Julius Evans v. Alex Jones ( 2021 )


Menu:
  •                               In the
    United States Court of Appeals
    For the Seventh Circuit
    ____________________
    No. 19-3466
    JULIUS EVANS,
    Petitioner-Appellee,
    v.
    ALEX JONES,
    Respondent-Appellant.
    ____________________
    Appeal from the United States District Court for the
    Northern District of Illinois, Eastern Division.
    No. 14-cv-03930 — Rebecca R. Pallmeyer, Chief Judge.
    ____________________
    ARGUED SEPTEMBER 25, 2020 — DECIDED MAY 5, 2021
    ____________________
    Before RIPPLE, BRENNAN, and ST. EVE, Circuit Judges.
    ST. EVE, Circuit Judge. An Illinois jury convicted Julius Ev-
    ans of the first-degree murder of Moatice Williams, who was
    killed in a drive-by shooting in Chicago. Only one eyewit-
    ness—Andrew Jeffers—connected Evans to the shooting. Jef-
    fers’s account of the shooting dramatically changed over time.
    Jeffers initially only provided a few general identifying de-
    tails of the shooter, and did not specifically identify any of the
    shooters. Eleven months later, however, the police
    2                                                  No. 19-3466
    approached him while he was incarcerated and Jeffers then
    identified Evans as the shooter. Then, at trial, Jeffers recanted
    that identification: he testified that he did not see the identity
    of the shooter but had identified Evans because the police told
    him to.
    During closing arguments, the prosecutor argued that Jef-
    fers’s trial testimony—that he did not see Evans shoot Wil-
    liams—was false and the jury should disbelieve it because Jef-
    fers only changed his story after being paid a visit by a de-
    fense investigator working for Evans’s co-defendant, Mario
    Young, who was a known gang member. Evans appealed his
    conviction, asserting that the prosecutor’s statements during
    closing argument deprived him of his right to a fair trial. He
    contended that the prosecutor’s statements were improper
    because there was insufficient evidence in the record to sup-
    port them, and that they were prejudicial because Jeffers’s
    credibility was of the utmost importance given the lack of
    other evidence against Evans. The state appellate court con-
    cluded that there was sufficient evidence in the record to sup-
    port the prosecutor’s statements during closing argument,
    and so they were not improper.
    Evans unsuccessfully petitioned the state court for post-
    conviction relief. He then filed a habeas petition in federal
    court, which the district court granted. Upon a close examina-
    tion of the record and giving deference to the state appellate
    court’s findings, we find that the state appellate court’s deter-
    mination that the prosecutor’s statements were proper was
    objectively unreasonable. While we “do not lightly grant pe-
    titions for a writ of habeas corpus brought by state prisoners,”
    Cook v. Foster, 
    948 F.3d 896
    , 899 (7th Cir. 2020), we agree with
    the district court that the facts of this case compel the
    No. 19-3466                                                  3
    conclusion that Evans was deprived of his right to a fair trial
    and he is entitled to relief. We therefore affirm.
    I. Background
    A. Factual Background
    On the evening of August 23, 1996, someone inside a vehi-
    cle opened fire on West Washington Street in Chicago. The
    bullets hit and instantly killed Moatice Williams, who had
    been sitting on his bicycle on the sidewalk. After arriving on
    the scene, Chicago Police Officer James Cianella interviewed
    two individuals who had witnessed the crime: Margaret Win-
    ton and Andrew Jeffers. Police did not identify any additional
    witnesses.
    Winton told Officer Cianella that she had been selling
    goods across the street from Williams. Right before the shoot-
    ing, she had observed Jeffers and a man named John “pitching
    quarters”—trying to see who could get the quarters closest to
    the line on the sidewalk—two or three feet from Williams. She
    then saw a gray car with tinted windows drive down the
    street, turn around the block, then proceed for a second time
    down West Washington Street. Someone then fired seventeen
    or eighteen gunshots from the front passenger-side window.
    Winton saw three men in the car (including the shooter) but
    did not see their faces and could not identify them.
    Jeffers told Officer Cianella that he saw the vehicle stop
    and fire shots, and then drive east. There were three black
    men in the car, one of whom was wearing a white t-shirt. He
    did not provide any other identifying information.
    4                                                     No. 19-3466
    B. Investigation
    Eleven months later—in July 1997—two Chicago Police
    Department “Cold Case Squad” detectives visited Jeffers in
    prison to get his recollection of the shooting. Jeffers was serv-
    ing a prison sentence for an unrelated drug offense at the
    time. At trial, the detectives testified that during this visit they
    showed Jeffers a six-person photo array and asked whether
    he recognized any of the individuals from the night of the
    shooting. By that stage of the investigation, the detectives had
    identified three primary suspects: Evans, Mario Young, and
    Royce Grant. Their photographs made up three of the six pho-
    tos in the array. The detectives testified that Jeffers identified
    Evans and Young—Evans as the shooter and Young as the
    front seat passenger in the vehicle. Jeffers initialed the backs
    of the photographs of Evans and Young to certify having iden-
    tified them.
    Three months later, the detectives paid Jeffers another
    visit—this time, to a boot camp where Jeffers was serving the
    remainder of his prison sentence. Assistant State’s Attorney
    Lorraine Scaduto accompanied. At trial, Scaduto testified that
    she showed Jeffers two photographs, one of Evans and one of
    Young. According to Scaduto, Jeffers again identified Evans
    as the shooter and Young as the vehicle’s front passenger. Sca-
    duto asked Jeffers to describe the events of the shooting and
    requested permission to transcribe his recollection into a writ-
    ten statement. Jeffers agreed.
    According to Jeffers’s written statement, on the night of
    the shooting he was pitching quarters with John across the
    street from where he lived on West Washington Street. The
    victim sat on his bicycle watching the game. Jeffers “bent
    down to pick up some quarters” and heard “four or five
    No. 19-3466                                                   5
    gunshots.” He looked up and “saw a gray Oldsmobile Cut-
    lass, two-door” driving slowly down the street. Jeffers saw
    three men in the car: one man was driving, another was lean-
    ing forward in the front passenger seat, so that the third man
    could reach over him and shoot toward the street from the
    front passenger window. The next day, Jeffers was pitching
    quarters with John again in the same location. Evans and
    Young—whom he recognized from the Oldsmobile Cutlass
    the day before—drove up in a different car. Evans apologized
    to John for having shot at him the day before, explaining that
    they mistakenly thought they were shooting at members of a
    rival gang. Jeffers signed the statement to certify that he had
    given the statement “freely and voluntarily and that no
    threats or promises were made to him in exchange for his
    statement.”
    The police arrested Evans, Young, and Grant on Novem-
    ber 14, 1997. A few weeks later, the detectives called upon Jef-
    fers to identify Evans and Young from an in-person lineup.
    Jeffers had been released from boot camp and was serving the
    remainder of his sentence on house arrest. At trial, a detective
    testified that Jeffers was shown a six-person lineup consisting
    of Evans, Grant, Young, and three non-suspects. The detective
    testified that Jeffers again identified Evans as the shooter and
    Young as the man sitting in the front seat.
