with whom WOLLMAN and BYE, Circuit Judges, join, dissenting.
I respectfully dissent from the judgment of the court because the court applies an incorrect standard in reviewing Mr. Nied-erstadt’s claim and reaches the wrong conclusion on its merits.
I.
The court does not acknowledge that there is an issue with respect to whether the Antiterrorism and Effective Death Penalty Act (AEDPA) standard of review, see 28 U.S.C. § 2254(d)(1), applies to Mr. Niederstadt’s claim: It passes over the matter entirely. After deciding that the state of Missouri had not properly preserved the issue of procedural default in state court, the court simply states that “[t]he task of a federal habeas court is to determine whether [a state court’s] conclusion was an ‘unreasonable application of ... clearly established federal law, as determined by the Supreme Court.’” But that standard applies only if Mr. Nieder-stadt’s claim was “adjudicated on the merits in state court.” 28 U.S.C. § 2254(d).
The court, in discussing the question of procedural default, relies on Muth v. Frank, 412 F.3d 808 (7th Cir.2005), cert. denied, 546 U.S. 988, 126 S.Ct. 575, 163 L.Ed.2d 480 (2005), for the proposition that if a state court denies a claim summarily we should presume that it decided it on the merits. But the motion for rehearing that Mr. Niederstadt filed in the Missouri Supreme Court contains claims asserting that the court’s interpretation and application of Missouri statutes was wrong both under Missouri law and the due process clause. The Missouri Supreme Court’s response was a one-line order stating simply that Mr. Niederstadt’s “motion for rehearing is overruled.” The order did not discuss or even acknowledge Mr. Niederstadt’s constitutional claim or any other claim that he raised.
Although in James v. Bowersox, 187 F.3d 866, 869 (8th Cir.1999), cert. denied, 528 U.S. 1143, 120 S.Ct. 994, 145 L.Ed.2d 942 (2000), we stated that “the summary nature of a state court’s ruling does not affect the § 2254(d)(1) standard of review,” the state court in James had reviewed the *842federal claim and “labelled]” it “ ‘without merit.’ ” Likewise, Weeks v. Angelone, 528 U.S. 225, 231, 237, 120 S.Ct. 727, 145 L.Ed.2d 727 (2000), is distinguishable: In that case, the state court had specifically referred to the claim at issue (issue “44”), along with several others, “considered” them, and found that they had “no merit.” Weeks v. Commonwealth, 248 Va. 460, 465, 450 S.E.2d 379, 383 (1994).
Since our decision in James, moreover, we have remarked that determining when a state court has decided an issue on the merits is “not so easy” and indicated that there are no hard-and-fast rules. Brown v. Luebbers, 371 F.3d 458, 460-61 (8th Cir.2004) (en banc), cert. denied, 543 U.S. 1189, 125 S.Ct. 1397, 161 L.Ed.2d 192 (2005). Indeed, in that case, both the court’s and the dissenting opinion discussed at length the question of whether the state court had reached petitioner’s federal claim or had decided only the state claim that was based on the same factual predicate. Because of all the labor devoted to that question, it is plain that both opinions in Brown proceeded on the assumption that if the state court exhibited no apparent awareness of the federal claim or gave no hint that it had considered it, we could not conclude that it had been adjudicated on the merits. And more recently we conducted a detailed examination of a state court opinion before determining that it resolved a particular claim on the merits. Weaver v. Bowersox, 438 F.3d 832, 838-39 (8th Cir.2006).
Thus, the premise of all of our previous relevant cases seems to be that when a state court has said nothing with respect to a federal claim, we can have no basis for concluding that it had adjudicated it. It therefore comes as a surprise to learn that we now have a binary approach to such matters: If a claim was dismissed for something other than a procedural default, it must necessarily have been adjudicated on the merits. Perhaps there is some justification for this rule, but the court offers none, and in the process overrules sub silentio a number of our cases that are inconsistent with any such principle. Or maybe our previous cases can be distinguished on the ground that in them the state court did discuss some claims and thus it might reasonably be inferred that the state court simply overlooked some claim or claims that went unmentioned. But the court here does not make that distinction, and I would not be inclined to make it either, for reasons that I allude to later.
