OPINION
WILKINSON, Circuit Judge:William Thomas Bauberger was convicted of second-degree murder and assault with a deadly weapon inflicting serious injury after he drove his car the wrong way down an exit ramp, killing one person and wounding another. Bauberger unsuccessfully challenged his murder conviction in state court after learning that the jurors read dictionary definitions of several words in the judge’s instructions.
Bauberger then sought federal habeas relief under 28 U.S.C. § 2254, arguing that the jurors’ dictionary use violated his federal constitutional rights and prejudiced his verdict. The district court agreed and granted the writ. Given that the dictionary definitions did not materially alter the instruction as a whole and that the government presented significant evidence of malice, any misconduct the jurors may have committed did not exert a “substantial and injurious effect ... in determining the jury’s verdict.” Brecht v. Abraham-son, 507 U.S. 619, 623, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993) (quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946)) (internal quotation marks omitted). Accordingly, the district court’s order granting Bauberger the writ is reversed, and the ease is remanded with directions to dismiss the petition.
I.
On February 3, 2002, William Bauberger attended a Super Bowl party at a friend’s house. Over the course of the five or so hours he was there, he drank more than ten beers. After the game Bauberger, despite his intoxicated condition, decided to drive to another friend’s house. He never made it. Instead, he drove his Cadillac — at a speed somewhere around 45 to 55 miles per hour — the wrong direction down an exit ramp off U.S. Highway 421 in Winston-Salem, North Carolina. Disregarding “Do Not Enter” and “Wrong Way” signs, as well as the honks and swerves of several cars traveling in the proper direction, Bauberger finally crashed his car into that of William and Carol Foy. William sustained several broken bones, and Carol died within minutes.
The government charged Bauberger with second-degree murder and assault with a deadly weapon inflicting serious injury. At trial the government introduced Bauberger’s troubled driving record. He had two prior driving-while-impaired (“DWI”) convictions, as well as a reckless driving conviction and other driving offenses. He also had disregarded prior court orders not to drive and was driving on a revoked license the night of the collision. Bauberger admitted that his blood-alcohol content that night was 0.20 and that he was aware of the dangers of driving while intoxicated. In light of this *103evidence, Bauberger conceded guilt to the lesser-included offense of involuntary manslaughter at trial but insisted that he lacked the malice necessary for a second-degree murder conviction under North Carolina law.
The jury convicted Bauberger of second-degree murder and assault with a deadly weapon inflicting serious injury, and he was sentenced to 189 to 236 months in prison. Shortly after the verdict came down, however, the parties and the court learned that the jury may have used a dictionary during its deliberations. Bauberger filed a postconviction Motion for Appropriate Relief (“MAR”), arguing that the dictionary was an impermissible extraneous influence on the jurors and that the dictionary definitions lowered the government’s burden of proof regarding malice. The MAR court determined from the jurors’ affidavits that the jury’s foreperson left the courthouse during a break in deliberations, went to a public library, and brought back the 1953 edition of Webster’s New Collegiate Dictionary. He read to the other jurors the dictionary’s definition of several terms in the judge’s malice instruction but not the definition of “malice” itself.
The MAR court denied Bauberger’s requested relief, reasoning that the jurors’ actions, though improper, were harmless. The North Carolina Court of Appeals affirmed, reasoning that Bauberger’s federal constitutional rights were not violated because the definitions “concerned legal terminology, not evidence developed at trial.” State v. Bauberger, 176 N.C.App. 465, 626 S.E.2d 700, 706 (2006). The North Carolina Supreme Court affirmed by an equally divided vote, leaving Bauberger’s convictions in place but stripping the lower court’s decision of precedential effect. See State v. Bauberger, 361 N.C. 105, 637 S.E.2d 536 (2006).
Bauberger filed for federal habeas relief under 28 U.S.C. § 2254. The district court granted his petition, holding that the jurors’ dictionary use violated Bauberger’s clearly established Sixth Amendment rights and that the error prejudiced him because the dictionary’s definitions of “recklessly” and “wantonly” may have lowered the government’s burden of proof regarding malice.
