Judge SACK concurs in a separate opinion.
CALABRESI, Circuit Judge.Edward Kruelski appeals the district court’s (Chatigny, J.) denial of his petition for a writ of habeas corpus. Kruelski argues that the Double Jeopardy Clause of the Fifth Amendment of the United States Constitution foreclosed his continued prosecution after the Connecticut trial court, at the close of evidence, acquitted him based on an erroneous interpretation of the applicable statute of limitations. We affirm the district court’s denial of the writ.
I.
Edward Kruelski was charged in Connecticut State Superior Court for the Dan-bury Judicial District with having committed the offense of “offering to make home improvements without being registered,” in violation of section 20-427(b)(5) of the Connecticut General Statutes. Before trial, Kruelski filed a Motion to Dismiss claiming that the charge against him was barred by the relevant statute of limitations, section 54-193 of the Connecticut General Statutes. This motion was left undecided by the trial court.
In May 1995, the case was tried to a jury. During the trial, Kruelski called to the stand a police officer, who testified that although the arrest warrant was signed by a judge on August 22, 1994, two days short of the one-year statute of limitations, it did not come to the attention of a police officer until late on the afternoon of August 25, one day after the year had run, and was not served until that same night. After the close of evidence, Kruel-ski moved for acquittal, arguing both that the State of Connecticut had failed to provide sufficient evidence of the elements of the offense charged and that the State had failed to initiate prosecution within the statute of limitations. The trial court rejected Kruelski’s adequacy of the evidence argument, finding that “[t]he State has introduced evidence sufficient to sustain a conviction in this prosecution for violation of section 20-427(b)(5) of the General Statutes.” But it granted Kruelski’s motion for acquittal based on the statute of limitations. The court was not convinced by the State’s contention that the issuance of an arrest warrant by a judge of the Superior Court satisfied the statute of limitations, and held instead that there must be proof that the appropriate police department had received the warrant by the statutory deadline.
The State appealed to the Connecticut Appellate Court, which reversed the trial court’s decision on the statute of limitations. The Appellate Court ruled that in order to meet the requirements of a Connecticut statute of limitations, an arrest warrant need only be issued within the time limitations and then executed without unreasonable delay. It sent the case back for a new trial. State v. Kruelski, 41 Conn.App. 476, 677 A.2d 951 (1996).
On remand, Kruelski, relying on the Fifth Amendment’s ban on double jeopardy, U.S. Const. amend. V, filed a motion to dismiss the prosecution. Holding that a second trial was permitted under United States v. Scott, 437 U.S. 82, 98 S.Ct. 2187, 57 L.Ed.2d 65 (1978), the trial court denied the motion. Kruelski appealed and the Appellate Court, also relying on Scott, affirmed the trial court's decision. State v. Kruelski, 49 Conn.App. 553, 715 A.2d 796 (1998). Kruelski appealed to the Connecticut Supreme Court, which affirmed the Appellate Court’s decision and its reading *105of Scott. State v. Kruelski, 250 Conn. 1, 737 A.2d 377 (1999).
In July 2000, Kruelski filed an application for a writ of habeas corpus based on Ms double jeopardy argument in the United States District Court for the District- of Connecticut. On May 31, 2001, the district court issued a decision denying the petition. Kruelski v. Connecticut Superior Court, 156 F.Supp.2d 185 (D.Conn.2001). The court held that the state trial court’s entry of a judgment of acquittal did “not constitute an acquittal barring further prosecution,” id. at 188, and that even if it did, the Connecticut Supreme Court’s decision was not an unreasonable application of clearly established federal law under 28 U.S.C. § 2254(d)(1), id. at 190. Kruelski appeals this decision.
II.
We review de novo a district court’s denial of habeas corpus relief. Washington v. Schriver, 255 F.3d 45, 52 (2d Cir.2001); Fama v. Comm’r of Corr. Servs., 235 F.3d 804, 808 (2d Cir.2000); Chalmers v. Mitchell, 73 F.3d 1262, 1266 (2d Cir.1996).