It might also be argued that we ought not to presume that a state court has not considered a claim just because it did not mention it. After all, it often happens that appellate courts fail to discuss assignments of error without mentioning them directly, but the court’s judgment presumably concludes them nevertheless. But this kind of reasoning probably makes less sense when, as here, the state and federal claims are based on the same factual predicate, and the court does not advance this argument in defense of its ruling.
I think, moreover, that the principle apparently adhered to in our previous eases has a better claim to recognition than the one that the court adopts. First of all, one apparent purpose of AEDPA is to promote comity and to provide a certain amount of deference to the place of state courts in our overall federal system. The statute was intended to increase respect for state court judgments: It is therefore the states’ interests that are at stake here, and they should be alert to their vindication. I see no reason to encourage state courts to be laconic. Nor do I believe that it burdens a state court significantly to require it to devise an order that will allow a federal court to discern whether the state *843court has adjudicated a federal claim. The statute, moreover, does not by its own terms apply unless the federal claim has been adjudicated; and if the relevant state court order is opaque with respect to whether it decides a matter, the condition in the statute is not satisfied. To put it another way, if a state wants to claim the shelter that the statute provides, it has the burden to show that it is entitled to it.
It is important to appreciate that in passing AEDPA Congress meant only to limit a state prisoner’s right to an unrestricted review of a habeas claim, not to eliminate it. By requiring that a claim be adjudicated on the merits in state court before the review by a federal court would be restricted, I believe that Congress made clear that a habeas petitioner must receive an unrestricted, ie., de novo, review of his or her constitutional claims by either a state or federal court. We therefore correctly conducted an intensive review in our previous cases to determine whether the state court in fact had decided the federal claim at issue. When forced to choose between possibly depriving a habe-as petitioner of unrestricted review by any court and requiring a state court simply to say that it conducted such a review, I think that Congress’s intent compels us to choose the latter.
I would therefore review Mr. Nieder-stadt’s claim de novo.
II.
Even assuming for the sake of argument that the due process claim here was adjudicated on the merits and AEDPA therefore applies, I believe that Mr. Niederstadt would be entitled to habeas relief. In those cases in which the state court decides an issue but provides no explanation, we “conduct an independent review of the record and applicable law to determine whether the state court decision is contrary to federal law [or] unreasonably applies clearly established law.” Harris v. Stovall, 212 F.3d 940, 943 (6th Cir.2000). For the reasons that follow, I believe that the state court’s denial of Mr. Nieder-stadt’s due process claim is an unreasonable application of clearly established Supreme Court due-process precedent.
Over forty years ago, the Supreme Court said that “[t]he basic principle that a criminal statute must give fair warning of the conduct that it makes a crime has often been recognized by this Court.” Bouie v. City of Columbia, 378 U.S. 347, 350-51, 84 S.Ct. 1697, 12 L.Ed.2d 894 (1964). While the due process clause does not incorporate the specific prohibitions of the ex post facto clause, Rogers v. Tennessee, 532 U.S. 451, 458, 121 S.Ct. 1693, 149 L.Ed.2d 697 (2001), the “concepts of notice [and] foreseeability” are at its core, id. at 459, 121 S.Ct. 1693. Thus the due process clause is violated when a court gives retroactive effect to “a judicial construction of a criminal statute [that] is ‘unexpected and indefensible by reference to the law which had been expressed prior to the conduct in issue.’ ” Bouie, 378 U.S. at 354, 84 S.Ct. 1697 (quoting Jerome Hall, General Principles of Criminal Law 61 (2d ed.1960)); see also Rogers, 532 U.S. at 461, 121 S.Ct. 1693.