II.
We review the district court’s decision to grant the writ de novo. Bell v. Ozmint, 332 F.3d 229, 233 (4th Cir.2003).
We shall assume without deciding that the North Carolina Court of Appeals’s rejection of Bauberger’s Sixth Amendment claims was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” satisfying the threshold demands for habeas relief under the Antiterrorism and Effective Death Penalty Act (“AEDPA”). 28 U.S.C. § 2254(d)(1); see also Golphin v. Branker, 519 F.3d 168, 189-90 (4th Cir. 2008) (leaving unresolved whether the state court unreasonably applied federal law because any error did not have prejudicial impact under Brecht.) By doing so, we skirt the problems long associated with unnecessary constitutional decisionmaking: we avoid wasting the parties’ and the courts’ limited resources on “questions that have no effect on the outcome of the case,” Pearson v. Callahan, 555 U.S. 223, 129 S.Ct. 808, 818, 172 L.Ed.2d 565 (2009), and we minimize the threat to good adjudication that arises when courts enter into thorny constitutional areas with inadequate briefing or in fact-bound dispositions, see id. at 819-20.
Assuming arguendo that the state court erred in rejecting Bauberger’s claim *104does not end our inquiry, however. “[M]ost constitutional errors can be harmless,” including those of the kind we shall assume occurred at Bauberger’s trial. Arizona v. Fulminante, 499 U.S. 279, 306, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991); see, e.g., Fullwood v. Lee, 290 F.3d 663, 678-83 (4th Cir.2002) (third-party influence on juror and jurors’ consideration of extraneous evidence subject to harmless error review). On direct review, the government has the burden of proving that a constitutional error was “harmless beyond a reasonable doubt.” Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967).
On collateral review, however, the calculus changes. Because of the threat collateral attacks pose to “finality, comity, and federalism,” Fry v. Pliler, 551 U.S. 112, 116, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007), habeas petitioners may secure the writ only if the error “actually] prejudice[d]” them, Brecht, 507 U.S. at 637, 113 S.Ct. 1710 (quoting United States v. Lane, 474 U.S. 438, 449, 106 S.Ct. 725, 88 L.Ed.2d 814 (1986)) (internal quotation marks omitted). In making this determination, courts ask whether the error had a “substantial and injurious effect or influence in determining the jury’s verdict.” Id. (quoting Kotteakos, 328 U.S. at 776, 66 S.Ct. 1239) (internal quotation marks omitted). In the “unusual” situation where “the matter is so evenly balanced that [the habeas judge] feels himself in virtual equipoise as to the harmlessness of the error [under Brecht ]” — 'that is, where the judge is in “grave doubt” — the court must grant the writ. O’Neal v. McAninch, 513 U.S. 432, 435, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995).
The Supreme Court recently addressed the relationship between the Chapman direct review standard, AEDPA’s deference requirement, and the Brecht collateral review standard. In Fry, the question before the Court was whether habeas courts should apply the AEDPA/Chapman test (asking whether a state court unreasonably applied Chapman’s direct review standard), the Brecht test (asking whether a constitutional error had a substantial and injurious effect), or both. See Fry, 551 U.S. at 119-20,127 S.Ct. 2321.
The Court made two critical determinations. First, it held that “in § 2254 proceedings a court must assess the prejudicial impact of constitutional error in a state-court criminal trial under the [Brecht ] standard ..., whether or not the state appellate court recognized the error and reviewed it for harmlessness under [Chapman].” Fry, 551 U.S. at 121-22, 127 S.Ct. 2321 (emphasis added) (citations omitted). Second, the Court stated that “it certainly makes no sense to require formal application of both tests (AEDPA/Chapman and Brecht) when the latter obviously subsumes the former.” Id. at 120, 127 S.Ct. 2321. That is to say, where an error is harmful under Brecht, any state court decision declaring it harmless must have unreasonably applied Chapman. As a result, any error satisfying Brecht will also satisfy AEDPA’s deference requirements. See Ruelas v. Wolfenbarger, 580 F.3d 403, 411-13 (6th Cir.2009).