    Jeffers met with Assistant State’s Attorney Ann Lorenz
    later that day. Lorenz asked Jeffers to tell her what he had wit-
    nessed in relation to the shooting. According to Lorenz, Jeffers
    recounted a version of events consistent with his written state-
    ment. Lorenz testified that she showed Jeffers two photo-
    graphs, one of Evans and one of Young. Jeffers confirmed to
    Lorenz that those were the men he saw in the Oldsmobile the
    6                                                  No. 19-3466
    night Williams was murdered. Jeffers then testified before a
    grand jury that same day. His testimony was consistent with
    his handwritten statement.
    C. Trial Court Proceedings
    At Evans’s trial in June 2000, Jeffers recounted a much dif-
    ferent version of what had happened. Despite the accounts he
    gave during the investigation nearly three years prior, Jeffers
    testified at trial that although he was pitching quarters on
    West Washington Street at the time of the murder, he did not
    actually see the shooting. Instead, as soon as he heard gun
    shots, he ducked and did not look up again until after the fir-
    ing stopped. He testified that he did not know the shots had
    been fired from a passing car, he did not see the car, and he
    certainly did not see who shot and killed Williams.
    The prosecutor pressed Jeffers about the statements he
    gave the detectives and identifications he made of Evans and
    Young. Jeffers testified initially that he never identified pho-
    tographs of Evans and Young, never signed a handwritten
    statement of his account of the shooting, and never met with
    Assistant State’s Attorney Lorenz. Jeffers testified that he had
    never seen Evans before the trial.
    Eventually, Jeffers admitted identifying Evans and Young
    from the lineup, but only because the detectives told him to,
    not because he recognized them from the night of the shoot-
    ing. According to Jeffers, the detectives told him to “stick with
    the story, tell the story,” but Jeffers did not elaborate further
    on the “story” he was told to “stick with.” Jeffers testified that
    the story of what he saw was simple, he: “Be[nt] down pitch
    quarter, shots were fired. Ran across the street, called the am-
    bulance. Ambulance came, detectives grabbed me, snatched
    No. 19-3466                                                   7
    me, threw me in the car.” Jeffers testified that any other details
    attributed to his story were made up by the detectives—
    “[t]hey added all that stuff on.”
    On cross-examination, Jeffers testified that he was not be-
    ing given anything in exchange for his trial testimony, nor had
    anyone threatened him to testify a certain way. Jeffers testified
    again on redirect examination that no one threatened or in-
    timidated him with respect to his trial testimony:
    Q:      It’s your testimony today that no gang
    members or no one intimidated you into giving
    this testimony, is that right?
    A:     Right.
    The prosecutor then asked Jeffers whether a private investi-
    gator working for Mario Young visited him before trial to
    speak with him about his testimony:
    Q:    But after you were released from boot
    camp you got a visit from someone working for
    Mario Young, the co-defendant, didn’t you?
    A:     No.
    Q:     Didn’t an investigator working for Mario
    Young’s lawyer come to your home and ask you
    questions about what happened and talk to you
    about the shooting?
    A:     No.
    On re-cross examination, defense counsel attempted to clarify
    Jeffers’s testimony about whether he had spoken with an in-
    vestigator working for Young:
    8                                                No. 19-3466
    Q:    The state’s attorney just asked you if you
    spoke to an investigator for Mario Young when
    you were released from custody, is that correct,
    sir?
    A:     When I was released from custody.
    Q:     Right, when you were out.
    A:     Right.
    Q:     And did you tell them, did you stick with
    the story with that investigator, sir?
    A:     No.
    Q:     Pardon me, sir?
    A:     No.
    Q:     So you didn’t tell that person what you
    told the grand jury, correct?
    A:     Yeah.
    Q:     Because you didn’t stick with the story?
    A:     Right.
    On further direct examination, the prosecutor again asked if
    Jeffers had been visited by an investigator working for Young:
    Q:     I thought you just said you don’t remem-
    ber being visited by an investigator for Mario
    Young, the co-defendant, after you were re-
    leased from custody; didn’t you just say that ten
    minutes ago?
    A:     You asked me if she came to my house.
    Q:     Who is it that came and visited you?
    No. 19-3466                                                9
    A:      Didn’t no one come to my house.
    Q:      Where did they visit you at?
    A:     I don’t remember. She didn’t come to my
    house.
    Q:      But you don’t remember where?
    …
    A:    Whoever she was, she didn’t come to my
    house, whatever you say.
    Q:     But you know it was an [i]nvestigator
    who worked for Mario Young, the defendant in
    this case, didn’t you?
    A:      No.
    Q:      You didn’t know that?
    A:      No.
    Q:      Who did you think it was?
    A:   I don’t know who it was. She just asked
    me questions and I talked to her.
    Q:     Asked you questions about the shooting,
    right?
    A:      Yeah.
    Q:     And that is when you decided to start
    saying that you didn’t see who did the shooting,
    right?
    A:      I told her the truth.
    This concluded Jeffers’s testimony. No further evidence was
    presented during trial about an investigator visiting Jeffers.
    10                                                 No. 19-3466
    Before proceeding with closing arguments, the court in-
    formed the jury that “[c]losing arguments are made by the at-
    torneys to discuss the facts and circumstances in the case and
    should be confined to the evidence and to reasonable infer-
    ences to be drawn from the evidence.” It further instructed
    the jury that “[a]ny argument made by the attorneys which is
    not based on the evidence should be disregarded.”
    With Jeffers being the only witness able to identify the men
    involved in the shooting, the closing arguments centered on
    his credibility. The state’s initial closing argument urged the
    jury to credit Jeffers’s earlier statements over his trial testi-
    mony. Defense counsel argued the opposite—encouraging
    the jury to believe Jeffers’s testimony that he did not witness
    the shooting and could not identify the shooter.
    In the state’s rebuttal closing argument, the prosecutor at-
    tempted to explain the change in Jeffers’s testimony. The
    prosecutor argued that Jeffers provided very little infor-
    mation to law enforcement at the crime scene out of fear of
    retaliation by gang members. In the prosecutor’s view, Jeffers
    only felt safe enough to tell the detectives what he witnessed
    and identify Evans when he was in state custody and “away
    from the street where the gang bangers dominate.” The pros-
    ecution argued that the change in Jeffers’s account could be
    traced to a visit from an investigator working for Young:
    Think about when the story changed. Andrew
    Jeffers didn’t just identify Julius Evans and
    Mario Young one time. He identified him four
    times …. It only changed after he was released
    from custody when lo and behold he gets a visit
    from an investigator working for the lawyer for
    Mario Young, the defendant’s co-offender.
    No. 19-3466                                               11
    Defense counsel objected and the court overruled the objec-
    tion without explanation. The prosecutor continued:
    An investigator comes to visit him, and that per-
    son apparently interviews Andrew Jeffers. For
    what purpose? You can draw your own conclu-
    sion. Andrew Jeffers now knows when the in-
    vestigator visits him they know how to find
    him.
