EASTERBROOK, Circuit Judge.
Patrick Rogers assembled a team to rob Bryant’s Cocktail Lounge in Milwaukee. He recruited Robert Cranmore, Brian Rolf, and Robert Prihoda to do the job, and his girlfriend agreed to drive the getaway car. All four armed themselves, and in the early morning hours of August 17, 1975, they burst into the bar. Rogers and Rolf commandeered the lounge on the first floor; Cranmore and Prihoda were assigned to the bar on the second floor. All four terrorized the bartenders and patrons by brandishing weapons and threatening immediate death to all who did not hand over their valuables.
Gerald Drefahl and Dennis O’Bradovich, two off-duty policemen, were on the first floor when the robbers entered. O’Brado-vich was accompanied by his fiancée, celebrating their engagement. All three prostrated themselves as the gunmen demanded. O’Bradovich surreptitiously drew his revolver and, thinking that he was sheltered by the bar, demanded that the intruders surrender. A gun battle ensued. O’Bradovich retreated into the vestibule. Cranmore and Prihoda charged down the stairs, guns blazing. Cranmore got an angle on O’Bradovich and put a bullet through his head. Four desperadoes pumped ten slugs into O’Bradovich. He never regained consciousness. Other officers, alerted by a silent alarm, appeared as the robbers emerged from the tavern. Rolf gave up; Rogers was shot as he fled. Cranmore and Prihoda got away but were tracked down within hours. All four were convicted of first degree murder and armed robbery, and each was sentenced to life imprisonment.
Rogers, Rolf, and Cranmore appealed, and the court of appeals affirmed in an exhaustive opinion. Cranmore v. State, 85 Wis.2d 722, 271 N.W.2d 402 (1978). Priho-da escaped from prison, so the court did not consider his appeal. After being recaptured, Prihoda filed in 1980 an application for collateral relief. The only ground given was that pattern jury instruction 1100, which had been given at the trial, shifted to the defendants the burden of showing that they did not intend O’Bradovich’s death. The other three defendants made this argument on direct appeal and lost, 85 Wis.2d at 770-71, 271 N.W.2d 402, so it is no surprise that Prihoda lost too.
He tried again in 1985, this time raising a battery of arguments, including challenges to the adequacy of his counsel at trial and the voluntariness of his plea of guilty to the charge of armed robbery. He lost again, this time on the basis of Wis.Stat. § 974.06(4), which provides:
All grounds for relief available to a prisoner ... must be raised in his original, supplemental or amended motion [for collateral relief]. Any ground finally adjudicated or not so raised, or knowingly, voluntarily and intelligently waived in the proceeding that resulted in the conviction or sentence or in any other proceeding the prisoner has taken to secure relief may not be the basis for a subsequent motion, unless the court finds a ground for relief asserted which for sufficient reason was not asserted or was inadequately raised in the original, supplemental or amended motion.
Prihoda, represented by counsel in 1985, argued that lack of legal assistance in 1980 was a “sufficient reason” for a second motion despite failure to raise all grounds in the first. The court of appeals disagreed and rejected the petition without reaching the merits. The Supreme Court of Wiscon[1382]*1382sin denied review without opinion, 134 Wis.2d 458, 401 N.W.2d 10 (1987). Priho-da’s petition in federal court under 28 U.S.C. § 2254 failed largely because of the state court’s invocation of a procedural ground. The district judge held that Priho-da has not established “cause” for the forfeiture in state court. Although Prihoda preserved his challenge to instruction 1100, three opinions of this court hold that this instruction is not unconstitutional. Pigee v. Israel, 670 F.2d 690 (7th Cir.1982); Dean v. Young, 777 F.2d 1239 (7th Cir.1985); Fencl v. Abrahamson, 841 F.2d 760 (7th Cir.1988). Pigee, Dean, and Fencl did not leave much room for Prihoda to maneuver, so the district court dismissed his petition.
