Opinion by Judge O’SCANNLAIN; Dissent by Judge THOMAS.
OPINION
We must decide whether erroneous jury instructions in a state murder trial were prejudicial as a matter of federal Constitutional law.
I
A
During the predawn hours of May 24, 1992, Ramon Flores, an attendant at a Shell gas station convenience store in San Mateo, California, was shot in the face with a .45-caliber bullet.1 He died almost instantaneously. There were no witnesses to the crime, but a neighbor heard a loud bang coming from near the station around 3:45 a.m. and then a voice yelling as if addressing someone else. A Coke can was found on the store’s counter. Michael Pu-lido’s thumbprint was on the can. A cash register, which had been removed from the store, was found later that morning in some bushes. Pulido’s fingerprints were on the cash register.
At the time, Pulido, who was sixteen years old, lived with his uncle, Michael Aragon, and Aragon’s cohabitant, Laura Moore. Aragon, Moore, two of their children, and a neighbor had previously seen Pulido with a .45-caliber Colt pistol. On two occasions, Pulido had observed to Aragon that the Shell station would be easy to rob because the attendant was always asleep. Aragon and Moore testified that *1010when they got up at 3 a.m. on the morning of the robbery to care for their baby, Pulido was not at home. However, when they awoke later that morning, Pulido was in the living room, already dressed. Pulido showed Aragon his wallet and said, “Look unc, almost all ones.” Later that day, Moore insisted that Pulido dispose of his gun. Pulido took the gun apart and discarded most of the pieces. Moore retained some of the pieces to prevent reassembly and turned them over to police, who determined that they came from a .45-caliber Colt.
When Pulido was arrested on an unrelated auto theft charge two weeks later, he told police he had information about the robbery and murder. He led police to some discarded .45-caliber cartridges with ejection marks resembling those on a cartridge found at the gas station. He also made several inconsistent exculpatory statements to police, first blaming a drug dealer named Carlos Vasquez, then his stepfather, Eduardo Alarcon, and finally, an unidentified Tongan male for the robbery and murder. Aragon testified, however, that Pulido had confessed to the crimes while the two were having dinner at a pizza parlor. Pulido told Aragon that he went to the gas station, bought a Coke, and left. He then returned and saw that the attendant was asleep. Although he considered shooting the attendant through the window, he decided instead to go inside and ask for another Coke. He shot Flores in the face, then ripped out the cash register and went back to his car. Pulido recanted this confession while in jail, however, writing in a letter to Moore: “If unc is reading this, tell him I didn’t kill that guy, I was just messing with him.”
At trial, Pulido claimed for the first time that Aragon had killed Flores. According to Pulido, he and Aragon were together the night of May 23 and ended up at the Shell station after Aragon smoked crack cocaine at Hunters Point. Pulido testified that he waited outside while Aragon went in to buy matches or cigarettes. After hearing a gunshot, Pulido ran into the store and saw Aragon holding Pulido’s gun. Flores was lying on the floor, bleeding from a shot to the face. Pulido yelled at Aragon, ran out of the store, and got back into the car. Moments later, Aragon came out with the cash register and threw it on Pulido’s lap. As they drove away, Aragon forced Pulido at gunpoint to open the cash register. Pulido handed the cash over to Aragon, then tossed the cash register into some bushes.
At the time of the shooting, Aragon was on probation for 1989 convictions for burglary, possession of cocaine, and contributing to the delinquency of a minor. Although Aragon denied using cocaine that night, Aragon’s sister testified that he was “on something” when she saw him either on May 24 or 25, and her son testified that Aragon was acting paranoid and smelled of crack cocaine. A police detective testified that Aragon had first claimed that he had gotten up at 12:15 a.m. to take care of the baby, but when interviewed with Moore at the police station, both said that it was around 3 a.m.
No physical evidence linked Aragon to the crime.
B
Pulido was convicted in the San Mateo County Superior Court of first-degree felony murder, robbery, receiving stolen property, and auto theft. The jury deadlocked on whether Pulido personally used a firearm and personally inflicted great bodily injury2 but unanimously returned a spe*1011cial-cireumstance finding of robbery-murder. Pulido was sentenced to life in prison without the possibility of parole.