    The court overruled another objection, and the prosecutor
    went on:
    They know where he is at. The gang bangers
    that executed Moatice Williams on the street
    right next to him know how to find Andrew Jef-
    fers. They can come and see him whenever they
    want. All of a sudden after getting a visit from
    an investigator working for them—
    Defense counsel objected again and this time the court sus-
    tained. The prosecutor nonetheless continued:
    For Mario Young, his co-defendant, his co-of-
    fender, all of a sudden Andrew Jeffers can’t re-
    member his name. He can’t remember a damn
    thing. ... What a surprise. What a surprise that
    Andrew Jeffers after being visited by Mario
    Young’s investigator would suddenly forg[e]t
    everything that he saw when–on the night of
    August 23, 1996, when Moatice Williams was
    gunned down by this defendant. ... That’s why
    Andrew Jeffers was such a pain when he was on
    the witness stand yesterday. You can draw your
    own conclusions. Andrew Jeffers now knows
    12                                                 No. 19-3466
    that Mario Young or his investigator knows
    how to find Andrew Jeffers.
    The prosecutor made a last attempt to bolster Jeffers’s out-of-
    court statements, and after the court overruled another objec-
    tion, he asserted:
    Andrew Jeffers was one hundred percent sure
    and entirely consistent in his identifications of
    what happened and who he saw do it … until
    he was paid a visit by Mario Young’s investiga-
    tor … that’s when suddenly he started to los[e]
    his memory which is not a big surprise or
    shouldn’t be a big surprise to anybody.
    The prosecutor concluded and the court instructed the
    jury. With respect to the parties’ closing statements, the court
    stated:
    Closing arguments are made by the attorneys to
    discuss the facts and circumstances in the case
    and should be confined to the evidence and to
    reasonable inferences to be drawn from the evi-
    dence. Neither opening statements nor closing
    arguments are evidence. Any statement or argu-
    ment made by the attorneys which is not based
    on the evidence should be disregarded.
    The jury issued a guilty verdict the following day. The trial
    court sentenced Evans to life in prison.
    D. Direct Appeal
    Evans appealed his conviction to the Illinois Appellate
    Court. He argued that the prosecution engaged in misconduct
    by arguing in its rebuttal closing statement that an
    No. 19-3466                                                      13
    investigator working for Evans’s co-defendant, Mario Young,
    caused Jeffers to recant his earlier statements. 1 The appellate
    court upheld the conviction.
    The state appellate court determined that the prosecutor
    “did not invent an allegation of witness intimidation without
    any basis in the record.” Instead, the prosecutor “made a rea-
    sonable inference from the evidence which demonstrated that
    Jeffers dramatically changed his testimony at trial after receiv-
    ing a visit from an investigator sent by codefendant Young, a
    known gang member.” Finding adequate support in the rec-
    ord for the prosecutor’s comments, the appellate court held
    the statements did not substantially prejudice Evans.
    The appellate court reasoned the record reflected that Jef-
    fers identified Evans as the shooter on multiple occasions,
    provided a written statement describing what he witnessed
    on the night of the murder, testified against Evans before a
    grand jury, and only “abruptly recanted after being visited by
    an investigator sent by codefendant Young.” The state court
    interpreted trial testimony as reflecting that “Jeffers admitted
    that he did, in fact, meet with [Mario Young’s] investigator.”
    The court further reasoned that Jeffers’s explicit testimony
    that he was not threatened by Evans did not render improper
    the prosecutor’s inference that Young’s investigator caused
    Jeffers to recant his prior statements at trial. As an initial mat-
    ter, the court explained, Jeffers “did not state, nor was he
    asked, whether he was threatened by the investigator.” More
    importantly to the appellate court, the record reflected that
    1 Evans also appealed his conviction on other grounds that are no
    longer at issue. Accordingly, we will not discuss them further.
    14                                                  No. 19-3466
    “Jeffers identified defendant as the shooter in this case on five
    separate occasions and that his testimony only changed after
    he was released from custody and visited by an investigator.”
    Given that “the investigator was sent by codefendant Young,”
    and Evans and Young “worked together to murder the vic-
    tim,” the prosecutor reasonably invited the jury “to infer that
    Jeffers’s reluctance to testify was the result of fear invoked by
    the investigator’s visit.”
    Evans filed a petition for leave to appeal, and the Illinois
    Supreme Court denied his petition without comment on De-
    cember 5, 2002.
    E. Post-Conviction Proceedings
    After unsuccessfully petitioning the state courts for post-
    conviction relief, Evans filed a pro se petition for writ of ha-
    beas corpus pursuant to 
    28 U.S.C. § 2254
     in federal court. As
    relevant to this appeal, Evans argued that the prosecutor’s ref-
    erences to Young’s investigator deprived him of his right to a
    fair trial. After appointing Evans counsel, the district court
    agreed and granted Evans’s petition. According to the court,
    it was objectively unreasonable for the state court to conclude
    that the prosecutor’s statements were supported by the record
    or based on a reasonable inference of the evidence in the rec-
    ord. The court thus determined that the statements were im-
    proper and potentially prejudicial, and had deprived Evans
    of his right to a fair trial. This appeal followed.
    II. Discussion
    We review the district court’s grant of a habeas petition de
    novo, “but our inquiry is an otherwise narrow one.” Schmidt
    v. Foster, 
    911 F.3d 469
    , 476 (7th Cir. 2018) (en banc). Under the
    Antiterrorism and Effective Death Penalty Act of 1996
    No. 19-3466                                                                15
    (AEDPA), a federal court may grant habeas relief only when
    a state court adjudication on the merits: (1) “was contrary to,
    or involved an unreasonable application of, clearly estab-
    lished Federal law, as determined by the Supreme Court of
    the United States,” or (2) “was based on an unreasonable de-
    termination of the facts in light of the evidence presented in
    the State court proceeding.” 
    28 U.S.C. §§ 2254
    (d)(1), (2). Only
    Supreme Court precedent—not circuit court precedent—con-
    stitutes clearly established federal law in § 2254 habeas cases.
    See Parker v. Matthews, 
    567 U.S. 37
    , 48–49 (2012). A state court
    unreasonably applies clearly established federal law if it “cor-
    rectly identifies the governing legal rule from Supreme Court
    case law, but unreasonably applies it to the facts of the case.”
    Clark v. Lashbrook, 
    906 F.3d 660
    , 664 (7th Cir. 2018).
    Darden v. Wainwright, 
    477 U.S. 168
     (1986), provides the
    clearly established federal law for Evans’s prosecutorial mis-
    conduct claim. 2 See Parker, 
    567 U.S. at 45
    . Under Darden, a
    prosecutor’s improper statements deprive a criminal defend-
    ant of his right to a fair trial if the remarks “so infected the
    trial with unfairness as to make the resulting conviction a de-
    nial of due process.” Darden, 
    477 U.S. at 181
     (quoting Donnelly
    v. DeChristoforo, 
    416 U.S. 637
    , 643 (1974)). A defendant seeking
    relief under Darden embarks on an “uphill battle; ‘improper
    statements during closing arguments rarely constitute
    2 The Illinois Appellate Court did not explicitly cite Darden in its anal-
    ysis of Evans’s prosecutorial misconduct claims—instead, it relied on
    state-court precedent to conduct a Darden-like analysis. We agree with the
    district court that the state court’s failure to specifically discuss Darden
    does not negate the applicability of Darden to this case. See Evans v. Lash-
    brook, 
    2019 WL 6117585
     at *6 n.7 (N.D. Ill. Nov. 18, 2019) (citing Ruvalcaba
    v. Chandler, 
    416 F.3d 555
    , 565 (7th Cir. 2005)).