The Anti-Terrorism and Effective Death Penalty Act of 1996 (AEDPA) revised the conditions under which federal courts may grant habeas relief to a person in state custody. 28 U.S.C. § 2254. Among the AEDPA’s new conditions is the requirement that an application for a writ of habeas corpus may be granted only if (1) the state decision “was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” or (2) the state decision “was based on an unreasonable determination of the facts in light of the evidence presented” to the state court. § 2254(d). The facts in this case are not in dispute. Consequently, the writ may be granted only if the Connecticut Supreme Court’s decision was contrary to or an unreasonable application of clearly established federal law. Cf. Williams v. Artuz, 237 F.3d 147, 151-53 (2d Cir.2001).
The concurrence argues that because the AEDPA applies, we should abstain from discussing the correct interpretation of Supreme Court precedent and limit ourselves to the question of whether the Connecticut court’s interpretation of that precedent was a reasonable one. But where reasonable minds can differ on a constitutional question, either because the Supreme Court has not yet addressed the issue or because it has addressed it in a way that leaves room for interpretation as to the constitutional rule, we often have an obligation to inform state courts what we believe the correct answer to be.
Comity, as recognized in the AEDPA, mandates that lower federal courts yield to many state court interpretations of federal law even when such interpretations are wrong, so long as they are reasonable. But just as state courts enjoy a special expertise in matters of state law, by which federal courts often wish to be guided,1 so federal district and *106circuit courts have a particular knowledge of federal law, which state courts faced with federal questions may want to consult. Given the absence of “reverse” certification from state courts to federal courts of appeals and the minimal likelihood of certiorari to the Supreme Court of the United States,2 the only way state courts can do this, should they want to, is if federal courts state their views of federal law, even when a different view is not unreasonable. For this reason, it is often appropriate in considering a habeas petition under the AEDPA for the federal court to go through two steps: first, the court determines what the correct interpretation of Supreme Court precedent is; second, if the state court’s understanding or application of that precedent is determined to be erroneous, the federal court must still ask whether that error was a reasonable one.3
The Supreme Court has stated that federal courts should take a similar approach, for similar reasons, when deciding claims of qualified immunity from suit under 42 U.S.C. § 1983, where the question is the reasonableness of the defendant’s belief that his or her action was constitutional. Saucier v. Katz, 533 U.S. 194, 201, 121 S.Ct. 2151, 150 L.Ed.2d 272 (2001) (“This is the process for the law’s elaboration from case to case, and it is one reason for our insisting upon turning to the existence or nonexistence of a constitutional right as the first inquiry. The law might be deprived of this explanation were a court simply to skip ahead to- the question whether the law clearly established that the officer’s conduct was unlawful in the circumstances of the case.”); see also Poe v. Leonard, 282 F.3d 123, 133 (2d Cir.2002) (“Were we immediately to decide whether [the defendant’s] actions were objectively reasonable, we would fail to provide any guidance to supervisors of future [potential plaintiffs] about what the law requires.”). These considerations apply pari ratione to a federal court’s consideration of state court decisions pursuant to habeas petitions, as governed by the AEDPA.
Under the AEDPA, the result of the first step of the analysis — our conclusions as to the correct interpretation of Supreme Court precedent — will not, of course, be binding on state courts. Mask v. McGinnis, 252 F.3d 85, 90 (2d Cir.2001) (per curiam) (noting that a petitioner cannot win habeas relief solely by demonstrating that the state court unreasonably applied Second Circuit precedent). In fact, what is said in the first step is not binding on federal courts either, since the reasonableness of the state court’s application of Supreme Court precedent — the decisive issue under the AEDPA — does not turn on the federal court’s view of the correct interpretation of that precedent. Cf. Horne v. *107Coughlin, 178 F.3d 603, 604 (2d Cir.1999) (petition for rehearing).
That the federal court’s reading of Supreme Court precedent is dicta does not mean, however, that it is necessarily unreliable. This is not an area in which “the presentation lacks the ‘concrete adverseness ... upon which the court so largely depends for illumination of difficult constitutional questions.’” Id. at 605 (quoting Baker v. Carr, 369 U.S. 186, 204, 82 S.Ct. 691, 7 L.Ed.2d 663 (1962) (discussing standing requirement)). In arguing that a state court’s application of Supreme Court precedent was reasonable or unreasonable, the parties must address the merits of them respective readings of that precedent. And to the extent that there is a danger that arguments will be overlooked, it is minimized by the fact that, as the concurrence rightly notes, our decision as to the correct interpretation of Supreme Court precedent will not bind a future panel of our court that is subsequently presented with an overlooked argument.