On direct appeal, Pulido argued that the jury instructions on aiding and abetting felony murder and robbery, read together, impermissibly allowed him to be convicted of felony murder even if he did not form the intent to aid and to abet the robbery until after the murder. The California Supreme Court agreed that the felony-murder rule did not “include aiders and abettors or conspirators who join the felonious enterprise only after the murder has been completed,” and that the instructions “could well suggest to a jury that a person who aids and abets only in the asportation phase of robbery, after the killing is complete, is nonetheless guilty of first degree murder under the felony-murder rule.” People v. Pulido, 15 Cal.4th 713, 63 Cal.Rptr.2d 625, 936 P.2d 1235, 1243, 1245 (1997) (“Pulido I”). The court upheld Pulido’s conviction, however, on the ground that Pulido did not suffer any prejudice from the instructional error. In particular, the court concluded that the jury’s robbery-murder special-circumstance finding “demonstrates the jury did not accept the theory defendant joined the robbery only after Flores was killed,” but rather “found — explicitly, unanimously and necessarily — 'that defendant’s involvement in the robbery, whether as a direct perpetrator or as aider and abettor, commenced before or during the killing of Flores.” Id., 63 Cal.Rptr.2d 625, 936 P.2d at 1244.
Pulido thereafter filed this federal habeas petition, which the district court granted after discovering that the other special circumstance instruction — not relied upon by the California Supreme Court or by Pulido — was also defective. Pulido v. Lamarque, No. 99-4933, 2005 WL 6142229, at *14-*20 (N.D.Cal. Mar.24, 2005) (“Pulido II ”). Namely, it contained a typographical error, using the word “or” instead of “and” between its two prongs, thus enlarging the scope of activity that would qualify as robbery felony murder under the special circumstance. Id. at *14. Because the district court could not “be reasonably certain that the jury, if required to do so, would have found that [Pulido’s] involvement in the robbery preceded the victim’s death,” the court was “left with grave doubt as to the likely effect of [the] error on the jury’s verdict.” Id. at *20 (internal quotation marks omitted). Therefore, the district court concluded that Pulido was entitled to habeas relief on this claim of instructional error. Id.
We affirmed, concluding that the instructional error was structural, and that the verdict must be reversed because we could not be “ ‘absolutely certain’ that the jury found that Pulido’s crime of robbery was committed contemporaneously with the murder.” Pulido v. Chrones, 487 F.3d 669, 676 (9th Cir.2007) (per curiam) (“Pulido III ”) (quoting Lara v. Ryan, 455 F.3d 1080, 1086 (9th Cir.2006)). The Supreme Court vacated our decision, however, holding that we were “wrong to categorize this type of error as ‘structural’ ” rather than “ask[ing] whether the flaw in the instructions ‘had substantial and injurious effect or influence in determining the jury’s verdict.’ ” Hedgpeth v. Pulido, 555 U.S. 57, 129 S.Ct. 530, 530-31, 172 L.Ed.2d 388 (2008) (per curiam) (“Pulido IV”) (quoting Brecht v. Abrahamson, 507 U.S. 619, 623, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993)). Thus, the Court remanded for us to determine, in the first instance, whether Pulido was actually prejudiced by the faulty jury instructions. Id. at 533 & n. *.
II
Pulido contends that the error in the felony-murder instructions, combined with the newly discovered error in the special-*1012circumstance instruction, prejudiced him by allowing the jury to convict him based on the invalid theory that he did not join in the robbery until after Flores was murdered (the “late-joiner theory”).
“When considering an allegedly erroneous jury instruction in a habeas proceeding, [we] first consider whether the error in the challenged instruction, if any, amounted to ‘constitutional error.’ ” Morris v. Woodford, 273 F.3d 826, 833 (9th Cir.2001) (quoting Calderon v. Coleman, 525 U.S. 141, 146, 119 S.Ct. 500, 142 L.Ed.2d 521 (1998) (per curiam)). Where, as here, constitutional error is conceded, we proceed directly to the question of prejudice. See id.
