nationwide company incorporated in Delaware and headquartered in New Jersey is a citizen of exactly two states, so a New York plaintiff can sue it in diversity; reorganize the identical business as an LLC with one New York investor among its members, and the same New York plaintiff's suit fails, because the LLC now carries New York citizenship wherever it goes. Belleville Catering, from the opening section, is that trap sprung after a full trial³².

Two further statutory rules assign citizenship to parties who litigate on behalf of someone else. The legal representative of a decedent's estate is deemed a citizen only of the decedent's state, and the representative of an infant or incompetent takes the citizenship of the person represented³³ — a rule that removes any incentive to hand-pick an out-of-state representative for the purpose of creating diversity. And in a direct action against a liability insurer where the insured is not joined as a defendant, the insurer is deemed a citizen of every state of which the insured is a citizen, in addition to its own³⁴, again closing a route by which an essentially local dispute could dress itself up as diverse.

Timing of the Diversity Determination

Diversity is measured once, at the moment the complaint is filed. In Grupo Dataflux v. Atlas Global Group, L.P., a Texas limited partnership sued a Mexican corporation, and because two of the partnership's partners were Mexican citizens, aliens stood on both sides of the case at filing and jurisdiction was absent from the start. The Mexican partners left the partnership before trial, the case was tried, and a jury returned a $750,000 verdict — yet the Court held the original defect incurable, because a post-filing change in a party's citizenship neither creates nor destroys jurisdiction, and the verdict fell³⁵.

The snapshot rule works in both directions, and the matched pair of outcomes differs by a single variable: whether the change happens before or after the filing. Suppose a plaintiff genuinely relocates from Idaho to Oregon with the intent to remain, and then files against an Idaho defendant. Diversity exists at the snapshot, so the case proceeds, and even moving back to Idaho a month later would not defeat jurisdiction that had properly attached³⁶. By contrast, a case that lacked diversity at filing cannot be repaired by anyone's later relocation. The one repair the doctrine allows is a change of parties rather than a change in a party's citizenship: dismissing a dispensable, jurisdiction-spoiling party from the case can save the action, because after the dismissal the judgment runs only between parties who were diverse all along. Grupo Dataflux drew exactly that line — the partnership remained a party throughout, so shedding its partners changed only the citizenship of a continuing litigant and cured nothing³⁷.

The rule earns its rigidity through what it prevents. Measuring jurisdiction once gives courts and litigants a stable answer that does not fluctuate as people move or entities reorganize; without it, parties could manufacture or destroy jurisdiction strategically in mid-case, and every relocation would invite a fresh round of jurisdictional litigation, consuming the very resources a threshold rule exists to conserve. The price, as Grupo Dataflux shows, is an occasional wasted trial, and the Court judged that price lower than the price of a jurisdiction that changes shape as the case proceeds.

Amount in Controversy

The final requirement is monetary: the matter in controversy must exceed $75,000, exclusive of interest and costs³⁸. The word "exceeds" does real work, because a complaint pleading exactly $75,000 fails the test. The plaintiff's good-faith allegation of the amount controls unless it appears to a legal certainty that the claim cannot exceed the threshold. The classic legal-certainty case is a claim under a statute that caps damages below the line — a $150,000 demand on a claim capped at $50,000 cannot support jurisdiction, because the law itself forbids recovery above the threshold no matter what the jury thinks. The rule runs the other way as well: a diverse plaintiff who pleads $150,000 in tort damages in good faith and ultimately recovers $40,000 has lost nothing jurisdictionally, because recovery of less does not retroactively destroy jurisdiction that existed when the case was filed.

The statute polices inflated allegations after the fact rather than up front, and § 1332(b) is the mechanism: a plaintiff who filed in federal court and is finally adjudged entitled to recover less than $75,000, computed without regard to setoffs or counterclaims, may be denied costs and may even have costs imposed on her³⁹. The two-part design is deliberate. Requiring proof of the amount at the pleading stage would spawn a damages mini-trial at the threshold of every diversity case, so the good-faith standard keeps the front-end inquiry cheap, while the cost sanction supplies the deterrent against gaming that a lenient pleading standard would otherwise invite. The monetary floor itself rations the federal forum: local bias may exist in a $5,000 dispute as much as in a $500,000 one, but Congress judged that the cost of running small disputes through federal court outweighs the benefit, and it reserved the grant for higher-stakes cases⁴⁰.