The Supreme Court, though undoubtedly conscious of the dangers of dicta, concluded in Saucier that these dangers were small in comparison to the benefits that guidance would give. As a result, it urged the use of a two-step approach to claims of qualified immunity. We conclude that the same is true in the case of habeas petitions under the AEDPA.
The Ninth Circuit, for reasons largely similar to those discussed above, appears to have mandated that appellate courts in that circuit employ a similar two-step approach to habeas petitions under the AED-PA. Tran v. Lindsey, 212 F.3d 1143, 1155 (9th Cir.2000) (“[W]hen analyzing a claim that there has been an unreasonable application of federal law, we must first consider whether the state court erred; only after we have made that determination may we then consider whether any error involved an unreasonable application of controlling law within the meaning of § 2254(d).”). The Fourth Circuit has disagreed and held that federal courts should never venture beyond the reasonableness inquiry. Bell v. Jarvis, 236 F.3d 149, 162 (4th Cir.2000) (“Our charge under the statute is only to determine whether the state court’s adjudication of the claims before it was a reasonable one in light of the controlling Supreme Court law.”). Both positions appear to us to be extreme. We agree with the Ninth Circuit that in many cases governed by the AEDPA, it is appropriate for federal courts first to determine the correct interpretation of Supreme Court precedent. But we decline to hold that federal courts are required to do so. Such a statement is more useful and hence more appropriate in some circumstances than in others.
Again, the Saucier analogy is instructive. While that opinion seems to speak in mandatory terms, we have not so read it. Koch v. Town of Brattleboro, 287 F.3d 162, 166 (2d Cir.2002) (“Although we normally apply this two-step test ... we retain the discretion to refrain from determining whether, under the first step of the test, a constitutional right was violated at all.”). Where, for example, the question is likely to come up with some frequency in cases in which qualified immunity is not an issue, then a statement of the law, in dicta, is much less useful. Id.; see also African Trade & Info. Ctr., Inc. v. Abromaitis, 294 F.3d 355, 359-60 (2d Cir.2002) (deciding whether claimed constitutional right was clearly established without first deciding whether it existed); Vega v. Miller, 273 F.3d 460, 468 (2d Cir.2001) (same). Similarly, when the AEDPA question either has come up or is likely to arise directly in federal cases, the use of dicta for the possible guidance of state courts has much less to be said for it.
*108III.
In the ease before us, clarification of the underlying issue is likely to prove useful to state courts should they wish to be guided by the lower federal courts. Accordingly, we first discuss what the correct rule is for deciding whether the dismissal, on statute of limitations grounds, of the case against Kruelski triggered the protections of the Double Jeopardy Clause. In other words, we first consider whether or not the Connecticut court’s finding that it did not trigger those protections was erroneous. Only after this do we consider whether, even if erroneous, that court’s application of Supreme Court precedent was reasonable.
A.
Two Supreme Court cases guide our consideration of whether a final disposition favorable to the defendant, which is reached after the close of evidence but before a jury verdict, entails that further prosecution constitutes prohibited double jeopardy. The first, Burks v. United States, 437 U.S. 1, 98 S.Ct. 2141, 57 L.Ed.2d 1 (1978), held that the Double Jeopardy Clause barred retrial after an appellate court, having made an incorrect finding that the prosecution had not submitted sufficient evidence to rebut the defendant’s insanity defense, vacated the conviction. Burks contrasted the effect of erroneous terminations based on an insanity defense from equally erroneous terminations based on procedural defects, such as a defective indictment.4 “[R]eversal for trial error, as distinguished from evidentia-ry insufficiency, does not constitute a decision to the effect that the government has failed to prove its case. As such, it implies nothing with respect to the guilt or innocence of the defendant.” 437 U.S. at 15, 98 S.Ct. 2141.