    16                                                    No. 19-3466
    reversible error.’” United States v. Klemis, 
    859 F.3d 436
    , 442 (7th
    Cir. 2017) (quoting United States v. Wolfe, 
    701 F.3d 1206
    , 1211
    (7th Cir. 2012)).
    Darden is a “highly generalized standard,” Parker, 
    567 U.S. at 49
    , and its application “can demand a substantial element
    of judgment.” Yarborough v. Alvarado, 
    541 U.S. 652
    , 664 (2004).
    Our analysis in this case thus requires that we acknowledge
    that “[t]he more general the rule, the more leeway courts have
    in reaching outcomes in case-by-case determinations.” 
    Id.
     We
    also note, however, “[e]ven in the context of federal habeas,
    deference does not imply abandonment or abdication of judi-
    cial review.” Miller-El v. Cockrell, 
    537 U.S. 322
    , 324 (2003).
    In applying Darden, we must “first look to the challenged
    comments to determine whether they were improper.” Ellison
    v. Acevedo, 
    593 F.3d 625
    , 636 (7th Cir. 2010). Only if the state-
    ments are improper must we decide if they “so infected the
    trial with unfairness” as to have denied the defendant his due
    process right to fair trial. 
    Id.
     (quoting Darden, 
    477 U.S. at 181
    ).
    A. Whether the Prosecutor’s Statements were Improper
    It is well established that a prosecutor may not reference
    facts not before the jury to bolster a witness’s credibility. See
    United States v. Alviar, 
    573 F.3d 526
    , 542 (7th Cir. 2009). A pros-
    ecutor may, however, “argue reasonable inferences from the
    evidence that the jury has seen and heard.” United States v.
    Waldemer, 
    50 F.3d 1379
    , 1383 (7th Cir. 1995).
    Evans challenges seven statements from the prosecutor’s
    rebuttal closing argument as improper. The essence of each
    challenged statement is the same: the prosecutor argued that
    Jeffers recanted his identification of Evans as the shooter dur-
    ing his trial testimony because an investigator working for co-
    No. 19-3466                                                  17
    defendant Mario Young visited him before that testimony.
    The prosecution implied that the visit intimidated Jeffers be-
    cause Young now knows where he lives. According to Evans,
    these statements were improper because there is no evidence
    in the record from which the prosecutor could reasonably in-
    fer that an investigator who worked for Young visited Jeffers.
    The Illinois Appellate Court disagreed and concluded that
    the prosecutor’s remarks were proper because Jeffers had tes-
    tified that he did, in fact, meet with an investigator sent by
    Young. Upon closely examining the content of Jeffers’s testi-
    mony, however, we conclude that the Illinois Appellate
    Court’s finding that Jeffers so testified misstates the record
    and lacks evidentiary support. We thus agree with Evans that
    the state appellate court’s determination that the prosecutor’s
    comments were proper is objectively unreasonable.
    When Jeffers was initially questioned by the prosecutor
    about a visit from an investigator working for Young, he twice
    explicitly denied such a visit occurred. Later, on re-cross, Ev-
    ans’s counsel asked Jeffers if he recalled the prosecutor’s ques-
    tions. In response, Jeffers testified that he did, in fact, meet
    with someone about the shooting, but he did not testify that
    the person was an investigator sent by Young. On further re-
    direct, the prosecutor questioned Jeffers about why he origi-
    nally denied that he was visited by an investigator who
    worked for Young. Jeffers testified that he spoke to a woman
    about the case, but that he did not know who she was:
    Q:     But you know it was an [i]nvestigator
    who worked for Mario Young, the defendant in
    this case, didn’t you?
    A:     No.
    18                                                 No. 19-3466
    Q:     You didn’t know that?
    A:     No.
    Q:     Who did you think it was?
    A:   I don’t know who it was. She just asked
    me questions and I talked to her.
    The Illinois Appellate Court acknowledged that Jeffers in-
    itially denied being visited by an investigator who worked for
    Young. Nonetheless, it determined that during re-cross, Jef-
    fers admitted that he spoke to an investigator who worked for
    Young. This interpretation mischaracterizes Jeffers’s testi-
    mony on re-cross and seemingly ignores his clarification on
    re-direct that he did not know who the investigator was or
    who she worked for. While Jeffers admitted to speaking with
    an investigator, he repeatedly denied that this investigator
    worked for Young. The Illinois Appellate Court’s finding to
    the contrary lacks evidentiary support, and so its determina-
    tion that the prosecutor’s comments were based on a reason-
    able inference from Jeffers’s testimony—and so were
    proper—is objectively unreasonable.
    First, focusing on Jeffers’s testimony on re-cross, we rec-
    ognize that his indirect responses to certain of defense coun-
    sel’s questions make the transcript somewhat difficult to fol-
    low. But there is no reasonable reading of this exchange in
    which Jeffers admits to speaking with an investigator work-
    ing for Mario Young. Instead, the testimony establishes only
    that Jeffers spoke to an investigator to whom he told the
    “story” consistent with his testimony at trial, and inconsistent
    with his prior statements to police.
    Second, even if Jeffers’s testimony on re-cross was ambig-
    uous as to whether the investigator worked for Young, his
    No. 19-3466                                                  19
    testimony on re-direct clarified the issue. The prosecutor
    asked Jeffers why he had previously denied being visited by
    an investigator who worked for Young. Jeffers explained that
    he had spoken to a woman about the case but flatly denied
    that he had any knowledge that she had a connection to
    Young. The Illinois Appellate Court seemingly did not take
    this testimony into account.
    In fact, Jeffers explicitly denied having met with an inves-
    tigator who worked for Young five separate times. Nonethe-
    less, the state appellate court weighed his testimony on re-
    cross—in which he never explicitly stated the investigator
    worked for Young—over his clear and repeated denials that
    such a visit occurred. The Illinois Appellate Court’s determi-
    nation that Jeffers testified that he met with an investigator
    who worked for Young therefore is not supported by the trial
    evidence.
    The state appellate court’s conclusion that the prosecutor’s
    comments were proper depended on its determination that
    Jeffers had testified that an investigator who worked for
    Young visited him. According to the state appellate court, be-
    cause there was evidence in the record that Jeffers had “dra-
    matically changed his testimony at trial after receiving a visit
    from an investigator sent by codefendant Young, a known
    gang member,” the prosecutor’s statements during closing ar-
    gument were proper because the prosecutor made a reasona-
    ble inference that the change in Jeffers’s testimony could be
    attributed to witness intimidation. When Jeffers’s testimony,
    however, is properly characterized—that he met with an in-
    vestigator but had no knowledge as to whom the investigator
    worked for—the prosecutor’s comments no longer rest on a
    reasonable inference from the evidence in the record. Without
    20                                                No. 19-3466
    evidence connecting the investigator to Young, it was not rea-
    sonable for the prosecutor to argue that Jeffers changed his
    testimony after a visit from an investigator who worked for
    Young because Jeffers feared that Young knew where to find
    him. While the record does not reflect who the investigator
    was or who she worked for, the only testimony about her em-
    ployer is that she did not work for Young. Close examination
    of the record reveals that the state court mischaracterized and
    misstated the content of Jeffers’s testimony. The prosecutor’s
    statements during closing argument about witness intimida-
    tion were therefore improper because the record did not sup-
    port them, nor do they reflect a reasonable inference from the
    record. The state appellate court’s determination to the con-
    trary therefore reflects an unreasonable application of Darden.
