delivered the opinion of the Court.
This case requires us to review the grant of a writ of habeas corpus to a state prisoner under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), 28 U. S. C. § 2254(d). The District Court in this case issued the writ to respondent Reginald Lett on the ground that his Michigan murder conviction violated the Double Jeopardy Clause of the Constitution, and the U. S. Court of Appeals for the Sixth Circuit affirmed. In doing so, however, these courts misapplied AEDPA’s deferential standard of review. Because we conclude that the Michigan Supreme Court’s application of federal law was not unreasonable, we reverse.
I
On August 29, 1996, an argument broke out in a Detroit liquor store. The antagonists included Adesoji Latona, a taxi driver; Charles Jones, a passenger who claimed he had been wrongfully ejected from Latona’s cab; and Reginald Lett, a Mend of Jones’s. After the argument began, Lett left the liquor store, retrieved a handgun from another friend outside in the parking lot, and returned to the store. He shot Latona twice, once in the head and once in the chest. Latona died from his wounds shortly thereafter. See People v. Lett, 466 Mich. 206, 208-209, 644 N. W. 2d 743, 745 (2002).
Michigan prosecutors charged Lett with first-degree murder and possession of a firearm during the commission of a *770felony. His trial took place in June 1997. From jury selection to jury instructions the trial took less than nine hours, spread over six different days. Id., at 209, 644 N. W. 2d, at 745.
The jury’s deliberations began on June 12, 1997, at 3:24 p.m., and ran that day until 4 p.m. Id., at 209, n. 1,644 N. W. 2d, at 745, n. 1. After resuming its work the next morning, the jury sent the trial court a note — one of seven it sent out in its two days of deliberations — stating that the jurors had “ ‘a concern about our voice levels disturbing any other proceedings that might be going on.’” Id., at 209, n. 2, 644 N. W. 2d, at 745, n. 2. Later, the jury sent out another note, asking “‘What if we can't agree? [M]istrial? [R]etrial? [W]hat?’ ” Id., at 209, 644 N. W. 2d, at 745.
The trial transcript does not reveal whether the judge discussed the jury’s query with counsel, off the record, upon receiving this last communication. Id., at 209, n. 3,644 N. W. 2d, at 745, n. 3. What is clear is that at 12:45 p.m. the judge called the jury back into the courtroom, along with the prosecutor and defense counsel. Once the jury was seated, the following exchange took place:
“THE COURT: I received your note asking me what if you can’t agree? And I have to conclude from that that that is your situation at this time. So, I’d like to ask the foreperson to identify themselves, please?
“THE FOREPERSON: [Identified herself.]
“THE COURT: Okay, thank you. All right. I need to ask you if the jury is deadlocked; in other words, is there a disagreement as to the verdict?
“THE FOREPERSON: Yes, there is.
“THE COURT: All right. Do you believe that it is hopelessly deadlocked?
“THE FOREPERSON: The majority of us don’t believe that—
“THE COURT: (Interposing) Don’t say what you’re going to say, okay?
*771“THE FOREPERSON: Oh, I’m sorry.
“THE COURT: I don’t want to know what your verdict might be, or how the split is, or any of that. Thank you. Okay? Are you going to reach a unanimous verdict, or not?
“THE FOREPERSON: (No response)
“THE COURT: Yes or no?
“THE FOREPERSON: No, Judge.” Tr. in No. 96-08252 (Recorder’s Court, Detroit, Mich.), pp. 319-320.
The judge then declared a mistrial, dismissed the jury, and scheduled a new trial for later that year. Neither the prosecutor nor Lett’s attorney made any objection.
Lett’s second trial was held before a different judge and jury in November 1997. This time, the jury was able to reach a unanimous verdict — that Lett was guilty of second-degree murder — after deliberating for only 3 hours and 15 minutes. Lett, supra, at 210, and n. 4, 644 N. W. 2d, at 746, and n. 4.