The Missouri sodomy statute, at the time of Mr. Niederstadt’s conduct, prohibited “deviate sexual intercourse with another person without that person’s consent by the use of forcible compulsion.” Mo. Rev.Stat. § 566.060.1 (Supp.1991). Deviate sexual intercourse included “any sexual act involving the genitals of one person and the mouth, tongue, hand, or anus of another person.” Mo.Rev.Stat. § 566.010(1) (Supp.1991). Forcible compulsion was defined as “physical force that overcomes reasonable resistance” or “a threat, express or implied, that places a *844person in reasonable fear of death, serious physical injury, or kidnapping of himself or another person.” Mo.Rev.Stat. § 556.061(12) (Supp.1991).
Mr. Niederstadt admits that the evidence would have been sufficient to convict him of deviate sexual assault in the first degree, Mo.Rev.Stat. § 566.070(1) (1986), which applied when the victim was an incapacitated person, or deviate sexual assault in the second degree, Mo.Rev.Stat. § 566.080(1) (1986), which applied to an assault of a sixteen-year-old victim by a person seventeen years or older. Neither of these crimes included the element of forcible compulsion. Mr. Niederstadt argues that he did not use forcible compulsion to accomplish the act, and that the Missouri Supreme Court’s construction of § 556.061(12)(a) to include his conduct diverged so widely from the law as it existed at the time of his offense as to violate his due process rights.
The Missouri Supreme Court held that Mr. Niederstadt used physical force against S.C., a minor dependent on him for care, in such a manner that her reasonable resistance to his sexual assaults was greatly reduced and then sodomized her while she slept. The act of penetrating S.C.’s vagina itself required force to be applied to her body, the court noted, and it held that that satisfied the physical force element of the statute. Niederstadt II, 66 S.W.3d at 14-16.
Under Bowie, we must determine what law “had been expressed prior to the conduct in issue.” We first examine the statutory language: The statute, after all, is the primary expression of the law applicable to this case. Section 566.060 required that the “deviate sexual intercourse” be accomplished “by the use of forcible compulsion” (emphasis added); it does not give notice that the intercourse itself can be the force used to accomplish the intercourse or that a sleeping or unconscious victim can be “compelled.” Under § 556.061(12), “forcible compulsion” is defined as “physical force that overcomes reasonable resistance,” but this definition does nothing to bring Mr. Niederstadt’s conduct within the scope of § 566.060: It provides no notice that “physical force” can be the intercourse itself or that the victim (whose “reasonable resistance” is overcome) may be asleep. Therefore we do not believe that the statutory language provided “fair warning” that Mr. Niederstadt’s conduct came within the terms of the crime with which he was charged.
Case law also plainly failed to provide the notice that due process requires. No Missouri case before Mr. Niederstadt’s holds or implies that the force necessary for forcible compulsion is equivalent to the performance of the sexual act which the statute states the forcible compulsion is used to accomplish. Nor had any case held or implied that a defendant could be convicted under § 566.060 when the victim was asleep and unaware of the acts alleged to be “forcible compulsion.” For example, in State v. R—D—G—, 733 S.W.2d 824, 827 (Mo.Ct.App.1987), the court’s discussion of force focused on the defendant’s grabbing and holding the victim’s arms and dragging her into a bedroom. Here S.C. testified that she was not even aware that Mr. Niederstadt was in the room until she awoke and, according to the Missouri Supreme Court, the deviate sexual intercourse had already been accomplished “by the use of forcible compulsion,” Mo.Rev. Stat. 566.060.
In holding that the force used to penetrate S.C.’s vagina could fulfill the element of forcible compulsion, the Missouri Supreme Court reconfigured the sodomy statutes in an unexpected way. Under this construction, the element of “forcible compulsion” collapses into the element of *845“deviate sexual intercourse,” Mo.Rev.Stat. § 566.060.1 (Supp.1991), and sodomy, when committed against a victim from whom virtually no reasonable resistance is expected, such as the sleeping victim here, becomes indistinguishable from the offense of deviate sexual intercourse in the second degree, which did not require proof of forcible compulsion. At the time of Mr. Niederstadt’s conduct, neither the plain language of the sodomy statute nor the case law supported such a construction. Where, as here, a court construes a criminal statute in a new way that removes an entire element from it, it violates the principle of fair warning that underlies the constitutional right to due process if, in the same case, the court uses that new statutory interpretation to uphold the defendant’s conviction.