    On appeal, the state makes two primary arguments. First,
    the state argues that because both the prosecutor and defense
    counsel characterized the investigator as working for Young,
    “both sides agreed that the person who spoke to Jeffers
    worked for Young” and the record is therefore susceptible to
    multiple interpretations. But even if the attorneys shared the
    implicit assumption that the investigator worked for Young,
    this shared belief is not evidence and does not introduce am-
    biguity into Jeffers’s clear and repeated denials that he spoke
    to an investigator who worked for Young.
    Second, the state argues that because Jeffers admitted he
    spoke to an investigator, it was proper for the prosecutor to
    link that investigator to Young because the investigator must
    have been working for the defense. In the state’s view, this
    must be true because Evans “does not contend that this inves-
    tigator worked for the State, nor could he, as it was a defense
    investigator who approached Jeffers.” The state makes this
    No. 19-3466                                                    21
    circular contention without citation to the record. A lack of
    evidence that the investigator worked for the state does not
    constitute evidence that the investigator worked for the de-
    fense. It is precisely the lack of evidence about who the inves-
    tigator worked for that makes the prosecutor’s comments im-
    proper. Evans does not need to prove the investigator worked
    for the state to successfully argue that there was no evidence
    in the record from which the prosecutor could reasonably in-
    fer that the investigator worked for Young. And without evi-
    dence linking the investigator to the defense, the prosecutor’s
    statements about why Jeffers’s testimony changed—that he
    was paid a visit by an investigator hired by a known gang
    member who now knew where to find him—was not a rea-
    sonable inference.
    Accordingly, the Illinois Appellate Court’s determination
    that the prosecutor’s statements were supported by the rec-
    ord, and therefore proper, was unreasonable. The dissent sug-
    gests that we have reached this conclusion without consider-
    ing Jeffers’s testimony about the investigator in the context of
    the entire trial. In doing so, however, it is the dissent which
    applies too narrow a lens to its analysis. The dissent hangs its
    hat on a single exchange in which defense counsel inquired
    whether “[t]he state’s attorney just asked if [Jeffers] spoke to
    an investigator for Mario Young” and Jeffers responded,
    “Right.” (emphasis added). In the dissent’s view, this was an
    affirmative response that can reasonably be viewed as affirm-
    ing either that he was just asked that question or that he in fact
    spoke to an investigator working for Mario Young. Because
    both of these inferences are possible, the dissent argues that
    the prosecutor’s statement in closing argument that Jeffers
    spoke to an investigator working for Young is a “logical con-
    clusion.”
    22                                                  No. 19-3466
    But that conclusion is only logical if one ignores the con-
    tent and context of the defense counsel’s question, Jeffers’s an-
    swer, Jeffers’s previous denial that he spoke to such an inves-
    tigator, and his subsequent clarification that he spoke to
    someone but he did not know if the investigator worked for
    Young (and did not know who she worked for). This context
    is not “an exercise in keeping score,” as the dissent suggests.
    Our determination that the prosecutor’s remarks were not
    based on a reasonable inference from record evidence is not
    based on the fact that Jeffers denied that he met with an in-
    vestigator working for Young more times than he admitted it.
    Rather, in considering Jeffers’s full testimony, we do not find
    such an inference of admission reasonable. Interpreting Jef-
    fers’s answer of “right” to a question about whether the state’s
    attorney had just asked a question as an affirmation about the
    very thing that he had just twice denied is not reasonable, es-
    pecially considering his subsequent clarifications.
    Rather than relying solely on Jeffers’s style of speaking to
    search for reasonable alternative inferences that might be
    drawn from his testimony, as the dissent proposes, we base
    our conclusion on the facts to which Jeffers did or did not tes-
    tify. Jeffers specifically did not admit to having spoken with
    Mario Young’s investigator, and he expressly stated the oppo-
    site: he did not speak to an investigator who worked for Mario
    Young. We decline the dissent’s invitation to base our decision
    on each of these facts in isolation, and instead consider the
    trial record in full. It was unreasonable for the Illinois Appel-
    late Court to ignore Jeffers’s explicit testimony which estab-
    lishes only that he spoke with a female investigator, and that
    no one threatened him prior to testifying at trial. Because
    there was no evidence in the record as to whom the investiga-
    tor worked for, there was insufficient evidence presented at
    No. 19-3466                                                  23
    trial to support the prosecutor’s arguments that Jeffers’s testi-
    mony changed because a defense investigator working for a
    known gang member visited him. While “[i]t is of course true
    that in closing counsel may make arguments reasonably in-
    ferred from the evidence presented,” at some point “the in-
    ference asked to be drawn will be unreasonable enough that
    the suggestion of it cannot be justified as a fair comment on
    the evidence.” United States v. Vargas, 
    583 F.2d 380
    , 385 (7th
    Cir. 1978) (holding that a prosecutor’s closing statement that
    the defendant had previously trafficked heroin was not based
    on a reasonable inference from the evidence presented be-
    cause “the only testimony on the subject was [the defendant’s]
    unrebutted statement that he had no prior criminal record”).
    That is precisely what occurred here. There was thus no rea-
    sonable basis for the state appellate court to conclude that the
    prosecutor’s comments in his closing statement were proper.
    B. Whether the Prosecutor’s Statements Violated Due Pro-
    cess
    Having found the challenged statements improper, we
    turn to whether—considering the record as a whole—the
    prosecutor’s comments deprived Evans of a fair trial. “The rel-
    evant question is whether the prosecutors’ comments so in-
    fected the trial with unfairness as to make the resulting con-
    viction a denial of due process.” United States v. Olson, 
    450 F.3d 655
    , 674 (7th Cir. 2006) (quoting Darden, 
    477 U.S. at 181
    ).
    Six factors guide our inquiry: (1) whether the prosecutor mis-
    stated evidence; (2) whether the remarks implicate specific
    rights of the accused; (3) whether the defense invited the com-
    ments; (4) the trial court’s instructions; (5) the weight of the
    evidence against the defendant; and (6) the defendant’s op-
    portunity to rebut the improper remarks. See Darden, 
    477 U.S. 24
                                                      No. 19-3466
    at 181; see also Howard v. Gramley, 
    225 F.3d 784
    , 793 (7th Cir.
    2000). We do not apply these factors in a rigid manner and
    rely on them only as a “guide to determine whether there was
    fundamental unfairness that infected the bottom line.” Hough
    v. Anderson, 
    272 F.3d 878
    , 903 (7th Cir. 2001). We generally
    consider the weight of the evidence to be “the most important
    consideration.” 
    Id.
     (quoting United States v. Morgan, 
    113 F.3d 85
    , 90 (7th Cir. 1997)).