Aggregation rules determine whose dollars count toward the "matter in controversy"⁴¹. In Snyder v. Harris the Supreme Court described the settled interpretation of that phrase: a single plaintiff may aggregate all of her own claims against a single defendant, related or not, while separate plaintiffs may not pool their separate and distinct claims unless they unite to enforce a single title or right in which they share a common and undivided interest. So a lone plaintiff's $50,000 contract claim and unrelated $30,000 tort claim against the same diverse defendant cross the line together, but two crash victims claiming $40,000 each against the same driver do not, because each presents her own controversy and each falls short on her own. The line tracks whose stake is genuinely in dispute: one plaintiff's several claims put that plaintiff's entire demand before the court, while letting strangers stack their small claims would manufacture a federal forum the statute withholds from each of them individually. Even so, a co-plaintiff whose claim falls short is not always shut out of federal court, because — as the next section explains — § 1367 can carry a below-threshold claim that shares a common factual core with a claim that qualifies⁴².

Supplemental Jurisdiction

Supplemental jurisdiction answers a problem the first two grants create. Real disputes rarely sort themselves into jurisdictional categories: the plaintiff with a federal discrimination claim usually also has a state wrongful-discharge claim growing out of the same firing, and neither § 1331 nor § 1332 reaches the state claim on its own. Supplemental jurisdiction lets a federal court that already has original jurisdiction over at least one claim — the anchor — hear related claims that could never have entered federal court by themselves. The doctrine is parasitic by design: it never supplies a freestanding basis for jurisdiction, so every analysis begins by identifying the anchor claim, and a case with no anchor has nothing for the related claims to attach to. The modern analysis runs in three steps, which the sections below take in order: the broad grant of § 1367(a), the diversity carve-out of § 1367(b), and the discretionary off-ramps of § 1367(c).

From Pendent Claims to Common Nucleus

The doctrine began as judge-made law about the meaning of a constitutional "case." In Hurn v. Oursler the Court allowed a state unfair-competition claim to travel with a federal copyright claim where both rested on the same alleged copying of the same play, reasoning that the two theories belonged to a single cause of action⁴³. That cause-of-action test proved cramped and unstable, and in United Mine Workers of America v. Gibbs the Court replaced it with the modern constitutional standard: Article III extends the judicial power to cases rather than to individual claims, so a federal court with a jurisdictionally sufficient federal claim may also hear state claims that derive from a "common nucleus of operative fact" with it — claims a plaintiff would ordinarily be expected to try in one judicial proceeding⁴⁴. In Gibbs itself, a mine superintendent sued a national union under federal labor law for shutting down his mine and added a state tort claim resting on the same campaign of interference, and the shared factual core carried both claims into a single constitutional case⁴⁵.

The shared-facts requirement is what keeps the doctrine honest. Tying the test to a factual core the plaintiff would naturally try in one proceeding ensures that the added claims genuinely belong to the same case rather than free-riding into federal court on an unrelated federal anchor, while sparing litigants the waste and inconsistent outcomes that come from splitting one real-world dispute across two court systems. The line is easy to apply at the extremes: a federal trademark claim and an unrelated landlord-tenant dispute between the same parties share no nucleus, so the state claim stays out, while a federal copyright claim and a state unfair-competition claim resting on the same copying share everything, so the state claim comes in⁴⁶. Gibbs added a second holding that still governs — the power is discretionary, informed by judicial economy, convenience, fairness to litigants, and comity toward state courts⁴⁷. Two older labels still appear in judicial opinions and are worth recognizing: a state claim that a plaintiff attaches to her own federal claim, as Gibbs did, is often called pendent, while reactive claims involving additional parties are called ancillary, and both kinds now travel together under the statute's single label of supplemental jurisdiction⁴⁸.

The judge-made regime carried a legitimacy problem, because federal courts were exercising a jurisdiction no statute had conferred, and the difficulty grew sharpest where supplemental claims would bring entirely new parties into the case. Congress resolved it by enacting 28 U.S.C. § 1367, which put the whole doctrine on a statutory footing and extended the grant expressly to claims involving the joinder or intervention of additional parties⁴⁹. The next section takes up that statute.

The Statutory Grant of Section 1367

Section 1367 works as a two-gate sequence: subsection (a) grants broadly, and subsection (b) subtracts narrowly. The grant provides that in any civil action of which the district courts have original jurisdiction, they shall have supplemental jurisdiction over all other claims so related to the anchor claims that they form part of the same case or controversy under Article III — and its last sentence states expressly that the grant includes claims involving the joinder or intervention of additional parties, the language that settled the old doubt about pendent-party jurisdiction⁵⁰. The same-case-or-controversy standard carries forward the Gibbs test, so the first gate asks the familiar question: do the anchor claim and the added claim grow out of a common nucleus of operative fact⁵¹?