The second governing case, decided on the same day as Burks, is United States v. Scott, 437 U.S. 82, 98 S.Ct. 2187, 57 L.Ed.2d 65 (1978). Scott determined that a trial court’s dismissal of charges at the close of evidence on the basis of preindictment delay did not trigger the prohibitions of the Double Jeopardy Clause. The Court in Scott distinguished Burks, and its reversal for insufficient evidence to rebut an insanity defense, from dismissals for preindictment delay as follows:
The defense of insanity, like the defense of entrapment, arises from the notion that Congress could not have intended criminal punishment for a defendant who has committed all the elements of a proscribed offense where other facts established to the satisfaction of the trier of fact provide a legally adequate justification for otherwise criminal acts. Such a factual finding does necessarily establish the criminal defendant’s lack of criminal culpability under the existing law; the fact that the acquittal may result from erroneous evidentiary rulings or erroneous interpretations of governing legal principles affects the accuracy of that determination, but it does not alter its essential character. By contrast, the dismissal of an indictment for preindictment delay represents a legal *109judgment that a defendant, although criminally culpable, may not be punished because of a supposed constitutional violation.
We think that in a case such as this the defendant, by deliberately choosing to seek termination of the proceedings against him on a basis unrelated to factual guilt or innocence of the offense of which he is accused, suffers no injury cognizable under the Double Jeopardy Clause if the Government is permitted to appeal from such a ruling of the trial court in favor of the defendant.
437 U.S. at 97-99, 98 S.Ct. 2187 (internal quotation marks, citations and footnotes omitted).
Together, Burks and Scott provide pole stars which guide the decision of whether a defendant-favorable termination triggers the protections of the Double Jeopardy Clause. Terminations more akin to insufficiency of evidence to rebut an insanity defense give rise to those protections. Terminations similar to dismissals on the basis of preindictment delay do not.5
Two distinctions support the reasoning of Scott and Burks and explain the different outcomes in the cases. Both are implicit in Scott’s repeated use of the pregnant phrase “factual guilt or innocence.” The first is the distinction between determinations that relate to a defendant’s culpability and those that are merely procedural and do not bear on the defendant’s blameworthiness. The other is the distinction between questions of fact and questions of law. Scott established that an erroneous decision on a matter of law that does not bear on the defendant’s culpability (such as one dealing with preindictment delay) does not trigger the protections of the Double Jeopardy Clause.6 Burks held that a factual determination that does implicate the defendant’s culpability (such as a finding of insufficient evidence to rebut the defendant’s insanity defense) triggers those protections. The questions that the holdings of those cases do not directly decide are, first, whether an acquitting factual error that does not bear on culpability bars further litigation under the Double Jeopardy Clause and, second, whether an acquitting legal error that does go to the question of blameworthiness.precludes such further litigation.
As* to the first, it is a given that our system affords jury findings of fact considerable deference and finality. Consequently, it would seem that a defendant is protected (perhaps even constitutionally) against relitigation of the factual findings on which any termination is based — even if that decision failed to establish the defendant’s nonculpability.7
*110But it would be a mistake to think that only factual determinations trigger the protections of the Double Jeopardy Clause. Contrary to the reading of some commentators,8 this is not the holding of Scott Of equal or greater importance in the Court’s reasoning in that case and in Burks is the question of whether the decision “establish[es] the criminal defendant’s lack of criminal culpability under the existing law,” Scott, 437 U.S. at 98, 98 S.Ct. 2187, that is, whether it “implies [something] with respect to the guilt or innocence of the defendant,” Burks, 437 U.S. at 15, 98 S.Ct. 2141. And in fact, subsequent cases seem to answer the second question left open in Scott by indicating that where a termination based on an erroneous view of the law does go to the defendant’s culpability, the Double Jeopardy Clause bars further prosecution. Arizona v. Rumsey, 467 U.S. 203, 104 S.Ct. 2305, 81 L.Ed.2d 164 (1984) (holding that trial court’s imposition of a life sentence, based on its incorrect finding that, as a matter of law, robbery did not amount to an aggravating circumstance warranting the death penalty, was an acquittal barring further prosecution); Sanabria v. United States, 437 U.S. 54, 98 S.Ct. 2170, 57 L.Ed.2d 43 (1978) (holding that an erroneous evidentiary ruling, which led to termination for insufficient evidence, constituted an acquittal barring further prosecution).
B.