I
Prihoda asks us to overrule Pigee, Dean, and Fencl, but it would do him no good if we did. See Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989). Prihoda’s constitutional argument stems from Sandstrom v. Montana, 442 U.S. 510, 99 S.Ct. 2450, 61 L.Ed.2d 39 (1979), which was decided two years after his conviction became final. Pigee, Dean, and Fencl show that “a state court considering [Prihoda’s] claim at the time his conviction became final would [not] have felt compelled by existing precedent to conclude that the rule [Prihoda] seeks was required by the Constitution.” Saffle v. Parks, - U.S.-, 110 S.Ct. 1257, 1260, 108 L.Ed.2d 415 (1990). Any federal decision holding instruction 1100 unconstitutional therefore would be a new rule for purposes of Teague and could not be applied on collateral review. See also Sawyer v. Smith, - U.S.-, 110 S.Ct. 2822, 111 L.Ed.2d 193 (1990).
At all events, instruction 1100 is next to irrelevant to Prihoda’s conviction. Two theories supported his culpability: that he fired at O’Bradovich intending the officer’s death, and that he aided and abetted Cranmore, who fired the fatal shot through the officer’s brain. Instruction 1100 was pertinent to the former theory but not the latter. It is inconceivable that the jury would have used instruction 1100 to convict on the basis of Prihoda’s own intent while believing that Prihoda was not culpable as an associate of Cranmore. Any error is harmless beyond a reasonable doubt. Rose v. Clark, 478 U.S. 570, 106 S.Ct. 3101, 92 L.Ed.2d 460 (1986).
II
Wisconsin contends that all of Prihoda’s remaining arguments were forfeited by his failure to present them when first seeking collateral relief in 1980. So the state’s court of appeals held. Harris v. Reed, 489 U.S. 255, 109 S.Ct. 1038, 103 L.Ed.2d 308 (1989), which holds that forfeiture in state court blocks review in federal court only when it meets the conditions of the independent-and-adequate-state-ground doctrine, potentially presents Prihoda with three ways to avoid the conclusion of the state court.
1. A state ground is “independent” only if the state court actually relied on a state rule sufficient to justify its decision. Harris holds that a federal court may be sure that the state relied on a sufficient procedural ground only if the state court says so. Which state court? Wisconsin’s court of appeals invoked a state ground explicitly and exclusively, but the Supreme Court of Wisconsin denied review without stating reasons. One court of appeals has held that when the highest state court that has examined the case does not give reasons, the decision of the state’s judicial system as a whole does not rest on an independent state ground. Nunnemaker v. Ylst, 896 F.2d 1200 (9th Cir.1990). Nunnemaker was amended on May 25, 1990, to limit its scope, but even as amended holds that in at least some circumstances the highest court’s silence prevents the state from invoking a procedural bar even though an inferior court relied explicitly on procedural grounds. So an unexplained rebuff by the state’s supreme court, or a summary affirmance by an intermediate court, may abandon all state grounds of decision.
Three other courts of appeals hold, to the contrary, that the state court whose explanation matters is the last to write an opin[1383]*1383ion. Harmon v. Barton, 894 F.2d 1268, 1272-74 (11th Cir.1990); Evans v. Thompson, 881 F.2d 117, 123 n. 2 (4th Cir.1989); Ellis v. Lynaugh, 873 F.2d 830, 838 (5th Cir.1989). We agree with these decisions. Harris requires a plain statement of a procedural ground, not to tell state courts whether or how to write opinions, but to ensure that the federal court knows whether the state is relying on its own law as the basis of decision. Once a state court has supplied the necessary plain statement, a federal court understands the basis of decision. Unexplained affirmances or denials of discretionary review do not retract a state-law basis of decision already given.