The second gate operates only when original jurisdiction is founded solely on diversity, and it operates only against claims by plaintiffs. In such cases the court has no supplemental jurisdiction over claims by plaintiffs against persons made parties under Rules 14, 19, 20, or 24, or over claims by persons proposed to be joined as plaintiffs under Rule 19 or seeking to intervene as plaintiffs under Rule 24, when exercising jurisdiction would be inconsistent with the requirements of § 1332⁵². The carve-out exists because subsection (a), left unchecked, would gut the complete-diversity rule: a plaintiff could sue the one diverse defendant, wait for the joinder devices to pull the non-diverse players into the case, and then assert the very claims she could never have filed directly. The paradigm is a diversity plaintiff who asserts a claim straight against the non-diverse third-party defendant her opponent impleaded under Rule 14 — § 1367(b) withdraws jurisdiction over exactly that claim⁵³. Whom the subsection leaves alone is equally telling. Defendants' reactive claims — the impleader claim itself, compulsory counterclaims, claims by third-party defendants — survive, because a defendant haled into the forum chose nothing and raises no evasion concern; the carve-out targets plaintiffs, who picked the forum and could otherwise engineer around § 1332.

Exxon Mobil Corp. v. Allapattah Services, Inc. resolved the statute's most contested application: co-plaintiffs whose claims fall short of the amount in controversy. The Court held that where at least one plaintiff's claim satisfies the jurisdictional amount and complete diversity is intact, § 1367 supplies supplemental jurisdiction over the related, below-threshold claims of the additional plaintiffs — including permissively joined plaintiffs and class members⁵⁴. The holding rests on a distinction sometimes called contamination. A non-diverse party poisons the whole action, because her presence eliminates the local-bias rationale that justifies the forum, while a co-plaintiff's low-value claim contaminates nothing — the case remains a genuinely diverse dispute with a smaller claim attached⁵⁵. The statutory text reinforced the result: § 1367(b)'s list of withdrawn claims conspicuously omits claims by plaintiffs joined under Rule 20, and the Court declined to read the omission out of the statute⁵⁶. So a $200,000 claimant and her $50,000 co-plaintiff from the same accident proceed together against the diverse defendant, while two plaintiffs suing one defendant lose the entire action if either of them shares the defendant's citizenship⁵⁷.

Discretion to Decline Supplemental Claims

Satisfying subsections (a) and (b) establishes power, not obligation, and § 1367(c) supplies the grounds on which a district court may nonetheless decline supplemental jurisdiction over a claim:

1. the claim raises a novel or complex issue of state law⁵⁸;
2. the claim substantially predominates over the claim or claims within the court's original jurisdiction⁵⁹;
3. the court has dismissed all claims over which it had original jurisdiction⁶⁰; or
4. in exceptional circumstances, other compelling reasons for declining exist⁶¹.

These grounds are Gibbs's discretionary values — judicial economy, convenience, fairness, and comity — given statutory form⁶², and each names a distinct situation in which power and propriety come apart. The first ground rests on comity: a state's highest court owns the development of its own law, and a federal court that resolves an unsettled question of state law is making a prediction the state's courts never get to correct in that case, so a claim turning on such a question is better decided where the answer can be authoritative⁶³. A supplemental claim that turns on a state constitutional provision the state's highest court has never construed fits that description; a garden-variety negligence claim governed by settled state law does not, however important the claim is to the parties, and the difference between the two is nothing but how settled the governing law is.

The second ground polices proportion rather than novelty. A supplemental claim substantially predominates when the proof it will require, the scope of the issues it raises, or the remedy it seeks dwarfs the federal anchor — the measure is what the litigation will actually consist of, not a count of the claims pleaded — and a case whose real dispute is the state claim is a state case wearing a federal badge, one that belongs in the courts whose law will decide it⁶⁴. Suppose a plaintiff attaches a sprawling state-law fraud case — dozens of witnesses, an accounting of years of transactions, and nearly all of the requested relief — to a single narrow federal claim about one disclosure form; the state claims substantially predominate, and the court may decline to hear them. Shrink the state side to a negligence count resting on the same facts, the same witnesses, and the same measure of damages as the federal anchor, and nothing predominates over anything, so the second ground supplies no basis to decline and the court should ordinarily keep the whole case. The fourth ground is a deliberately narrow residual clause: it demands both exceptional circumstances and compelling reasons, so it cannot become a license to decline whenever the state claims look inconvenient⁶⁵.