Thus the Court took a somewhat convoluted debate over the appropriate standard in harmless error habeas cases and opted for simplicity itself. Federal habeas courts must always review constitutional errors in state trials under Brecht, but they need not debate whether a state court’s harmless error determination also unreasonably applied Chapman, as most circuits since Fry have explicitly or implicitly recognized. See Welch v. Workman, 607 F.3d 674, 686 (10th Cir.2010); Wesbrook v. Thaler, 585 F.3d 245, 255-56 (5th Cir.2009); Ruelas v. Wolfenbarger, 580 *105F.3d 403, 411-13 (6th Cir.2009); Moses v. Payne, 555 F.3d 742, 755 (9th Cir.2009); Farley v. Bissonnette, 544 F.3d 344, 347-48 (1st Cir.2008); Bond v. Beard, 539 F.3d 256, 275-76 (3d Cir.2008). But see Perkins v. Herbert, 596 F.3d 161, 175-77 (2d Cir.2010) (leaving undecided whether both AEDPA/Chapman and Brecht must be used); Johnson v. Acevedo, 572 F.3d 398, 403-04 (7th Cir.2009) (requiring AED-PA/Chapman and Brecht where a state court addressed harmlessness).
Of course, most successful habeas petitioners must still go through “AEDPA[’s] mandate[d] ... two-step analysis” by demonstrating that the state court’s resolution of their constitutional claim was contrary to or unreasonably applied clearly established federal law and that the error was prejudicial under Brecht. Baum v. Rushton, 572 F.3d 198, 205 (4th Cir.2009); see also Barbe v. McBride, 521 F.3d 443, 453 (4th Cir.2008). But this two-step process does not require us to address the first prong where the petitioner’s claims fail on the second, see Golphin, 519 F.3d at 189— 90, and as explained above, Fry absolves us of any need to consider both AEDPAJChapman unreasonableness and Brecht prejudice in the harmless error context. Accordingly, we need only determine whether the jury’s dictionary use prejudiced Bauberger under the Brecht standard.1
III.
In analyzing whether the prejudice from the jury’s dictionary use meets the Brecht standard, we begin by laying out the jury’s actions in greater detail.
A.
The trial judge instructed the jurors on the meaning of “malice,” the only element of second-degree murder that Bauberger disputed, as follows:
Malice is a necessary element which distinguishes second degree murder from manslaughter. Malice arises when an act which is inherently dangerous to human life is intentionally done so recklessly and wantonly as to manifest a mind utterly without regard for human life and social duty and deliberately bent on mischief.
Shortly after beginning its deliberations, the jury requested a copy of the elements of second-degree murder and manslaughter. In response, the judge re-read the instructions to the jurors and returned them to their deliberations. However, they soon sent another note to the judge asking for a copy of the instruction regarding malice and any other contested element, stating that “[m]any of us are visual people.” The judge informed them he would prepare a copy during their upcoming lunch recess and dismissed them for that recess.
As previously indicated, during lunch the jury’s foreperson retrieved a dictionary and read some of its definitions to the other jurors when deliberations resumed. Though he did not read the definition for “malice” itself to the jury, he did read them the definitions for some of the words in the judge’s instruction, including “recklessly” and “wantonly.”2 The dictionary *106defined the former as “lack of due caution” and the latter as “arrogant recklessness of justice or the feelings of others.” Two hours into their resumed deliberations the jury informed the judge that they had resolved one count but were split seven to five on the other. The judge instructed them to do their best to “reconcile [their] differences ... without the surrender of conscientious convictions.” Within an hour of this instruction the jury stood at ten to two, and within two hours they had convicted Bauberger of second-degree murder and assault with a deadly weapon inflicting serious injury.