Lett appealed his conviction to the Michigan Court of Appeals. He argued that the judge in his first trial had announced a mistrial without any manifest necessity for doing so. Because the mistrial was an error, Lett maintained, the State was barred by the Double Jeopardy Clause of the U. S. Constitution from trying him a second time. The Michigan Court of Appeals agreed with Lett and reversed his conviction.
The State appealed to the Michigan Supreme Court, which reversed the Court of Appeals. The court explained that under our decision in United States v. Perez, 9 Wheat. 579 (1824), a defendant may be retried following the discharge of a deadlocked jury, even if the discharge occurs without the defendant’s consent. Lett, 466 Mich., at 216-217, 644 N. W. 2d, at 749. There is no Double Jeopardy Clause violation in such circumstances, it noted, so long as the trial court exercised its “ ‘sound discretion’ ” in concluding that the jury was deadlocked and thus that there was a “ ‘manifest necessity’ ” *772for a mistrial. Ibid, (quoting Perez, supra, at 580; emphasis deleted). The court further observed that, under our decision in Arizona v. Washington, 434 U. S. 497, 506-510 (1978), an appellate court must generally defer to a trial judge's determination that a deadlock has been reached. 466 Mich., at 218-222, 644 N. W. 2d, at 750-752.
After setting forth the applicable law, the Michigan Supreme Court determined that the judge at Lett's first trial had not abused her discretion in declaring the mistrial. Id., at 223, 644 N. W. 2d, at 753. The court cited the facts that the jury “had deliberated for at least four hours following a relatively short, and far from complex, trial,” that the jury had sent out several notes, “including one that appears to indicate that its discussions may have been particularly heated,” and — “[m]ost important” — “that the jury foreperson expressly stated that the jury was not going to reach a verdict.” Ibid.
Lett petitioned for a federal writ of habeas corpus. Again he argued that the trial court's declaration of a mistrial constituted an abuse of discretion because there was no manifest necessity to cut short the jury's deliberations. He further contended that the Michigan Supreme Court's rejection of his double jeopardy claim amounted to “an unreasonable application of . . . clearly established Federal law, as determined by the Supreme Court of the United States,” and thus that he was not barred by AEDPA, 28 U. S. C. § 2254(d)(1), from obtaining federal habeas relief. The District Court agreed and granted the writ. 507 F. Supp. 2d 777 (ED Mich. 2007). On appeal, a divided panel of the U. S. Court of Appeals for the Sixth Circuit affirmed. 316 Fed. Appx. 421 (2009). The State petitioned for review in our Court, and we granted certiorari. 558 U. S. 1047 (2009).
II
It is important at the outset to define the question before us. That question is not whether the trial judge should have declared a mistrial. It is not even whether it was an *773abuse of discretion for her to have done so — the applicable standard on direct review. The question under AEDPA is instead whether the determination of the Michigan Supreme Court that there was no abuse of discretion was “an unreasonable application of . . . clearly established Federal law.” §2254(d)(1).
We have explained that “an unreasonable application of federal law is different from an incorrect application of federal law.” Williams v. Taylor, 529 U. S. 362, 410 (2000). Indeed, “a federal habeas court may not issue the writ simply because that court concludes in its independent judgment that the relevant state-court decision applied clearly established federal law erroneously or incorrectly.” Id., at 411. Rather, that application must be “objectively unreasonable.” Id., at 409. This distinction creates “a substantially higher threshold” for obtaining relief than de novo review. Schriro v. Landrigan, 550 U. S. 465, 473 (2007). AEDPA thus imposes a “highly deferential standard for evaluating state-court rulings,” Lindh v. Murphy, 521 U. S. 320, 333, n. 7 (1997), and “demands that state-court decisions be given the benefit of the doubt,” Woodford v. Visciotti, 537 U. S. 19, 24 (2002) (per curiam).1
The “clearly established Federal law” in this area is largely undisputed. In Perez, we held that when a judge discharges a jury on the grounds that the jury cannot reach a verdict, the Double Jeopardy Clause does not bar a new trial for the defendant before a new jury. 9 Wheat., at 579-580. We explained that trial judges may declare a mistrial “whenever, in their opinion, taking all the circumstances into *774consideration, there is a manifest necessity” for doing so. Id., at 580. The decision to declare a mistrial is left to the “sound discretion” of the judge, but “the power ought to be used with the greatest caution, under urgent circumstances, and for very plain and obvious causes.” Ibid.