In Fry v. Pliler, 551 U.S. 112, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007), the Supreme Court clarified that the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”) did not replace the traditional test for prejudice on collateral review — i.e., whether the error “had substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 623, 113 S.Ct. 1710. Moreover, Fry explained that we need not conduct an analysis under AEDPA of whether the state court’s harmlessness determination on direct review — which is governed by the “harmless beyond a reasonable doubt” test set forth in Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967) — was contrary to or an unreasonable application of clearly established federal law. Fry, 551 U.S. at 119-20, 127 S.Ct. 2321 (citing 28 U.S.C. § 2254(d)(1)). This is because the Brecht test “obviously subsumes” the “more liberal AEDPAJChapman standard which requires only that the state court’s harmless-beyond-a-reasonable-doubt determination be unreasonable.” Id. at 120, 127 S.Ct. 2321. Accordingly, we apply the Brecht test without regard for the state court’s harmlessness determination.3 See id. at 121-22, 127 S.Ct. 2321.
Under Brecht, an instructional error is prejudicial and habeas relief is appropriate only if, after reviewing the record as a whole, we conclude that there was a substantial and injurious effect or influence on the verdict, or if we are “left in grave doubt” as to whether there was such an effect. Kotteakos v. United States, 328 U.S. 750, 765, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946). Grave doubt exists in the “unusual” circumstance where, “in the judge’s mind, the matter is so evenly balanced that he feels himself in virtual equipoise as to the harmlessness of the error.” O’Neal v. McAninch, 513 U.S. 432, 435, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995).
We must “take account of what the error meant to [the jury], not singled out and standing alone, but in relation to all else that happened.” Kotteakos, 328 U.S. at 764, 66 S.Ct. 1239. Hence, “we accept at the outset the well established proposition that a single instruction to a jury may not be judged in artificial 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). This proposition recognizes that “not only is the challenged instruction but one of many such instructions, but the process of instruction itself is but one of several components of the trial which may result in the judgment of conviction,” including “testimony of witnesses, argument of counsel, [and] receipt of exhibits in evidence.” Id. at 147, 94 S.Ct. 396.
*1013III
We begin by considering whether the jury’s special-circumstance finding rendered the instructional errors harmless.
A
The introductory special-circumstance instruction, California Jury Instructions— Criminal (“CALJIC”) 8.80.1, began as follows:
If you find the defendant in this case guilty of murder of the first degree, you must then determine if the following special circumstance is true or not true: that the murder was committed while the defendant was engaged in or was an accomplice in the commission of, attempted commission of, or the immediate flight after committing or attempting to commit robbery.
(emphasis added). As the California Supreme Court observed, this statement properly instructed the jury that “the robbery-murder special-circumstance allegation could not be found true unless [Pulido] was engaged in the robbery at the time of the killing.” Pulido I, 63 Cal.Rptr.2d 625, 936 P.2d at 1243.
CALJIC 8.80.1 further instructed the jury that if it could not determine whether Pulido was “the actual killer or an aider and abettor,” it could not find the special circumstance unless (1) “the defendant with the intent to kill aided, abetted, ... or assisted any actor in the commission of the murder in the first degree,” or (2) “with reckless indifference to human life and as a major participant, aided, abetted, ... or assisted in the commission of the crime of robbery.” (emphases added). If the jury found the special circumstance on the basis of the “intent to kill” prong, it must have concluded that Pulido’s involvement in the robbery preceded the murder. Therefore, Pulido focuses on the “reckless indifference/major participant” prong, contending that the jury construed it in such a way as to find the special circumstance even if it believed his late-joiner theory.
Taking Pulido’s theory of events as true, it is not implausible that the jury would have considered Pulido a “major participant” in the two-person robbery. The question thus boils down to whether the jury also found that Pulido acted with “reckless indifference to human life” based solely on his post-killing actions.