B.
Bauberger argues that the dictionary definitions of “recklessly” and “wantonly” lowered the government’s burden of proof on the issue of malice, an issue with which he contends the jury apparently struggled. As a result, Bauberger claims that the erroneous use of those dictionaries meets the Brecht standard: one or more of the jurors may have applied an unacceptably low malice standard in convicting Bauberger of second-degree murder, and therefore the dictionary use had a substantial and injurious effect upon his verdict. We disagree: the dictionary definitions did not materially alter the meaning of the instruction as a whole, and the government introduced significant evidence of malice.
1.
In assessing the harm from jurors’ use of a dictionary we look in part to the difference between the dictionary definition and the legal definition. See Henderson v. Kibbe, 431 U.S. 145, 154, 97 S.Ct. 1730, 52 L.Ed.2d 203 (1977) (looking to the difference between the actual instruction given and the correct instruction to determine prejudice); McNeill v. Polk, 476 F.3d 206, 226-27 (4th Cir.2007) (King, J., concurring in part and concurring in the judgment) (examining difference between dictionary and legal definitions); McNeill, 476 F.3d at 229-30 (Gregory, J., dissenting in part and concurring in part) (same). Under North Carolina law the “distinction between ‘recklessness’ indicative of murder and ‘recklessness’ associated with manslaughter ‘is one of degree rather than kind.’” State v. Rich, 351 N.C. 386, 527 S.E.2d 299, 303 (2000) (quoting United States v. Fleming, 739 F.2d 945, 948 (4th Cir.1984)). North Carolina courts have indicated, however, that the difference in degree is large. See, e.g., id.; State v. Wilkerson, 295 N.C. 559, 247 S.E.2d 905, 918 (1978). Nonetheless, they have upheld second-degree murder convictions even where a portion of the malice instruction veered toward culpable negligence in defining recklessness, provided the instruction as a whole conveyed the difference. See Rich, 527 S.E.2d at 301-03 (upholding a second-degree murder conviction where a portion of the recklessness instruction suggested that “recklessness of consequences” sufficed).
If Bauberger’s jury had only the dictionary’s “lack of due caution” definition to go on in determining the recklessness portion of the malice element, he would have more of a ease. But as the Supreme Court has long emphasized in the analogous context of jury instructions, “a single instruction to a jury may not be judged in *107artificial isolation, but must be viewed in the context of the overall charge.” Cupp v. Naughten, 414 U.S. 141, 146-47, 94 S.Ct. 396, 38 L.Ed.2d 368 (1973). Viewed in this light, the instruction, even as possibly modified by the definitions the jurors consulted, fully conveyed the essence of North Carolina law concerning malice. The modified version still referred to an “intentionally done,” “inherently dangerous” act. It still described the “lack of due caution” as of such a nature as to “show a mind fully or totally without respect or consideration for human life” or “social duty.” And it still spoke of an individual whose acts were so “arrogantly] reckless[ ] of justice or the feelings of others” as to indicate a mind “deliberately bent on mischief.”
This is not the language of mere culpable negligence. Phrases like these impose a much higher standard than the “thoughtless disregard of consequences” and “heedless indifference to the safety and rights of others” culpable negligence requires. State v. Mack, 697 S.E.2d 490, 494 (N.C.Ct.App.2010) (quoting State v. Wade, 161 N.C.App. 686, 589 S.E.2d 379, 382 (2003)) (internal quotation marks omitted). Bauberger’s verdict was not substantially and injuriously affected by the dictionary’s definition of “recklessly” because the altered instruction as a whole remained materially equivalent to the one given by the judge.