Harris implements the rule of Wainwright v. Sykes, 433 U.S. 72, 97 S.Ct. 2497, 53 L.Ed.2d 594 (1977), that a procedural default bars federal review unless there is cause for and prejudice from that default. Like other doctrines controlling the scope of collateral review, Sykes is based on principles of comity and economy. Neither is served by supposing that a silent affirmance or denial of relief withdraws a sufficient state-law reason already given. The rules concerning independent and adequate state grounds evolved in cases in which the Supreme Court reviews judgments of state courts under its appellate jurisdiction. Harris applies to collateral attacks the principles developed in these cases, especially Michigan v. Long, 463 U.S. 1032, 103 S.Ct. 3469, 77 L.Ed.2d 1201 (1983). When reviewing the judgment of a state court, the Supreme Court examines the opinion given by the highest state court that has supplied an explanation. It looks straight through a denial of discretionary review, on the authority of 28 U.S.C. § 1257(a), which speaks of the final judgment “rendered by the highest court of a State in which a decision could be had” — that is, the highest court that has rendered decision on the merits. Sullivan v. Texas, 207 U.S. 416, 28 S.Ct. 215, 52 L.Ed. 274 (1908); American Railway Express Co. v. Levee, 263 U.S. 19, 20-21, 44 S.Ct. 11, 12-13, 68 L.Ed. 140 (1923). See also Keith v. Clark, 97 U.S. 454, 455-57, 24 L.Ed. 1071 (1878) (same approach when the state’s highest court affirmed without opinion). Cf. Robert L. Stern, Eugene Gressman & Stephen M. Shapiro, Supreme Court Practice 139— 43 (6th ed. 1986). Before Long the Supreme Court also looked to the decisions of intermediate state courts to determine whether a summary affirmance or denial of review rested on state grounds. Cicenia v. Lagay, 357 U.S. 504, 507 n. 2, 78 S.Ct. 1297, 1299 n. 2, 2 L.Ed.2d 1523 (1958); Stembridge v. Georgia, 343 U.S. 541, 547, 72 S.Ct. 834, 837-38, 96 L.Ed. 1130 (1952). In light of Long’s, plain-statement rule, the Court will no longer try to interpret silence; although the Court has not said so, this makes it sensible to look past the silence to the decision of the next state court in the chain. Federal courts taking over that jurisprudence on instructions from Harris should use the same approach the Supreme Court does on direct appeal. Wisconsin’s court of appeals furnished a ground of decision relying on state law; silence from the Supreme Court of Wisconsin does not undercut that basis.
2. A state ground is “adequate” only if the state court acts in a consistent and principled way. A basis of decision applied infrequently, unexpectedly, or freakishly may be inadequate, for the lack of notice and consistency may show that the state is discriminating against the federal rights asserted. Johnson v. Mississippi, 486 U.S. 578, 587-89, 108 S.Ct. 1981, 1987-88, 100 L.Ed.2d 575 (1988); Barr v. Columbia, 378 U.S. 146, 149, 84 S.Ct. 1734, 1736, 12 L.Ed.2d 766 (1964). Cf. Howlett v. Rose, — U.S.-, 110 S.Ct. 2430, 2440-41, 110 L.Ed.2d 332 (1990).
A requirement of consistency poses special problems for forfeiture rules, because every state uses exceptions such as “plain error” or “sufficient reason” or even “cause and prejudice” to deal with cases in which application of the ordinary rules would produce unacceptably harsh consequences. Sometimes the court gives alternative procedural and substantive grounds (“The claim has been waived; but even if it had not been we would find no error.”). Although this could be understood to say that the state does not really enforce its procedural rule, Harris tells us, 109 S.Ct. [1384]*1384at 1044 n. 10, that in alternative-grounds cases the federal court must respect the procedural basis, so long as the state court says that each ground is sufficient. See also Rogers-Bey v. Lane, 896 F.2d 279, 281 (7th Cir.1990); Williams v. Chrans, 894 F.2d 928, 934 (7th Cir.1990); Phillips v. Lane, 787 F.2d 208, 211-13 (7th Cir.1986); Russell v. Rolfs, 893 F.2d 1033, 1040-41 (9th Cir.1990) (Kozinski, J., dissenting).
If in alternative-grounds cases the federal court respects the procedural ground, it must follow that when the only ground given is procedural the federal court must respect it, even though in some other cases the state court ignores the potential procedural basis and addresses the merits. After all, what is an alternative-ground case but those two possibilities rolled into a single case? If the procedural basis prevails even when, by addressing the merits, the state court indicates doubt about the propriety of using its forfeiture rule, then it must prevail when the doubt is expressed in other cases but not this one.