1
Because the prosecution did not rely on the “reckless indifference” prong, that term was not defined for the jury.4 However, the California Supreme Court has deemed the legal meaning of “reckless indifference to human life” to be exactly the same as its commonsense meaning — “a defendant’s subjective awareness of the grave risk to human life created by his or her participation in the underlying felony” — thus eliminating any need for clarifi*1014cation for the jury. People v. Estrada, 11 Cal.4th 568, 46 Cal.Rptr.2d 586, 904 P.2d 1197, 1203 (1995).
It strains credulity to argue that post-killing participation in a robbery, by itself, created a “grave risk” to a life that had already been taken.5 Under the late-joiner theory, the only conduct creating any risk to human life was Aragon’s. That risk had already ripened into Flores’ death by the time Pulido stumbled into the crime scene. Moreover, Pulido testified that he “knew a person ... could not withstand a .45 caliber shot to the face.” If Pulido did not join the robbery until after Aragon shot Flores, Pulido could not have been subjectively aware that his assistance to Aragon was creating a “grave risk to human life.” Consequently, we cannot conclude that the jury stretched its “commonsense understanding” of “reckless indifference to human life” to include Pulido’s post-killing participation in the robbery. Boyde v. California, 494 U.S. 370, 381, 110 S.Ct. 1190, 108 L.Ed.2d 316 (1990).
2
Pulido seeks to minimize the significance of CALJIC 8.80.1 by deeming it a “general introductory instruction” over which “a specific substantive instruction” — the flawed version of CALJIC 8.81.17 — would prevail in the jury’s mind. We disagree.
First, the specificity and length of CALJIC 8.80.1 refutes Pulido’s assertion that this introductory instruction was prefatory and nonsubstantive.6 This two-page *1015instruction not only defined the special circumstance, but also set forth the specific requirements for finding the special circumstance if the jury could not determine whether Pulido was the actual killer. It is highly unlikely that the jury would have ignored the text of this detailed instruction merely because it was captioned “SPECIAL CIRCUMSTANCES — INTRODUCTORY.” (emphasis added).
Pulido mischaracterizes the content of CALJIC 8.81.17 in an effort to elevate its significance. According to Pulido, CALJIC 8.81.17 “was the substantive instruction defining the robbery special circumstance” and contained “the specific rules governing the special circumstance introduced in 8.80.1.” (second emphasis added). But CALJIC 8.81.17 did not “define” the special circumstance or “govern” CALJIC 8.81.17. To the contrary, CALJIC 8.81.17 merely provided:
To find that the special circumstance, referred to in these instructions as murder in the commission of robbery is true, it must be proved:
1. The murder was committed while the defendant was engaged in the commission or attempted commission of a robbery; or
2. The murder was committed in order to carry out or advance the commission of the crime of robbery or to facilitate the escape therefrom or to avoid detection. In other words, the special circumstance referred to in these instructions is not established if the robbery was merely incidental to the commission of the murder.7
The first sentence, which Pulido points to as “defining” the special circumstance, merely named it (“murder in the commission of robbery”). It is the first sentence of CALJIC 8.80.1 which actually defined the special circumstance: “[Y]ou must ... determine if the following special circumstance is true or not true: that the murder was committed while the defendant ... was an accomplice in the commission of ... robbery.” Furthermore, the body of CALJIC 8.81.17 did not, as Pulido contends, “govern” CALJIC 8.80.1. By its terms, CALJIC 8.81.17 merely set forth an additional determination for the jury to make, not an alternative one that somehow supplanted CALJIC 8.80.1.
The sequence of these instructions also belies Pulido’s contention that the jury would have focused on CALJIC 8.81.17 to the exclusion of CALJIC 8.80.1. Because CALJIC 8.80.1 — the introductory instruction — came first, the jury would have considered it first. If the jury had believed the late-joiner theory, it would not have found the special circumstance based on the introductory instruction because Pulido did not aid and abet the murder with intent to kill, or aid and abet the robbery with reckless indifference and as a major participant. Having already determined that the special circumstance did not apply, there would have been no reason for the jury to consider the next special-circumstance instruction, CALJIC 8.81.17, especially in light of the judge’s admonition to “[disregard any instruction which applies to facts determined by you not to exist.” Therefore, the jury would not have even considered CALJIC 8.81.17, let alone relied on it to convict Pulido on the invalid late-joiner theory.