The question before us, therefore, is this: Did the trial court’s acquittal on the statute of limitations defense—which was an error of law—mean that the state had tried but failed to prove that Kruelski was culpable for the crimes charged? Since Kruelski was indicted for violating Connecticut criminal law, the proper course in answering this question is to look first to the intent of the Connecticut legislature, as this has been interpreted by Connecticut courts.
The Connecticut Supreme Court, in ruling on Kruelski’s double jeopardy claim, understood the issue in precisely the way we have described: “[W]e must determine whether the trial court’s judgment was based on legal grounds unrelated to a determination of the sufficiency of the evidence regarding the defendant’s factual innocence or guilt.” 737 A.2d at 380. The Connecticut High Court held that further prosecution was not barred, because under Connecticut law “[a] statute of limitations defense is not a defense that, when proved, negates any element of a charged offense or establishes a legal justification for an otherwise criminal act. Instead, this type of defense ‘represents a legal judgment that a defendant, although criminally culpable, may not be punished....’” Id. at 382 (quoting Scott, 437 U.S. at 98, 98 S.Ct. 2187). Connecticut, therefore, has determined that the expiration of its one-year statute of limitations governing the misdemeanor of offering to perform home improvements without being registered does not mean that a defendant is not culpable for the violation. It only provides that the *111defendant can no longer be prosecuted for it.9
This was not the only possible decision that the Connecticut Supreme Court could have reached. While statutes of limitations are often “intended to foreclose the potential for inaccuracy and unfairness that stale evidence and dull memories may occasion,” id. at 380 (quoting State v. Coleman, 202 Conn. 86, 519 A.2d 1201, 1204 (1987)), this need not be the only purpose they serve. Thus, it would be perfectly reasonable for a legislature to find that some forms of criminal culpability do not endure indefinitely and that after a certain amount of time prosecution is inappropriate, not because of the danger of error but because it would be wrong to punish a defendant for acts done in the distant past, acts as to which blame no longer persisted.10 Where a statute of limitations is based on reasoning of this sort, acquittal on a limitations defense might well be a determination of the defendant’s lack of continued culpability and hence might bar further prosecution. The Connecticut Supreme Court’s reasonable interpretation of the statute of limitations at issue in this case, however, means that this is not the situation before us.
We do not, of course, hold that, whenever a state legislature or court says that the grounds for a defendant-favorable determination do not concern the defendant’s culpability, such a statement decides the issue for the purposes of the federal Double Jeopardy Clause. That clause is necessarily governed by federal standards of culpability. This being the case, we doubt whether a state’s finding that, say, the insanity defense did not concern the defendant’s culpability would be determinative of whether further prosecution was allowed, In this case, however, the Connecticut Supreme Court’s conclusion — that the trial court’s incorrect finding that the statute of limitations had been exceeded by one day was not a determination that went to Kruelski’s culpability — is clearly within the bounds of the permissible under Scott and under Burks.
Given the Connecticut Supreme Court’s reasonable determination that the trial court’s acquittal on the statute of limitations did not bear on the defendant’s culpability, its decision is hard to fault under Scott. Certainly the Connecticut ruling was neither contrary to nor an unreasonable application of clearly established federal law, as § 2254(d) requires for a grant of a writ of habeas corpus on behalf of a person in custody pursuant to the judgment of a state court.
Accordingly, we AffiRM the district court’s judgment denying the writ.
. Such guidance is most frequently obtained by certifying a question of state law to the state's highest court. See, e.g., Cweklinsky v. Mobil Chemical Co., 297 F.3d 154, 160 (2d Cir.2002) ("[T]he principles of federalism and comity demand that federal courts give a state’s highest court the opportunity to determine state law authoritatively, if it wishes to do so.”); Allstate Ins. Co. v. Serio, 261 F.3d 143, 152 (2d Cir.2001) ("It is possible ... that under, the applicable state law, state courts have more flexibility and broader interpretive power than do the federal courts. In such circumstances, federal courts ought not to deprive the state courts of the opportunity to construe their own statutes, using the interpretive tools, presumptions, and standards they deem proper. For this too would in*106fringe on the sovereign authority of the states.”); Israel v. State Farm Mut. Auto. Ins., 239 F.3d 127, 135 (2d Cir.2000) (concluding that "Connecticut has a strong interest in deciding the issues certified rather than having the only precedent on point be that of the federal court, which may be mistaken.” (internal quotation marks and alterations omitted)).