Statements such as the one in Barr that a state procedural ground will be respected only when “strictly or regularly followed”, 378 U.S. at 149, 84 S.Ct. at 1736, turn out to be too strong, unless the “procedural ground” is understood to be a complex of rules and qualifications. Wisconsin employs a rule that failure to raise a ground forfeits it unless there are “sufficient” reasons. A rule of forfeiture-unless-good-reasons may be “strictly” followed when the state is scrupulous about treating particular reasons as “sufficient” routinely and others as insufficient with equal regularity.
Searching for “regularity” in the state’s employment of excuses and exceptions would embroil the federal court deeply in questions of state law and procedure. An alternative to the comprehensive survey of state cases every time a state invokes waiver is to recharacterize what it means for a state to apply its rule strictly. Both the Supreme Court and the inferior courts respect state procedural grounds unless they are regularly disregarded (or seemingly have been manufactured for the occasion, as in Johnson). A state ground that is solidly established will be respected even though not “strictly” followed. Any other approach would discourage state courts from applying plain error doctrines, lest giving one prisoner a break disable the state from enforcing its procedural rules with respect to many others. Neither prisoners’ interests nor the interests of federalism would be served by such a development.
Prihoda insists that § 974.06(4) is not “strictly” enforced. The statute itself allows a court to disregard the procedural bar for “sufficient reason”, so strict application is impossible under the statute’s own terms. Sometimes Wisconsin treats a prisoner’s lack of representation as a “sufficient reason”. State v. Wills, 69 Wis.2d 489, 493, 230 N.W.2d 827 (1975) (prisoner unrepresented and issue important). According to Prihoda, that toleration means that Wisconsin must give him the same boon. Especially, he says, because the form furnished to prisoners does not warn them of the effect of § 974.06(4) and refers to claims in the singular, implying that only one argument is to be made per case. Yet as the court of appeals observed in Priho-da’s own case, the statute does not contain an exception for prisoners proceeding without counsel. Wills relied on the importance of the issue as well as the lack of counsel. Other cases reiterate the message that pro se filing is not sufficient excuse. E.g., State ex rel. Dismuke v. Kolb, 149 Wis.2d 270, 441 N.W.2d 253 (Ct.App.1989). The form does not tell prisoners they may make but one claim; apparently many prisoners use it to make multiple arguments. No state case holds that the form supplies a reason adequate under the statute.
In the end, Prihoda is demanding automatic exemption from § 974.06(4) for prisoners who sought collateral relief pro se. Wisconsin is entitled to interpret its law so that “sufficient reason” means something more. In federal law, multiple collateral attacks are difficult to mount without a colorable showing of innocence. See Kuhlmann v. Wilson, 477 U.S. 436, 106 S.Ct. 2616, 91 L.Ed.2d 364 (1986), in which four Justices conclude that such a showing is [1385]*1385essential. Wisconsin does not make such a strict demand, but neither does it allow unrepresented prisoners to file an endless stream of requests for collateral attack, as Prihoda says it must. The difficulty of drawing lines, especially under standards such as “sufficient reason”, makes uncertain application inevitable. Uncertainty is not enough to disqualify a state’s procedural ground as one “adequate” under federal law. If it were, states would be induced to make their rules draconian rather than to allow prisoners the latitude now available.
3. A procedural ground also must be substantively sufficient if it is to be deemed “adequate”. The assumption animating the adequate-and-independent-state-ground doctrine was that when a state ground supplied a sufficient basis of decision, resolution of a federal issue would be an advisory opinion. If the state court says “This evidence must be suppressed under state and federal constitutions”, then discussion of the federal question by a federal court is gratuitous; it cannot affect the outcome. The state ground is a stopper.
Although Harris assumes that the doctrine may be applied to collateral attacks as well as appellate review, it loses something in the translation. Federal courts may override some state grounds of decision; not everything that is a stopper for state purposes is so for federal. Consider three related problems: (a) Does a procedural reason why a state may refuse to entertain a collateral attack on its own judgment imply that a federal court may not review the state judgment? (b) Is a default adequate for state purposes also adequate for federal purposes? (c) Does the fact that a state default is adequate only for state purposes imply that a federal court must reach the merits when other state grounds went unmentioned?