Since Perez, we have clarified that the “manifest necessity” standard “cannot be interpreted literally,” and that a mistrial is appropriate when there is a “‘high degree’” of necessity. Washington, supra, at 506. The decision whether to grant a mistrial is reserved to the “broad discretion” of the trial judge, a point that “has been consistently reiterated in decisions of this Court.” Illinois v. Somerville, 410 U. S. 458, 462 (1973). See also Gori v. United States, 367 U. S. 364, 368 (1961).
In particular, “[t]he trial judge’s decision to declare a mistrial when he considers the jury deadlocked is . . . accorded great deference by a reviewing court.” Washington, 434 U. S., at 510. A “mistrial premised upon the trial judge’s belief that the jury is unable to reach a verdict [has been] long considered the classic basis for a proper mistrial.” Id., at 509; see also Downturn v. United States, 372 U. S. 734, 736 (1963) (deadlocked jury is the “classic example” of when the State may try the same defendant twice).
The reasons for “allowing the trial judge to exercise broad discretion” are “especially compelling” in eases involving a potentially deadlocked jury. Washington, 434 U. S., at 509. There, the justification for deference is that “the trial eourt is in the best position to assess all the factors which must be considered in making a necessarily discretionary determination whether the jury will be able to reach a just verdict if it continues to deliberate.” Id., at 510, n. 28. In the absence of such deference, trial judges might otherwise “employ coercive means to break the apparent deadlock,” thereby creating a “significant risk that a verdict may result from pressures inherent in the situation rather than the considered judgment of all the jurors.” Id., at 510, 509.
*775This is not to say that we grant absolute deference to trial judges in this context. Perez itself noted that the judge’s exercise of discretion must be “sound,” 9 Wheat., at 580, and we have made clear that “[i]f the record reveals that the trial judge has failed to exercise the ‘sound discretion’ entrusted to him, the reason for such deference by an appellate court disappears,” Washington, 434 U. S., at 510, n. 28. Thus “if the trial judge acts for reasons completely unrelated to the trial problem which purports to be the basis for the mistrial ruling, close appellate scrutiny is appropriate.” Ibid. Similarly, “if a trial judge acts irrationally or irresponsibly,. . . his aetion cannot be condoned.” Id., at 514 (citing United States v. Jorn, 400 U. S. 470 (1971), and Somerville, supra, at 469).
We have expressly declined to require the “mechanical application” of any “rigid formula” when trial judges decide whether jury deadlock warrants a mistrial. Wade v. Hunter, 336 U. S. 684, 691, 690 (1949). We have also explicitly held that a trial judge declaring a mistrial is not required to make explicit findings of “ ‘manifest necessity’ ” nor to “articulate on the record all the factors which informed the deliberate exercise of his discretion.” Washington, supra, at 517. And we have never required a trial judge, before declaring a mistrial based on jury deadlock, to force the jury to deliberate for a minimum period of time, to question the jurors individually, to consult with (or obtain the consent of) either the prosecutor or defense counsel, to issue a supplemental jury instruction, or to consider any other means of breaking the impasse. In 1981, then-justice Rehnquist noted that this Court had never “overturned a trial court’s declaration of a mistrial after a jury was unable to reach a verdict on the ground that the ‘manifest necessity’ standard had not been met.” Winston v. Moore, 452 U. S. 944, 947 (opinion dissenting from denial of eertiorari). The same remains true today, nearly 30 years later.