    As an initial matter, the parties disagree about the level of
    deference we owe the state appellate court’s determination
    that Evans was not denied the right to a fair trial. Evans argues
    that a Darden analysis has two prongs because “we first deter-
    mine if the comments were … improper” and then “[i]f they
    were improper, we consider the record as a whole to deter-
    mine whether the comments deprived the defendant of a fair
    trial.” Olson, 
    450 F.3d at 673
    . According to Evans, because the
    state appellate court never reached the second prong—it de-
    termined only that the comments were proper—we should
    consider whether Evans was denied the right to a fair trial de
    novo. In support, Evans cites to how federal courts review in-
    effective assistance of counsel claims arising under Strickland
    v. Washington, 
    466 U.S. 668
     (1984). A Strickland claim has two
    prongs—a petitioner must show both that his counsel pro-
    vided constitutionally deficient performance (the “perfor-
    mance” prong) and that her was prejudiced by it (the “preju-
    dice” prong). When a state appellate court denies post-con-
    viction relief based on a petitioner’s failure to meet one prong
    but does not reach the merits of the other prong, federal courts
    review the unreached prong de novo. See Wiggins v. Smith, 
    539 U.S. 510
    , 534 (2003). Thus, in Evans’s view, because the state
    appellate court made “no determination” about whether the
    prosecutor’s comments (if improper) prejudiced him, we
    No. 19-3466                                                              25
    need not defer to state court’s determination that he was not
    deprived the right to a fair trial.
    Conversely, the state argues that, unlike Strickland, Darden
    did not set forth a dual-pronged inquiry but instead articu-
    lated a “very general” standard. Parker, 
    567 U.S. at 48
    . We
    have interpreted Darden as having “established a two-prong
    test for determining whether a prosecutors’ comments in clos-
    ing argument constitute a denial of due process.” Ellison, 
    593 F.3d at
    635–36. Nevertheless, according to the state, the rea-
    sonableness of the state appellate court’s adjudication turns
    on clearly established Supreme Court precedent, and there is
    no such precedent establishing a dual-pronged analysis.
    We need not decide this question to resolve this case. As
    described below, even if we owe deference to the state appel-
    late court’s determination, the prosecutor’s comments de-
    prived Evans of a fair trial and any decision to the contrary
    would be an unreasonable application of Darden.
    To be sure, not all of Darden’s factors weigh in Evans’s fa-
    vor. Specifically, the remarks did not “implicate other specific
    rights of the accused such as the right to counsel or the right
    to remain silent,” see Darden, 
    477 U.S. at 182
    , and the trial court
    instructed the jury to disregard any argument made by the
    attorneys that is not based on evidence. 3 Yet the remaining
    four factors tip the scale heavily enough to require that we
    3 Evans concedes that the only “specific right” arguably implicated in
    the prosecutor’s remarks is the defendant’s right to investigate the case
    against him. See Michigan v. Harvey, 
    494 U.S. 344
    , 348 (1990). That right is
    not implicated here, however, as the prosecutor’s comments concerned an
    investigator allegedly working for Evans’s co-defendant, Mario Young,
    not Evans himself.
    26                                                No. 19-3466
    conclude that the prosecution’s improper statements de-
    prived Evans of his right to a fair trial.
    As we have already explained, the prosecutor’s comments
    misstated the evidence. At no point does Jeffers testify that he
    spoke with an investigator working for Mario Young. Instead,
    he testified to the opposite: no one threatened or intimidated
    him, and he never spoke with an investigator working for
    Mario Young. The defense did not invite the prosecutor’s re-
    marks, either. In other words, the prosecutor’s comments can-
    not be seen as offsetting “improper statements from the de-
    fense that might have disposed the jury to favor the defend-
    ant’s position,” United States v. Alexander, 
    741 F.3d 866
    , 871
    (7th Cir. 2014), because the defense did not make such im-
    proper statements. The timing of the statements is particu-
    larly concerning, because the prosecution made the improper
    remarks during rebuttal closing statements—the very last
    time either party would address the jury.
    Most significant to our analysis, however, is the weight of
    the evidence against Evans. The evidence here was not “plen-
    tiful and compelling” as it was in United States v. Klemis, 
    859 F.3d 436
    , 443 (7th Cir. 2017), where “multiple witnesses” tes-
    tified against the defendant, and phone and text records doc-
    umented the defendant’s crime. Instead, the prosecution’s
    only evidence linking Evans to the shooting is Jeffers’s identi-
    fications. The only other eyewitness, Margaret Winton, was
    certain she never saw the shooter’s face and could not de-
    scribe him or either of the other two men in the car. In addi-
    tion, no physical evidence retrieved from the scene of the
    crime was traced to Evans.
    The state contends that the prosecutor’s remarks were not
    prejudicial because the jury would have found Jeffers’s
    No. 19-3466                                                  27
    “multiple out-of-court identifications” to be more credible
    than his trial testimony regardless. We are not persuaded.
    When Jeffers was first asked to identify the perpetrators, he
    had a 50% chance of picking one of the three people the police
    already suspected. In the next meeting, Jeffers was shown a
    photo of Evans again. Therefore, by the time Jeffers picked
    Evans out of an in-person lineup, he had seen his photo—and
    not the photo of any other person in the lineup—twice. Thus,
    the fact that Jeffers “identified” Evans multiple times before
    trial does not mean the evidence against him was compelling.
    Rather, the prosecution’s case rested heavily on a witness
    who initially did not provide specific details of the shooting,
    then told one version of events before trial and a completely
    different version of events at trial. Given the lack of other ev-
    idence tying Evans to the shooting, which version the jury be-
    lieved was crucial to the outcome. Consequently, any reason
    the jury had to credit one version over the other was likely to
    be influential. The prosecutor made Jeffers’s pre-trial version
    of events appear more credible to the jury than his trial testi-
    mony by arguing that Jeffers had been threatened by an in-
    vestigator working for Evans’s co-defendant, when there was
    no evidence in the record supporting that argument. The state
    appellate court acknowledged that “prosecutorial comments
    which suggest that a witness is afraid to testify truthfully be-
    cause of threats or intimidation by the defendant or on the
    defendant’s behalf, when not based upon evidence in the rec-
    ord, are highly prejudicial.” This is exactly what occurred
    here. Accordingly, the prosecutor’s improper comments de-
    prived Evans of his right to a fair trial.
    AFFIRMED
    28                                                  No. 19-3466
    BRENNAN, Circuit Judge, dissenting. The habeas petition
    before us focuses on the closing argument of the prosecutor,
    who theorized that a key witness, Andrew Jeffers, changed
    his testimony because he was intimidated by an investigator
    working for a co-defendant. The parties examine Jeffers’s trial
    testimony to assess whether the prosecutor’s remarks were
    reasonably inferred from the trial evidence. My colleagues
    conclude they are not, and therefore grant the habeas petition.
    Viewing the entire trial, I read Jeffers’s testimony as ad-
    mitting that he spoke with the investigator. That means the
    prosecutor’s closing argument was properly grounded in the
    trial evidence, and the Illinois Appellate Court’s decision rea-
    sonably applied clearly established federal law. Under the
    Antiterrorism and Effective Death Penalty Act of 1996
    (AEDPA), 
    110 Stat. 1214
    , we should defer to that decision.