. While there were 7,924 filings to the Supreme Court in the 2001 Term, the Court heard argument in only 88 cases. REHNQUIST, C. J., 2002 Year-End Report on the Federal Judiciary 8 (Jan. 1, 2003).
. While this two-step approach is appropriate in the federal district courts no less than in the federal courts of appeal, it is not mandatory in either. See infra. It is, moreover, quite understandable if a district court, burdened with a heavy docket, should prefer not to go through such a process and rule directly on the reasonableness of the state court's approach, leaving any future guidance solely to the courts of appeals.
. The Court fastened onto the example of a defective indictment in order to distinguish Ball v. United States, 163 U.S. 662, 16 S.Ct. 1192, 41 L.Ed. 300 (1896). In doing so, Burks also explicitly rejected the approach of an intervening line of cases, which had established the principle that ‘‘[a] defendant who requests a new trial as one avenue of relief may be required lo stand trial again, even when his conviction was reversed due to failure of proof at the first trial.” 437 U.S. at 10, 98 S.Ct. 2141. The Buries Court found that these cases had “not properly construed the [Double Jeopardy] Clause.” Id. at 12, 98 S.Ct. 2141.
. This question does not turn on how the terminating court characterizes its own decision — whether, for example, it uses the word “acquittal," as opposed to "dismissal.” "[T]he trial judge's characterization of his own action cannot control the classification of the action.” Scott, 437 U.S. at 96, 98 S.Ct. 2187 (quoting United States v. Jorn, 400 U.S. 470, 478 n. 7, 91 S.Ct. 547, 27 L.Ed.2d 543 (1971) (opinion of Harlan, J.)).
. This remains so even where a legally erroneous decision also depended on factual findings that are themselves immune from relitigation. The question is not whether the terminating decision rested in part on a finding of fact, but whether the error was a factual one or a legal one. In the case before us, therefore, it is immaterial that the trial judge's erroneous decision — that the statute of limitations barred prosecution because there was no proof that the warrant was received by the police department by the statutory deadline — rested in part on an unchallenged jury finding as to when the police first received the warrant. The error was purely legal.
.Cf. U.S. Const. amend. VII (establishing that for most suits at common law, "no fact tried by a jury, shall be otherwise reexamined in any Court of the United States, than accord*110ing to the rules of the common law”). For an application of this rule, see Gasperini v. Ctr. for Humanities, Inc., 518 U.S. 415, 116 S.Ct. 2211, 135 L.Ed.2d 659 (1996).
. See, e.g., 15B Charles Alan Wright, Arthur R. Miller & Edward H. Cooper, Federal Practice and Procedure § 3919.5, 650-51 (2d ed. 1991) (''[T]hat government appeals are generally available when dismissal rests on a ruling of law sought by the defendant, unmixed with any facts” is a conclusion "cemented by the decision in United States v. Scott.”); id. at 670 ("[T]he Scott decision allows [the burden of further trial proceedings] to be imposed if the defendant wins dismissal on a matter of law, not fact.”).
. Other courts have reached similar conclusions with respect to the meanings of statutes of limitations. See, e.g., Nesbitt v. Hopkins, 907 F.Supp. 1317, 1325 n. 9 (D.Neb.1995); Cox v. State, 585 So.2d 182, 189-92 (Ala.Crim.App.1991); Jackson v. State, 92 Md.App. 304, 608 A.2d 782, 784-87 (1992), disapproved on other grounds, Armstead v. State, 342 Md. 38, 673 A.2d 221, 241 n. 32 (1996).
. This is how we understand Justice Holmes’s statement, in a case concerning the federal crime of concealing assets from a bankruptcy trustee, that ''[a] plea to the statute of limitations is a plea to the merits, and ... after judgment upon it, it [can] not be reopened in a later prosecution.” United States v. Oppenheimer, 242 U.S. 85, 87-88, 37 S.Ct. 68, 61 L.Ed. 161 (1916) (citation omitted). Whether Oppenheimer extends beyond the context of bankruptcy or is still good law as to the meaning of that federal statute of limitations are questions we need not address in this case, in which the issue is how Connecticut reads its statute of limitations.