We next turn to the special-circumstance verdict form, which stated that “[w]e, the jury ..., find the special circumstance that [Pulido] engaged in or was an accomplice in the commission of or attempted commission of robbery during the commission of the crime charged in count 1 [first-degree murder], to be TRUE.” (block capitalization omitted; emphasis added). Although the verdict form correctly sets forth the timing requirement, Pulido nevertheless contends that it failed to ameliorate the instructional errors.
Pulido argues that the language of the verdict form has little significance in the prejudice inquiry because “[i]t is the instructions, not the verdict forms, which state the substantive definitions of the charge.” However, given that the prejudice inquiry must encompass “the record as a whole,” Pulido IV, 129 S.Ct. at 533 n. *, it is entirely appropriate to consider the verdict form in conjunction with the jury instructions and the trial record as a whole, see Cupp, 414 U.S. at 147, 94 S.Ct. 396; cf. Mills v. Maryland, 486 U.S. 367, 375-76, 108 S.Ct. 1860, 100 L.Ed.2d 384 (1988). The verdict form is especially relevant because “[v]erdict forms are, in essence, instructions to the jury,” United States v. Reed, 147 F.3d 1178, 1180 (9th Cir.1998), and thus in some cases “can cure problems created by defective instructions,” United States v. Alghazouli, 517 F.3d 1179, 1190 (9th Cir.2008).
Pulido also argues that the jury would not have assumed that the verdict form contained the authoritative definition of the special circumstance. Therefore, contrary to the plain meaning of the form, the jury would have simply assumed that it encompassed the late-joiner theory. It is implausible, however, that the jury would have cavalierly made this assumption without seeking clarification. During deliberations, the jury “demonstrated that it was not too shy to ask questions, suggesting that it would have asked another” if faced with a verdict form that required it to assume that the word “during” also meant “after.” Weeks v. Angelone, 528 U.S. 225, 235-36, 120 S.Ct. 727, 145 L.Ed.2d 727 (2000). We therefore reject the notion of the jury’s “extremely gullible acceptance of a result that makes no conceivable sense.” Middleton, 541 U.S. at 438, 124 S.Ct. 1830.
Furthermore, the trial record refutes Pulido’s assertion that the text of the verdict form was disregarded by the jury. Prior to deliberations, the trial judge read the verdict form to the jury and informed the jury that the form would be available in the jury room for “use in arriving at a verdict.” After the jury returned its verdicts, the judge ordered the clerk to read them all aloud and asked the jury “to listen to them because you will be polled and asked if these are your verdicts.” The clerk then read the verdicts, including the special-circumstance verdict that Pulido “engaged in or was an accomplice in the commission of or attempted commission of robbery during the commission of [murder].” (emphasis added). When polled individually, each juror affirmed that the clerk had read his or her true verdicts. Based on the foregoing, we cannot assume that the verdict form was a practical nullity. See Humphries v. Dist. of Columbia, 174 U.S. 190, 194, 19 S.Ct. 637, 43 L.Ed. 944 (1899) (noting that the object of a jury poll “is to ascertain for a certainty that each of the jurors approves of the verdict as returned ” (emphasis added)).
C
Neither the introductory special-circumstance instruction nor the special-circumstance verdict form was consistent with the late-joiner theory. If the jury had followed the court’s instructions using its *1017eommonsense understanding of the terms, it would not have based its special-circumstance finding on the late-joiner theory. We therefore conclude that the special-circumstance finding militates in favor of harmlessness.
IV
We next consider whether the jury’s questions during deliberations reflected any prejudicial effect of the instructional errors.
A
During deliberations, the jury asked a series of questions about the meaning of the aiding and abetting and felony murder instructions, CALJIC 3.01 and CALJIC 8.27. No questions were asked, however, about the meaning of either of the special circumstance instructions, CALJIC 8.80.1 and CALJIC 8.81.17.