    I
    A foundational principle of our federal system remains
    that “[s]tate courts are adequate forums for the vindication of
    federal rights.” Burt v. Titlow, 
    571 U.S. 12
    , 19 (2013). Congress
    enacted AEDPA to ensure that federal habeas review of state
    court adjudication is “narrow … and not the broad exercise of
    supervisory power.” Donnelly v. DeChristoforo, 
    416 U.S. 637
    ,
    642 (1974) (internal quotation marks omitted). Under AEDPA,
    a federal court does not have authority to issue a writ of ha-
    beas corpus unless the state court decision (1) “was contrary
    to, or involved an unreasonable application of, clearly estab-
    lished Federal law, as determined by the Supreme Court of
    the United States” or (2) “was based on an unreasonable de-
    termination of the facts in light of the evidence presented in
    the State court proceeding.” 
    28 U.S.C. § 2254
    (d).
    No. 19-3466                                                     29
    A state court unreasonably applies clearly established fed-
    eral law if “it correctly identifies the governing legal rule from
    Supreme Court case law, but unreasonably applies it to the
    facts of the case.” Clark v. Lashbrook, 
    906 F.3d 660
    , 664 (7th Cir.
    2018) (citing Williams v. Taylor, 
    529 U.S. 362
    , 407–08 (2000)). A
    federal court reviewing habeas under the “unreasonable ap-
    plication” prong of § 2254(d)(1) must first “determine what
    arguments or theories supported or, … could have sup-
    ported, the state court’s decision.” Harrington v. Richter, 
    562 U.S. 86
    , 102 (2011). The court then must ask “whether fair-
    minded jurists could disagree on the correctness of the state
    court’s decision if based on one of those arguments or theo-
    ries.” Shinn v. Kayer, 
    141 S. Ct. 517
    , 524 (2020) (internal quota-
    tion marks omitted) (citing Richter, 
    562 U.S. at 101
    ). A peti-
    tioner can satisfy this inquiry “only by showing that ‘there
    was no reasonable basis’” for the state court’s decision. Cullen
    v. Pinholster, 
    563 U.S. 170
    , 188 (2011) (quoting Richter, 
    562 U.S. at 98
    ). The state court’s factual determination is “presumed to
    be correct,” and the petitioner bears the burden of rebutting
    that presumption “by clear and convincing evidence.” 
    28 U.S.C. § 2254
    (e)(1).
    AEDPA sets a high bar for when a state court decision con-
    stitutes an unreasonable application of Supreme Court prece-
    dent. To meet this bar, a state court decision must be “so lack-
    ing in justification … beyond any possibility for fairminded
    disagreement.” Schmidt v. Foster, 
    911 F.3d 469
    , 477 (7th Cir.
    2018) (en banc) (quoting Richter, 
    562 U.S. at 103
    ). Habeas relief
    is appropriate solely when “state courts veer well outside the
    channels of reasonable decision-making about federal consti-
    tutional claims.” 
    Id.
     (internal quotation marks omitted). This
    standard is “difficult to meet.” White v. Woodall, 
    572 U.S. 415
    ,
    419 (2014) (internal quotation marks omitted). That is because
    30                                                    No. 19-3466
    AEDPA deference directs federal courts to “presum[e] that
    state courts know and follow the law.” Woodford v. Visciotti,
    
    537 U.S. 19
    , 24 (2002) (per curiam).
    The clearly established federal law in this case, Darden v.
    Wainwright, provides the relevant “framework to evaluate
    ‘whether the prosecutors’ comments so infected the trial with
    unfairness as to make the resulting conviction a denial of due
    process.’” Bartlett v. Battaglia, 
    453 F.3d 796
    , 800 (7th Cir. 2006)
    (quoting Darden, 
    477 U.S. at 181
    ). To succeed under Darden, a
    petitioner must prove that the prosecutor’s remarks were im-
    proper and that those statements deprived him of a fair pro-
    ceeding. Darden, 
    477 U.S. at 181
    .
    A prosecutor’s closing remarks unsupported by trial evi-
    dence are improper. See Berger v. United States, 
    295 U.S. 78
    , 88–
    89 (1935). Federal courts defer to state courts in making this
    assessment. Cf. Miller-El v. Cockrell, 
    537 U.S. 322
    , 340 (2003). A
    state appellate court’s decision that prosecutorial statements
    were supported by evidence “will not be overturned on fac-
    tual grounds unless objectively unreasonable in light of the
    evidence presented in the state-court proceeding.” 
    Id.
     This
    deference is necessary because “[a] criminal trial does not un-
    fold like a play with actors following a script,” Geders v. United
    States, 
    425 U.S. 80
    , 86 (1976), and habeas review cannot take
    into account all the nuances of live testimony. Though “rea-
    sonable minds reviewing the record might disagree about the
    finding in question, on habeas review that does not suffice to
    supersede the trial court’s” determination. Brumfield v. Cain,
    
    576 U.S. 305
    , 314 (2015) (cleaned up). So “even a strong case
    for relief does not mean the state court’s contrary conclusion
    was unreasonable.” Richter, 
    562 U.S. at 102
    .
    No. 19-3466                                                    31
    A prosecutor may comment on the credibility of a witness
    if the remarks were reasonably inferred from the evidence at
    trial. United States v. Klemis, 
    859 F.3d 436
    , 443 (7th Cir. 2017).
    This court has recognized that “[a]ttorneys have more leeway
    in closing arguments to suggest inferences based on the evi-
    dence.” Soltys v. Costello, 
    520 F.3d 737
    , 745 (7th Cir. 2008). A
    prosecutor’s inference “need not always be introduced, nor
    immediately followed, by a direct reference to the trial rec-
    ord.” United States v. Wolfe, 
    701 F.3d 1206
    , 1213 (7th Cir. 2012).
    Rather, reasonable inference must be “defined contextually.”
    United States v. Waldemer, 
    50 F.3d 1379
    , 1384 (7th Cir. 1995).
    That context should dispose of this case.
    II
    The appropriate context from which the prosecutor may
    render a reasonable inference is the entire trial, not just por-
    tions of it. See United States v. Young, 
    470 U.S. 1
    , 11–12 (1985)
    (explaining that prosecutor’s comments or conduct must be
    viewed in “the context of the entire trial”).
    That context includes the following exchanges. On the first
    re-direct examination, the prosecutor asked Jeffers whether
    he had spoken to an investigator working for Young. Jeffers
    answered no. On re-cross, defense counsel followed up and
    this time Jeffers admitted that he did speak to an investigator.
    Then the prosecutor readdressed Jeffers’s encounter with the
    investigator. Jeffers responded that he did not know the iden-
    tity of the investigator and that the investigator did not visit
    his home. In his closing argument, the prosecutor offered a
    theory on why Jeffers recanted. He theorized that Jeffers had
    given a complete story when he was in custody because he
    felt safe, and later recanted his testimony upon a visit from
    Young’s investigator because he was intimidated.