*776The legal standard applied by the Michigan Supreme Court in this case was whether there was an abuse of the “broad discretion” reserved to the trial judge. Somerville, supra, at 462; Washington, supra, at 509. This type of general standard triggers another consideration under AEDPA. When assessing whether a state court’s application of federal law is unreasonable, “the range of reasonable judgment can depend in part on the nature of the relevant rule” that the state court must apply. Yarborough v. Alvarado, 541 U. S. 652, 664 (2004). Because AEDPA authorizes federal courts to grant relief only when state courts act unreasonably, it follows that “[t]he more general the rule” at issue — and thus the greater the potential for reasoned disagreement among fair-minded judges — “the more leeway [state] courts have in reaching outcomes in case-by-case determinations.” Ibid.) see also Knowles v. Mirzayance, 556 U. S. 111, 123 (2009).
III
In light of all the foregoing, the Michigan Supreme Court’s decision in this case was not unreasonable under AEDPA, and the decision of the Court of Appeals to grant Lett a writ of habeas corpus must be reversed.
The Michigan Supreme Court’s adjudication involved a straightforward application of our longstanding precedents to the facts of Lett’s case. The court cited our own double jeopardy cases — from Perez to Washington — elaborating upon the “manifest necessity” standard for granting a mistrial and noting the broad deference that appellate courts must give trial judges in deciding whether that standard has been met in any given case. Lett, 466 Mich., at 216-222,644 N. W. 2d, at 749-752. It then applied those precedents to the particular facts before it and found no abuse of discretion, especially in light of the length of deliberations after a short and uncomplicated trial, the jury notes suggesting heated discussions and asking what would happen “if we can’t agree,” and — “[m]ost important” — “the fact that the *777jury foreperson expressly stated that the jury was not going to reach a verdict.” Id., at 223, 644 N. W. 2d, at 753. In these circumstances, it was reasonable for the Michigan Supreme Court to determine that the trial judge had exercised sound discretion in declaring a mistrial.
The Court of Appeals for the Sixth Circuit concluded otherwise. It did not contest the Michigan Supreme Court's description of the objective facts, but disagreed with the inferences to be drawn from them. For example, it speculated that the trial judge may have misinterpreted the jury's notes as signs of discord and deadlock when, read literally, they expressly stated no such thing. 316 Fed. Appx., at 427. It further determined that the judge’s brief colloquy with the foreperson may have wrongly implied a false equivalence between “mere disagreement” and “genuine deadlock,” and may have given rise to “inappropriate pressure” on her to say that the jury would be unable to reach a verdict. Id., at 426-427. The trial judge’s mistakes were so egregious, in the Court of Appeals’ view, that the Michigan Supreme Court’s opinion finding no abuse of discretion was not only wrong but objectively unreasonable. Id., at 427.
The Court of Appeals’ interpretation of the trial record is not implausible. Nor, for that matter, is the more inventive (surely not “crude”) speculation of the dissent. Post, at 789. After all, the jury only deliberated for four hours, its notes were arguably ambiguous, the trial judge’s initial question to the foreperson was imprecise, and the judge neither asked for elaboration of the foreperson’s answers nor took any other measures to confirm the foreperson’s prediction that a unanimous verdict would not be reached.2
But other reasonable interpretations of the record are also possible. Lett’s trial was not complex, and there is no reason that the jury would necessarily have needed more than *778a few hours to deliberate over his guilt. The notes the jury sent to the judge eértainly could be read as reflecting substantial disagreement, even if they did not say so outright. Most important, the foreperson expressly told the judge— in response to her unambiguous question “Are you going to reach a unanimous verdict, or not?” — that the jury would be unable to agree. Lett, supra, at 210, 644 N. W. 2d, at 745.