With respect to aiding and abetting, the jury first asked, “Is ‘aiding and abetting’ a robbery equivalent to a guilty conclusion on count 2 of (robbery)?” The court answered, “Yes — See attached copy of Instruction 3.01,” which defined aiding and abetting, in relevant part, as follows:
A person aids and abets the commission of a crime when he or she,
(1) with knowledge of the purpose of the perpetrator and
(2) with the intent or purpose of committing, encouraging, or facilitating the commission of the crime, by act or advice aids, promotes, encourages or instigates the commission of the crime.
The jury also inquired about CALJIC 8.27, the felony-murder instruction,8 which provided: -
If a human being is killed by any one of several persons engaged in the commission or attempted commission of the crime of robbery, all persons who either directly and actively commit the act constituting such crime, or who with knowledge of the unlawful purpose and with the intent or purpose of committing, encouraging, or facilitating the commission of the offense, aid, promote, encourage or instigate by act or advise its commission, are guilty of murder in the first degree, whether the killing is intentional, unintentional, or accidental.
The jury noted that “Felony homicide defi[ni]tion is unclear to us.... Attached are two different interpretations of the meaning of the sentence. Which is correct? A or B[?]” These interpretations, labeled “A” and “B,” appeared as follows:
A
EITHER:
—Actively commit robbery
—Both robbery intended and facilitates by aiding
B
ANY:
—Actively commit robbery
—Robbery intended
—Facilitate by aiding
In response, the court wrote: “We suggest you reread 8.27 with the definition of aiding and abetting in mind from instruction 3.01.”
The jury followed up by asking, “Does point 1 [of CALJIC 3.01] imply that the *1018‘knowledge of the purpose’ is knowledge prior to the commission of the crime or during the commission of the crime?” The court responded, “We are unable to answer the question as to an implication of whether ‘knowledge of unlawful purpose’ must be before or during[.] You should read 3.01 for its plain meaning in relation to all the other instructions bearing on this point.”
Finally, the jury asked whether “ ‘actively commit the act constituting such crime’ ” and “ ‘aid its commission’ ” in CALJIC 8.27 referred to the robbery or the murder. The court replied that “ ‘such crime’ refers to robbery mentioned in the previous line of instruction 8.27.”
B
According to Pulido, these questions indicate that the late-joiner theory “was the central focus” of the jury deliberations. In particular, Pulido points to the questions relating to the “facilitate by aiding” prong of CALJIC 8.27 and the “knowledge of the purpose” prong of CALJIC 3.01.
First, Pulido argues that the diagrams the jury drew corresponding to the defective felony-murder instruction, CALJIC 8.27, were intended “to confirm the availability of the invalid theory.” Id. at 12. However, neither diagram addresses the timing question. The difference between the diagrams is that Diagram B would have incorrectly supported a finding of felony murder if Pulido had merely “facilitate[d] by aiding” without any intent to commit robbery. In other words, the jury appeared to be grappling not with timing but rather intent, consistent with the duress defense presented at trial. By pointing the jury to CALJIC 3.01, which clearly required both intent and assistance under subpart 2, the court effectively invalidated the incorrect diagram. See Waddington v. Sarausad, 555 U.S. 179, 129 S.Ct. 823, 834, 172 L.Ed.2d 532 (2009). Furthermore, the guilty verdict on the robbery count confirms that the jury rejected the duress defense and properly found that Pulido had the requisite intent to commit robbery.
Pulido also contends that the question relating to “knowledge of the purpose” in CALJIC 3.01 goes to the crux of the timing issue. As noted above, the jury asked whether “knowledge of the purpose of the perpetrator” must be “knowledge prior to the commission of the crime or during the commission of the crime.” According to Pulido, “[t]he answer to the jurors’ question should have been a firm, “Yes’: To incur felony-murder liability, a defendant must have knowledge of the perpetrator’s purpose and assist in the robbery before or during the killing.”