As often happens, the state makes arguments and relies on authorities in its presentation to the en banc court that it never made to the district court or to the panel of our court that heard the case originally. The fact that this tack is late-blooming is evidence enough that the state did not originally regard these authorities as relevant, and rightly so. The court nevertheless adopts the state’s new arguments when it holds, ante at 837-38, that some old cases dealing with rape are helpful here. See State v. Welch, 191 Mo. 179, 89 S.W. 945 (1905); State v. Atkins, 292 S.W. 422 (Mo.1926). But those cases are inap-posite because they do not deal with a calibrated statutory scheme that deliberately distinguishes forcible sexual acts from acts that are committed on the incapacitated, including, presumably, the insensate.
The court also maintains, ante at 837 n. 1, that the Missouri Supreme Court preserved the distinction between the two offenses at issue here, but in fact the state court’s attempt to distinguish State v. Daleske, 866 S.W.2d 476 (Mo.App.1993), was at best startling and at worst entirely fanciful. The court also remarks, ante at 837 n. 1, that collapsing the two offenses would not be unconstitutional unless it was unexpected and indefensible. But surely the court does not mean to say that it expects state courts to collapse statutes or that it would defend state courts that did so.
I would therefore conclude that the Missouri Supreme Court’s decision was an unreasonable interpretation of United States Supreme Court due-process jurisprudence in Bouie and other cases because I believe that it is quite obvious that neither the plain language of the statute nor state case law at the time of Mr. Nieder-stadt’s conduct defined “forcible compulsion” as encompassing his conduct. I agree with the Missouri Court of Appeals, which stated in its opinion reversing Mr. Niederstadt’s conviction that “there are no facts in this case of persuasion or force. Defendant appeared while S.C. was sleeping. He initiated the sexual act while she slept. Defendant’s actions caused her to awaken. There was no evidence of forcible compulsion, as § 556.061(12) defines that term, before or after S.C. awoke.” Niederstadt I, 2001 WL 995937, at *3. Significantly, one member of the panel, after concurring in the state court of appeals’s unanimous opinion, wrote separately to express his “chagrin” that the reversal resulted from the prosecutor’s decision to pursue this charge: “Why the prosecutor chose to undertake the burden of proving forcible compulsion — an impossible task on the evidence here — defies explanation.” Id. at *4 (Shrum, J., concurring).
I recognize that the method of the common law will cause legal principles to migrate and expand somewhat as words and principles encounter new facts. But in the present case the Missouri Supreme *846Court’s application of the sodomy statute represented not the gradual evolution of a principle but a quantum leap that essentially redefined a statutory crime.
I recognize, too, that it is the sole province of the Missouri courts to construe their own statutes and that such constructions as those courts may give them are authoritative and binding on federal courts. We have no appellate jurisdiction over state courts and would not presume to venture an opinion that the interpretation that the Missouri Supreme Court gave the statute relevant to this case was incorrect as a matter of Missouri law: Indeed, it was correct by definition. The Missouri Supreme Court’s interpretation of the statute would remain intact, unaffected by a ruling in favor of the Mr. Niederstadt, and the courts of Missouri would be free to apply that interpretation to any conduct that occurs after the Missouri Supreme Court’s ruling in Mr. Niederstadt’s appeal. I would hold only that in applying its construction to Mr. Niederstadt, the Missouri Supreme Court violated his right to due process.
Mr. Niederstadt’s crime was grievous. Perhaps in some abstract moral sense he got what he deserved. But we are here to do law, not enforce morals, and Mr. Nied-erstadt’s sentence was at least eighteen years longer than it would otherwise have been because of the construction that the Missouri Supreme Court gave to the phrase “forcible compulsion.” In any event, Mr. Niederstadt did not get part of what he deserved, namely the due process of law that the Constitution secures for persons accused of crime. The maxim nulla poena sine lege, no punishment without a statute, is not a modern innovation intended to maximize the number of miscreants who go unpunished. It is a cherished principle of Anglo-American law, and ought not to be lightly shoved aside.
I would affirm the judgment of the district court.