(a) The procedural ground deployed against Prihoda is one limiting the availability of collateral relief. Wisconsin’s court did not give reasons of the kind envisaged by Sykes, such as the contemporaneous-objection rule, or the requirement that claims be made in due course through the state’s system of appeals, Murray v. Carrier, 477 U.S. 478, 106 S.Ct. 2639, 91 L.Ed.2d 397 (1986); Smith v. Murray, 477 U.S. 527, 106 S.Ct. 2661, 91 L.Ed.2d 434 (1986). Statutes such as § 974.06(4) might be thought to speak more to the question of exhaustion of state remedies than to the sort of omission that would bar federal collateral relief. Why should a prisoner who preserved his claims at trial and on appeal, offering the state ample opportunity to do things right the first time, be turned away from federal court merely because the state has a rule limiting to one the number of collateral attacks it will consider? A state may disable its own courts from revisiting criminal convictions without putting a similar damper on federal courts.
Although it is not clear as a matter of first principles that failure to abide by the state’s rules for waging collateral attack should forfeit under § 2254 a claim that was properly preserved at trial and on direct appeal, this court has held that an omission has this consequence. Zellers v. Duckworth, 763 F.2d 250, 252 (7th Cir.1985); Williams v. Duckworth, 724 F.2d 1439, 1442 (7th Cir.1984). Prihoda does not ask us to revisit these holdings. See also Justus v. Murray, 897 F.2d 709, 713 (4th Cir.1990); Gilmore v. Armontrout, 861 F.2d 1061, 1064-66 (8th Cir.1988); Tacho v. Martinez, 862 F.2d 1376, 1380 n. 5 (9th Cir.1988); Deutscher v. Whitley, 884 F.2d 1152, 1156 (9th Cir.1989); Presnell v. Kemp, 835 F.2d 1567 (11th Cir.1988). We therefore must resolve the case on the assumption that Prihoda needs to establish “cause and prejudice” for limiting his initial collateral attack to a single ground.
(b) One leg of Prihoda’s efforts to establish “cause” is planted firmly in the Twilight Zone. He maintains that developments in the law after 1980 could not have been anticipated when he filed his first petition, making the changes “cause” under Reed v. Ross, 468 U.S. 1, 104 S.Ct. 2901, 82 L.Ed.2d 1 (1984). Yet if the petition depends on changes in the law, then according to Teague they may not be applied on collateral attack unless they fit into one of two exceptions (neither of which [1386]*1386is applicable). If the law has changed, then Prihoda may have “cause” but cannot prevail because of Teague. If the law has not changed, then Prihoda escapes Teague but cannot establish “cause”.
Although the Supreme Court has not noticed this interaction, it is inescapable that Teague apparently nullifies the effect of Reed. The exceptions in Teague deal with changes so substantial, or so strongly suggesting factual innocence, that they would allow collateral relief under the “fundamental miscarriage of justice” exception to Sykes, see Carrier, 477 U.S. at 495-96, 106 S.Ct. at 2649-50, whether or not the prisoner establishes “cause”. That leaves no independent role for a doctrine treating legal change as “cause”. It may be, then, that Reed is no longer authoritative for cases on collateral attack. See Joseph L. Hoffman, The Supreme Court’s New Vision of Federal Habeas Corpus for State Prisoners, 1989 Sup.Ct.Rev. 165, 183. (Reed may retain vitality for cases in which a change of law occurs during direct appeal after a failure to preserve a legal point at trial.)
The other string to Prihoda’s bow is his pro se status in 1980. Ineffective assistance of counsel may be “cause”, see Carrier, 477 U.S. at 488-89, 106 S.Ct. at 2645-46. Prihoda pleads the inadequacy of his own assistance. There are two problems. First, someone who chooses to represent himself may not turn around and contend that he did not give himself the quality of legal advice a lawyer could have supplied. Faretta v. California, 422 U.S. 806, 834 n. 46, 95 S.Ct. 2525, 2541 n. 46, 45 L.Ed.2d 562 (1975). Second, ineffective assistance supplies “cause” only when the Constitution requires the state to assure adequate legal assistance. Prihoda dropped the ball on collateral attack. There is no right to assistance of counsel in waging a collateral attack. Pennsylvania v. Finley, 481 U.S. 551, 107 S.Ct. 1990, 95 L.Ed.2d 539 (1987). Perforce there is no right to “effective” assistance of counsel. Poor (or no) assistance therefore is not “cause” for default. Morrison v. Duckworth, 898 F.2d 1298, 1301 (7th Cir.1990).