    32                                                             No. 19-3466
    Based on this sequence, all can agree that Jeffers spoke to
    an individual who had visited him after he was released from
    custody but before trial. And all can agree that the individual
    was likely an investigator. What is unclear, the majority opin-
    ion concludes, is whether Jeffers knew that the investigator
    was someone working for Young.
    The majority opinion rejects the decision of the Illinois Ap-
    pellate Court—that Jeffers admitted to speaking with Young’s
    investigator—as lacking evidence, pointing to portions of Jef-
    fers’s testimony. On re-cross, defense counsel asked Jeffers,
    “The state’s attorney just asked if you spoke to an investigator
    for Mario Young when you were released from custody, is
    that correct, sir?” Jeffers responded, “Right.” Though the
    majority opinion acknowledges that Jeffers’s responses at var-
    ious parts of the testimony “make the transcript somewhat
    difficult to follow,” it surmises that “there is no reasonable
    reading of this exchange in which Jeffers admits to speaking
    with an investigator working for Mario Young.” (emphasis
    added).
    The context of the full trial weakens this conclusion.
    Throughout the trial, Jeffers used “right” and “yeah” inter-
    changeably to answer in the affirmative. 1 So when he replied
    1   Below are some examples from Jeffers’s testimony:
    “Q: As of ‘96 when this happened, you had been living there for about
    five or six years? A: Right.” R. 13-8 at 89–90.
    “Q: You’re saying you heard a total of four or five shots altogether? A:
    Right. Q: That is all you heard? A: Right.” 
    Id. at 95
    .
    “Q: You don’t write your name that way, AJ? A: Right.” 
    Id. at 104
    .
    No. 19-3466                                                          33
    to defense counsel’s question on re-cross with “right,” Jeffers
    could have meant either:
    1. Yes, it is correct that the state’s attorney just
    asked if I spoke to an investigator for Mario
    Young when I was released from custody; or
    2. Yes, it is correct that I spoke to an investiga-
    tor for Mario Young when I was released
    from custody.
    The first inference focuses on the state attorney’s question; the
    second inference focuses on the substance of that question.
    Both are reasonable readings of the exchange, and the latter
    provides a reasonable basis to infer that Jeffers admitted to
    speaking with an investigator working for Young.
    The point of the “reasonable inference” standard is to cap-
    ture the nuances of the trial—what parties attempt to com-
    municate and to understand. From Jeffers’s testimony, it can
    logically be concluded that he spoke to Young’s investigator.
    Inference number 1 is reasonable, but that does not render in-
    ference number 2 unreasonable. Indeed, given the sequence
    of the questions, inference number 2 is arguably more reason-
    able than inference number 1.
    The majority opinion concludes that the state court’s de-
    termination “lacks evidentiary support.” In doing so, it high-
    lights that there is “[a] lack of evidence about who the
    “Q: And on the bottom of page 6 it says Andrew Jeffers there, right?
    A: Right. Q: And you’re saying you didn’t put that signature there
    either, right? A: Yeah.” 
    Id. at 108
    .
    “Q: And you didn’t put that there either, right? A: Right. Q: And two
    lines below that, right? A: Yeah.” 
    Id. at 112
    .
    34                                                   No. 19-3466
    investigator worked for” and that “the only testimony about
    her employer is that she did not work for Young.”
    But by dismissing the state court’s interpretation of the
    testimony, the majority opinion applies an overly restrictive
    definition of “reasonable inference.” To be a “reasonable in-
    ference” does not require Jeffers to have explicitly said that he
    knew the investigator worked for Young. Black’s Law Diction-
    ary defines “inference” as “[a] conclusion reached by consid-
    ering other facts and deducing a logical consequence from
    them.” BLACK’S LAW DICTIONARY (11th ed. 2019). That Jeffers
    spoke to Young’s investigator is a logical conclusion:
       Jeffers decided to recant his prior statements
    and identification after his release from custody
    and before trial; an investigator visited Jeffers
    during that time;
       Young had an interest in the case as a co-defend-
    ant; and
       Jeffers responded in the affirmative when de-
    fense counsel asked whether he had spoken to
    an investigator working for Young.
    In response to this conclusion, the majority opinion under-
    scores that five times Jeffers denied meeting with Young’s in-
    vestigator. Such denials do not automatically amount to clear
    and convincing evidence. This is especially true if the state
    court renders its factual determination based on a reasonable
    inference from the trial record, as the Illinois Appellate Court
    did here.
    Evaluating whether an inference is reasonable is not an ex-
    ercise in keeping score. Five explicit denials do not obviate the
    weight of a single inference of admission to the contrary. Cf.
    No. 19-3466                                                      35
    United States v. Edwards, 
    581 F.3d 604
    , 612 (7th Cir. 2009) (not-
    ing that “the trier of fact must consider whether … particular
    falsehoods in a witness’s testimony so undermine his credi-
    bility as to warrant disbelieving … a critical part” of his testi-
    mony); United States v. Mejia, 
    82 F.3d 1032
    , 1038 (11th Cir.
    1996) (“A proper inference the jury can make from disbe-
    lieved testimony is that the opposite of the testimony is
    true.”).
    One last point on this score. The majority opinion empha-
    sizes that Jeffers “expressly stated … he did not speak to an
    investigator who worked for Mario Young.” But Jeffers only
    denied knowing the identity of the investigator who “just
    asked [him] questions,” not speaking with an investigator. In-
    deed, the majority opinion acknowledges that Jeffers said he
    “spoke to someone but he did not know if the investigator
    worked for Young.” Jeffers neither did nor could have ex-
    pressly denied speaking with Young’s investigator because
    he did not know that investigator’s identity.
    AEDPA “demands that state-court decisions be given the
    benefit of the doubt.” Renico v. Lett, 
    559 U.S. 766
    , 773 (2010)
    (quoting Woodford, 
    537 U.S. at 24
    ). The Illinois Appellate
    Court’s decision does not simply “rest[] on thin air.” Mendiola
    v. Schomig, 
    224 F.3d 589
    , 592 (7th Cir. 2000). It relies on a rea-
    sonable inference from evidence presented at trial—that Jef-
    fers admitted to speaking with an investigator working for
    Young—to conclude that the prosecutor’s comments were not
    improper. And when the record is subject to multiple inter-
    pretations, as here, federal courts should defer to state court’s
    reasonable interpretation. See, e.g., Bartlett, 
    453 F.3d at 802
    . Be-
    cause the full context of the trial provides a reasonable basis
    to characterize the prosecutor’s statements as not improper,
    36                                                   No. 19-3466
    the Illinois Appellate Court did not unreasonably apply
    Darden under § 2254(d)(1). We should defer to the state
    court’s decision, and bound by AEDPA, deny the petition.
    ***
    Under AEDPA, federal courts must not mistake habeas
    corpus as “a substitute for ordinary error correction through
    appeal.” Richter, 
    562 U.S. at
    102–03. Rather, it is “[a] guard
    against extreme malfunctions in the state criminal justice sys-
    tems.” Jackson v. Virginia, 
    443 U.S. 307
    , 332 n.5 (1979) (Stevens,
    J., concurring in judgment). No such breakdown happened
    here. The decision of the Illinois Appellate Court rested on
    record evidence, so under AEDPA I would deny the petition.
    I respectfully dissent.