The third ground is the workhorse. When the federal anchor dies early — on the pleadings, before any discovery — courts usually send the state claims to state court, because once the anchor is gone little federal interest remains in adjudicating the residue⁶⁶. The ground permits and never compels, though, and the timing of the anchor's dismissal is what moves the answer: a court that dismisses the federal claim on the eve of trial, after three years of litigation, may retain the state claims, because at that point economy and fairness cut hard toward finishing what has nearly been finished. Contrast this discretion with the analysis under subsections (a) and (b), which is mandatory in both directions — a claim outside the grant must be dismissed however convenient retention would be, while a claim within the grant may be declined only on the enumerated grounds, and decline decisions are reviewed for abuse of discretion. Declining is made fair by § 1367(d), which tolls the state limitations period while the supplemental claim is pending in federal court and for at least thirty days after its dismissal, so a plaintiff sent to state court is never time-barred by the round trip⁶⁷.

Removal and Remand

Everything to this point has assumed the plaintiff chose federal court. Removal is the defendant's counterweight to that choice, and it exists because the reasons for a federal forum — protection against local bias and a bench experienced in federal law — apply as strongly to defendants as to plaintiffs. Removal jurisdiction is entirely statutory and entirely derivative: a defendant can remove only a case the plaintiff could have filed in federal court to begin with, so the federal-question and diversity analysis of the earlier sections is the input to every removal question. This section works through the statutes in the order a removed case encounters them — what may be removed under § 1441, how and when under § 1446, and how a case returns to state court under § 1447.

The Right to Remove

Section 1441(a) permits the defendant or defendants to remove any civil action brought in state court of which the district courts of the United States have original jurisdiction, to the district court for the district and division embracing the place where the action is pending⁶⁸. Each clause carries its own limit. The original-jurisdiction condition means the removed case must itself satisfy § 1331 or § 1332: removal shifts the forum without enlarging federal power, so a defendant gains no broader access to federal court than the plaintiff would have had. A New York plaintiff's $500,000 state-law suit against a California corporation is removable because it could have been filed in federal court under diversity; cut the demand to $50,000 and removal fails with the original jurisdiction it depends on⁶⁹. The embracing-district language fixes venue automatically — the case goes to the federal district covering the state courthouse — and the phrase "the defendant or the defendants" means only defendants remove, so a plaintiff facing a federal counterclaim cannot repent of her own forum choice⁷⁰.

Because removability is tested against original jurisdiction, the well-pleaded complaint rule governs here with full force. A defendant whose only federal hook is a federal defense — preemption, for instance — cannot remove, because a complaint pleading only state-law claims does not arise under federal law whatever the answer will say⁷¹, and a defendant's own federal counterclaim can no more support removal than it could support original jurisdiction⁷². The plaintiff thus remains master of the claim at the removal stage as at filing: by pleading only state theories, a plaintiff who also holds federal theories can keep the case at home, though nothing she pleads can erase diversity that actually exists.

Diversity removals carry one additional limit. Under § 1441(b)(2), a civil action removable solely on the basis of diversity may not be removed if any party properly joined and served as a defendant is a citizen of the state where the action was brought⁷³. The rule follows from the grant's purpose: diversity jurisdiction protects outsiders from local prejudice, and a defendant sued at home needs no protection from her own state's courts. The polarity pair therefore turns on geography alone. A Texas plaintiff who sues an Oklahoma defendant in Texas state court on a $200,000 state-law claim faces removal, because the out-of-state defendant is the very party the diversity grant protects; file the identical suit in Oklahoma state court and removal is barred, even though complete diversity and the amount requirement are both satisfied, because the defendant is at home⁷⁴. The bar is specific to diversity — a home-state defendant may still remove a case arising under federal law — and removal itself concedes nothing else, since a removing defendant does not waive objections to personal jurisdiction by making the move.