Given the foregoing facts, the Michigan Supreme Court’s decision upholding the trial judge’s exercise of discretion— while not necessarily correct — was not objectively unreasonable.3 Not only are there a number of plausible ways to interpret the record of Lett’s trial, but the standard applied by the Michigan Supreme Court — whether the judge exercised sound discretion — is a general one, to which there is no “plainly correct or incorrect” answer in this case. Yarborough, supra, at 664; see also Knowles, supra, at 128. The Court of Appeals’ ruling in Lett’s favor failed to grant the Michigan courts the dual layers of deference required by AEDPA and our double jeopardy precedents.
The Court of Appeals also erred in a second respect. It relied upon its own decision in Fulton v. Moore, 520 F. 3d 522 (CA6 2008), for the proposition “that Arizona v. Washington sets forth three factors that determine whether a judge has exercised sound discretion in declaring a mistrial: whether the judge (1) heard the opinions of the parties’ counsel about the propriety of the mistrial; (2) considered the alternatives to a mistrial; and (3) acted deliberately, instead *779of abruptly.” 316 Fed. Appx., at 426. It then cited Fulton’s interpretation of Washington to buttress its conclusion that the Michigan Supreme Court erred in concluding that the trial judge had exercised sound discretion. 316 Fed. Appx., at 428.
The Fulton decision, however, does not constitute “clearly established Federal law, as determined by the Supreme Court,” §2254(d)(1), so any failure to apply that decision cannot independently authorize habeas relief under AEDPA. Nor, as the dissent suggests, can Fulton be understood merely to “illuminat[e]” Washington. Post, at 796. Washington nowhere established these three factors as a constitutional test that “determine^]” whether a trial judge has exercised sound discretion in declaring a mistrial. 316 Fed. Appx., at 426.
In concluding that Lett is not entitled to a writ of habeas corpus, we do not deny that the trial judge could have been more thorough before declaring a mistrial. As the Court of Appeals pointed out, id., at 427-428, she could have asked the foreperson additional followup questions, granted additional time for further deliberations, or consulted with the prosecutor and defense counsel before acting. Any of these steps would have been appropriate under the circumstances. None, however, was required — either under our double jeopardy precedents or, by extension, under AEDPA.
* *
AEDPA prevents defendants — and federal courts — from using federal habeas corpus review as a vehicle to second-guess the reasonable decisions of state courts. Whether or not the Michigan Supreme Court’s opinion reinstating Lett’s conviction in this case was correct, it was clearly not unreasonable. The judgment of the Court of Appeals is reversed, and the case is remanded for further proceedings consistent with this opinion.
It is so ordered.
The dissent correctly points out that AEDPA itself “never uses the term ‘deference.’ ” Post, at 797 (opinion of Stevens, J.). But our cases have done so over and over again to describe the effect of the threshold restrictions in 28 U. S. C. § 2254(d) on granting federal habeas relief to state prisoners. See, e. g., Wellons v. Hall, 558 U. S. 220, 223-224, n. 3 (2010) (per curiam); Smith v. Spisak, 558 U. S. 139, 142-143 (2010); McDaniel v. Brown, 558 U. S. 120, 132-133 (2010) (per curiam); Cone v. Bell, 556 U. S. 449, 463, 472 (2009); Knowles v. Mirzayance, 556 U. S. 111, 112, n. 2 (2009); Waddington v. Sarausad, 555 U. S. 179, 194 (2009).
We do not think it reasonable, however, to contend that “the foreperson had no solid basis for estimating the likelihood of deadlock.” Post, at 790. She had, after all, participated in the jury’s deliberations.
It is not necessary for us to decide whether the Michigan Supreme Court’s decision — or, for that matter, the trial judge’s declaration of a mistrial — was right or wrong. The latter question, in particular, is a dose one. As Lett points out, at a hearing before the Michigan Court of Appeals, the state prosecutor expressed the view that the judge had in fact erred in dismissing the jury and declaring a mistrial. The Michigan Supreme Court declined to accept this confession of error, People v. Lett, 463 Mich. 939, 620 N. W. 2d 855 (2000), and in any event — for the reasons we have explained — whether the trial judge was right or wrong is not the pertinent question under AEDPA.