So too with “wantonly.” Under North Carolina law, “wantonness” describes “intentional wrongdoing,” conduct undertaken in “conscious and intentional disregard of and indifference to the rights and safety of others.” State v. Williams, 284 N.C. 67, 199 S.E.2d 409, 412 (1973); see also State v. Young, 148 N.C.App. 462, 559 S.E.2d 814, 818 (2002). Again, it would be one thing if the jurors had only considered “arrogant recklessness of justice or the feelings of others” in determining whether Bauberger’s conduct was wanton. But the possibly modified instruction in its entirety still conveyed the knowing disregard of others’ safety central to wantonness under North Carolina law. It spoke of “intentionally done” acts, ones performed with such “arrogant recklessness” toward others as “to show a mind fully or totally without respect or consideration for human life ... and deliberately bent on mischief.” (emphases added). Any modification of the instruction that came about by virtue of the dictionary’s definition of “wantonly” did not materially affect that instruction’s malice standard.
The other jury instructions also indicate that the jurors could not have read the isolated definitions so as to alter the malice instruction. “[T]he challenged instruction [is often] but one of many such instructions .... ” Henderson, 431 U.S. at 152 n. 10, 97 S.Ct. 1730 (quoting Cupp, 414 U.S. at 147, 94 S.Ct. 396) (internal quotation marks omitted). Here, those other instructions illustrate that the jurors were well aware of the difference between malice and culpable negligence. The jurors were instructed that if they acquitted Bauberger of second-degree murder they must consider whether he was guilty of involuntary manslaughter, which requires a showing of culpable negligence. To find culpable negligence, the trial court instructed the jurors that they must find that Bauberger drove while impaired, that he “willful[ly,] wanton[ly,] or intentionally]” violated the law, or that his “inadvertent or unintentional violation of the law ... [was] accompanied by recklessness of probable consequences of a dangerous nature ... amounting altogether to a thoughtless disregard of consequences or a heedless indifference to the safety of others.” The jurors also knew that Bauberger had conceded guilt to involuntary manslaughter, including its lower culpable *108negligence standard, making malice the central disputed issue before them.
Given these circumstances, it is unlikely the jurors seized on isolated dictionary definitions to transmute the malice standard- — which they knew to be the disputed issue in the case — into the culpable negligence standard, to which Bauberger had already conceded. In other words, Bauberger’s verdict was not substantially affected because the dictionary definitions, viewed both in light of the instruction as a whole and in light of the entire trial, did not materially alter the malice standard.
2.
We also look to the strength of the evidence in assessing whether the dictionary use substantially and injuriously affected Bauberger’s verdict. See McNeill, 476 F.3d at 226 (King, J., concurring in part and concurring in the judgment); id. at 229 (Gregory, J., dissenting in part and concurring in part). This approach makes sense: if the evidence presented was such that the issue of malice was not likely a close one, it is less likely that the error impacted the jury’s decision.
On this score, Bauberger’s claim to prejudice is particularly weak. His is the paradigmatic case of second-degree murder via drunk driving. As detailed above, he had several prior drunk driving convictions, prior court orders not to drive, and a revoked license on the night in question. He stipulated to a high blood-alcohol content and admitted knowledge of alcohol’s dangerous effects upon drivers. He also drove at a high speed, into oncoming traffic, and in disregard of several signs. North Carolina courts have routinely accepted evidence less formidable than this in upholding second-degree murder convictions. See, e.g., Rich, 527 S.E.2d at 304 (finding sufficient evidence of malice where the defendant “drove his vehicle at a high rate of speed while impaired, on the wrong side of the road, in a no-passing zone and in violation of right-of-way rules”); State v. Westbrook, 175 N.C.App. 128, 623 S.E.2d 73, 78-79 (2005) (finding sufficient evidence of malice where the defendant had a prior DWI conviction and sped through a traffic light on the wrong side of the road). While we are not analyzing Bauberger’s conviction for the sufficiency of the evidence, such compelling evidence of malice makes it considerably less likely that the dictionary use affected the jury’s ultimate decision.