Removal Procedure and Timing

The mechanics live in § 1446, and they run on a timeline. A removing defendant files in the federal district court a notice of removal, signed under Rule 11 and containing a short and plain statement of the grounds for removal, together with copies of all process, pleadings, and orders served in the state action⁷⁵. The short-and-plain-statement standard deliberately tracks ordinary pleading, so the notice need only plausibly allege the jurisdictional facts — the amount in controversy, for example — and need not prove them with evidence at the outset⁷⁶. The defendant then gives written notice to all adverse parties and files a copy of the notice with the state court clerk, and that filing itself effects the removal: the state court may proceed no further unless and until the case is remanded, and no order of the federal court granting permission is required⁷⁷.

Three timing rules stack on top of the filing requirements. The notice must be filed within thirty days after the defendant receives the initial pleading through service or otherwise⁷⁸. When removal rests solely on § 1441(a), all defendants who have been properly joined and served must join in or consent to the removal — the unanimity rule⁷⁹ — and each defendant gets its own thirty days measured from its own service, so a defendant served six months into the case may still remove within thirty days of that service if the earlier-served defendants consent⁸⁰. A case that was not removable as originally pleaded can become removable later, and a fresh thirty-day window opens when the defendant first ascertains from an amended pleading, motion, order, or other paper that the case has become removable⁸¹. For diversity cases, though, the later-removal route has an outer wall: no removal on diversity grounds more than one year after commencement of the action, unless the district court finds that the plaintiff acted in bad faith to prevent removal⁸².

Each rule answers a distinct temptation. The short deadline forces the forum contest to the front of the case, before either court has invested in the merits, and keeps a defendant from litigating in state court, sizing up the judge, and removing only after things go badly. Unanimity keeps one defendant from dragging unwilling co-defendants out of the plaintiff's chosen forum, while the later-served-defendant rule spares a defendant who never had a fair chance to remove. The one-year cap protects state proceedings from disruption after substantial progress has been made, and its bad-faith exception answers the plaintiff who joins a non-diverse defendant, waits out the year, and then dismisses him for purely tactical reasons — the manipulation the cap would otherwise reward⁸³. So a defendant who files a complete notice twenty-five days after service, with every served co-defendant joining, has removed properly; one who files at day forty-five, or over a co-defendant's refusal, has committed a defect — though, as the next section shows, a defect of a fundamentally different character from a missing grant of jurisdiction.

Remand to State Court

Section 1447(c) sorts remand into two tracks that reprise this chapter's opening theme. A motion to remand on the basis of any defect other than lack of subject-matter jurisdiction — an untimely notice, a missing consent, a forum-state defendant — must be made within thirty days after the filing of the notice of removal, while lack of subject-matter jurisdiction requires remand whenever it appears at any time before final judgment⁸⁴. Procedural defects are therefore waivable, and jurisdictional defects never are. A plaintiff who waits sixty days to object that the notice was filed beyond the thirty-day removal window has forfeited the objection, and the case stays in federal court; the same plaintiff who discovers ten months into the case that the parties are not completely diverse obtains a mandatory remand, because a federal court cannot keep a case Congress never gave it, however late the defect surfaces⁸⁵. After final judgment the procedural track closes entirely, so a purely procedural removal defect raised for the first time after trial supplies no basis for undoing the result⁸⁶.

The two-track design balances the pressures this chapter has been tracing. Efficiency demands that objections to how removal was carried out be raised promptly so the litigation can settle into one forum and proceed; those defects protect orderly procedure, and parties who sleep on them lose them. The jurisdictional track answers to structure rather than efficiency, so it follows the Rule 12(h)(3) principle that the boundary between court systems belongs to no litigant to waive⁸⁷. The thirty-day clock thus does for removal mechanics exactly what Rule 12(h)(1) does for personal jurisdiction and venue — it converts a party-protective objection into something that must be used promptly or lost — while the jurisdictional ground remains, as it was in the opening section, immune to consent, delay, and litigation conduct.

Two final features keep the forum contest short. An order remanding the case may require payment of just costs and actual expenses, including attorney's fees, incurred as a result of the removal, which disciplines objectively unreasonable removals filed for delay: a defendant who removes a plainly non-removable case buys the plaintiff's fees for the detour⁸⁸. And with narrow statutory exceptions, an order remanding a removed case is not reviewable on appeal or otherwise⁸⁹. That bar is a deliberate design choice rather than an oversight: a remand decides only where the case will begin, not who wins it, and Congress chose to end the skirmish over the starting line rather than let appeals about it consume the litigation itself. A defendant who believes the remand was wrong therefore has no recourse — the case simply proceeds in state court, where the federal issues it may contain remain subject to ordinary appellate review at the end of the state proceedings.