Based on the jury’s later question whether “such crime” in CALJIC 8.27 referred to the murder or the robbery, it is entirely possible that the “crime” referred to in the question was the killing. But Pulido’s reading of the question as one calling for a yes-or-no answer — with “yes” meaning knowledge before or during, and “no” meaning knowledge after — is highly improbable. Read naturally, this is an either-or question — must the knowledge be before, or must it be during? If the jury question is taken at face value, it does not show any reliance on the invalid late-joiner theory, given that neither possibility contemplated by the jury is consistent with that theory. See Pulido I, 63 Cal.Rptr.2d 625, 936 P.2d at 1245. All this question shows is that the jury was asking whether premeditation was required, which is an entirely different timing question altogether.
C
At most, these jury questions suggest that the jury was uncertain about the application of the felony-murder instruction *1019in the context of aiding and abetting. However, there is no indication in the record that the jury imported this uncertainty into its application of the special-circumstance instructions. That the jury failed to ask a single question about the meaning of those instructions, notwithstanding the potential internal inconsistency created by the typographical error in CALJIC 8.81.17, indicates that it did not rely on a post-killing aiding-and-abetting theory to convict Pulido.
V
Finally, we consider whether the evidence in the trial record made it likely that the instructional errors had a substantial and injurious effect on the verdict.
Pulido asserts that because there was “substantial evidence that [he] did not join in the robbery until after the shooting,” the jury must have relied on the late-joiner theory in convicting him. But most of the evidence cited by Pulido for this proposition — including the neighbor’s testimony suggesting the presence of a second person at the crime scene, evidence of Aragon’s crack cocaine use (a possible motive for the robbery), and discrepancies in Aragon’s alibi statements- — -merely supports the theory that Aragon was also involved in the crime, and is equally consistent with Pulido’s post-killing and pre-killing participation. The only evidence that supports Pulido’s theory as to when he joined the robbery, therefore, is Pulido’s own uncorroborated testimony.
The State presented evidence that Pulido possessed the murder weapon both before and after the robbery. Moreover, Pulido had previously observed that the Shell station would be easy to rob because the attendant was always asleep. Pulido’s fingerprints — not Aragon’s — were found on the cash register, discrediting Pulido’s claim that Aragon had carried it out of the store. While this evidence does not definitively establish when Pulido joined the robbery, one piece of physical evidence, in our view, is sufficient to dispel any doubt as to whether the jury relied on the invalid theory — Pulido’s thumbprint on the Coke can found lying on the store counter.
The Coke can corroborated the prosecution’s theory that Pulido had used it as a diversion for the robbery, indicating prekilling participation. Although the defense conceded that the Coke can “is a crucial piece of evidence,” Pulido could not explain why his print was on the can. He denied ever touching it that night and speculated that he must have touched it while buying another Coke during a prior visit to the store. Pulido’s counsel acknowledged during closing argument that “that’s not a good explanation” and suggested that Pulido’s testimony about not touching the can was the result of a memory lapse. He could not explain the presence of the print, either. .
Pulido’s late-joiner story was at least the fourth version of the events he had offered and found no support in any evidence other than his own self-serving testimony. Moreover, Pulido could offer only speculation as to how his thumbprint ended up on “a crucial piece of evidence” tying him to the robbery before the murder. Consequently, the record evidence fails to persuade us that the instructional errors caused Pulido any prejudice. See Morales v. Woodford, 388 F.3d 1159, 1173 (9th Cir.2004) (“Mere speculation is insufficient to grant the writ under Brecht, because speculation does not give rise to a ‘grave doubt’ whether the error had a substantial effect in determining the jury’s verdict.”).
VI
For the foregoing reasons, we can say, “with fair assurance, after pondering all that happened without stripping the erroneous action from the whole, that the judg*1020ment was not substantially swayed” by the instructional errors. Kotteakos, 328 U.S. at 765, 66 S.Ct. 1239. Because Pulido did not suffer any actual prejudice, he is not entitled to habeas relief. See Brecht, 507 U.S. at 637, 113 S.Ct. 1710.
Accordingly, the judgment of the district court is
REVERSED and REMANDED.