Bauberger discounts the importance of this evidence by pointing to the struggles the jury apparently had in reaching its decision. He contends that the actual jury’s decisionmaking process — its requests for printed instructions, its foreperson’s decision to retrieve the dictionary, and its initially divided votes — is what counts under Brecht, not the actions of a hypothetical jury looking at the evidence. We agree in principle and have accordingly looked to the jury’s difficulty in reaching a decision in assessing prejudice in prior dictionary use cases. See McNeill, 476 F.3d at 226 (King, J., concurring in part and concurring in the judgment); id. at 229 (Gregory, J., dissenting in part and concurring in part).
However, as applied here, this principle cannot do the work Bauberger sets out for it because the record of the jury’s deliberations cannot bear this much weight. The jurors heard the definitions only an hour into their deliberations. And yet despite Bauberger’s contentions that the definitions lowered the malice standard to a level to which Bauberger had essentially conceded, the jury still needed an additional four hours before reaching a decision. This timeline tells us little about the issues around which the jurors’ struggles revolved. We are unwilling ultimately to put *109dispositive weight on a jury’s ambiguous actions. That would be altogether too speculative. Where the modified instruction fully conveyed the gist of North Carolina law and where the evidence presented was so strong, it takes clearer evidence of jury difficulty than this to satisfy Brecht.
IV.
“The principle that collateral review is different from direct review resounds throughout [the Supreme Court’s] habeas jurisprudence,” Brecht, 507 U.S. at 633, 113 S.Ct. 1710, and understandably so. Casually upending state convictions in federal court threatens “finality, comity, and federalism,” Fry, 551 U.S. at 116, 127 S.Ct. 2321, words which singly and together convey a sense of respect toward state judiciaries and the significant responsibilities assigned them. Thus we have respected state judgments where an error’s impact, considered in the context of the trial as a whole, was not significant. See, e.g., Golphin, 519 F.3d at 190-92 (overwhelming evidence rendered any error from admitting a confession harmless under Brecht); Wilson v. Ozmint, 357 F.3d 461, 469 (4th Cir.2004) (sentencing record rendered any error from not admitting one mitigating statement harmless under Brecht).
The error we assume arguendo occurred at Bauberger’s trial does not justify the relief he now seeks. Neither the court nor the government had any role in bringing about the erroneous dictionary use. The potentially modified instruction, taken as a whole, conveyed the essence of North Carolina law regarding malice. And BaubergeFs disturbing driving record, combined with his activities on the night of the accident, would more than permit a jury to find a “mind utterly without regard for” the life-threatening and ultimately life-ending risks created by his conduct. Given these circumstances, there was no substantial and injurious effect on his verdict from the jurors’ dictionary reading.3 The district court’s order granting Bauberger the writ is therefore reversed and the case remanded with directions to dismiss the petition.
REVERSED AND REMANDED
. We therefore need not resolve the parties' dispute about whether we should "look through” the North Carolina Court of Appeals’s decision (which did not conduct a harmless error analysis) to the MAR Court’s decision (which did) to perform AEDPA/Chapman analysis. See Ylst v. Nunnemaker, 501 U.S. 797, 803, 111 S.Ct. 2590, 115 L.Ed.2d 706 (1991) (habeas courts should look through summary affirmances to the "last reasoned decision” to determine whether a state procedural rule bars habeas review).
. The jurors also heard the definitions of "manifest” ("show”), "utterly” ("fully, total*106ly"), and "regard” ("respect or consideration for”), and earlier in Bauberger's trial one juror looked up the definition of "malice” in a dictionary at his home but did not copy it down, recall it, or share it with others. Following the district court, Bauberger focuses on the prejudicial impact of the dictionary definitions for "recklessly” and "wantonly” on appeal.
. To the extent the dissent wishes to ascribe some larger significance to our holding, we should note simply that we have assumed for purposes of argument the existence of the error and rested decision upon the application of the Brecht standard to the facts and circumstances of the case. The deference due state court judgments of conviction upon collateral attack has of course not absolved the court of the need to conduct a careful review of the matter, and we thank our distinguished colleague in dissent for his willingness to do likewise.