(c) The difficulties we have wrestled with are two aspects of a problem endemic to using the independent-and-adequate-state-grounds doctrine on collateral attack: What happens when a state denies relief on a basis sufficient under state law but irrelevant under federal law? A “plain statement” of Ground A, which turns out to be sufficient for the state’s purposes but insufficient for federal ones, may conceal the existence of Ground B, which would be redundant under state law but dispositive under federal law. Take a simple case: a state abolishes collateral attack in state courts. Such a decision would not affect federal jurisdiction under § 2254; the state rule would mean only that every prisoner has automatically exhausted state remedies. A prisoner who sought collateral relief in state court would be rebuffed with a citation to the law repealing the court’s authority. That procedural ground would be neither here nor there under Sykes. A state may have an additional and sufficient ground, such as failure to make a timely objection. Must the state court recite this too, and “plainly”, even though the state court would not deem it relevant for state purposes?
How would Harris require a federal court to treat the state’s decision? In one sense, the state court’s citation of the re-pealer would be the invocation of a procedural ground; yet because the state’s procedural ground would be irrelevant to Sykes, should it follow that the federal court goes straight to the merits, treating the absence of the kind of procedural ground relevant on federal attack as a concession that there is no procedural ground? Suppose that but for the lack of collateral review in state court, there would have been procedural obstacles aplenty: the defendant did not object at trial, did not raise issues on appeal, and so on. A strict requirement of “plain statement” of each ground would produce this curious consequence: by invoking a procedural ground sufficient to support its decision under state law, and most appropriate to the occasion, the state court would surrender all of the procedural grounds pertinent under federal law. Such a conclusion would be [1387]*1387inconsistent with the reasons the Court applied the “cause and prejudice” rule to collateral attack — (a) to require defendants to make their arguments at the time and place required by state law, so that errors may be avoided or cured at the first opportunity, and (b) to respect states’ legitimate interest in the enforcement of their own rules.
Such a possibility lurks in the background of this case. Suppose we were to agree with Prihoda that he had “cause” for not raising all of his claims in 1980. There would remain other, even more powerful, state grounds, which the state court did not need to mention because § 974.06(4) was adequate for its own purposes. One such ground shines through: Prihoda forfeited his direct appeal by taking to the lam. Carrier holds that failure to preserve a claim on direct appeal cuts out federal review if the state requires arguments to be raised on appeal on pain of forfeiture. Given § 974.06(4), Wisconsin’s court of appeals did not need to address this question. Unless under Harris raising a procedural barrier (see 109 S.Ct. at 1043, 1044 n. 12) preserves all procedural barriers, then the state must render advisory opinions on all potential procedural issues to avoid forfeiting its claims of forfeiture. Because we have held that, based on Teague, Prihoda failed to establish “cause” for omitting other grounds in 1980, we need not, and do not, resolve the problem of additional state procedural grounds presumably available to, but not invoked by, the state court.
Since 1985 Prihoda, with the assistance of vigorous and capable counsel, has been seeking collateral relief from his conviction. Three state courts and two federal courts have rejected these claims with scarcely a word about the merits. The length of this opinion testifies to the difficulty of administering a system of rules that sometimes (albeit rarely) excuses forfeitures in state court. Labors separating cases into forfeiture-enforced and forfeiture-waived categories may mask the broader point: Claims not finally resolved at trial and on direct appeal are almost always gone forever. Since 1977, when the Supreme Court decided Sykes, the trial and appeal have become the principal forum for the decision of all constitutional questions. Teague closes what was the last principal exception. No one should weep for Priho-da: his escape from custody, and not Wisconsin’s devotion to niggling rules, is the principal source of his inability to obtain further review. Other defendants should not miss the lesson of Sykes, Teague, and cases such as this one. The vise has closed.
AFFIRMED.