United States v. Hill , 749 F.3d 1250 ( 2014 )


Menu:
  •                                                                                 FILED
    United States Court of Appeals
    PUBLISH                              Tenth Circuit
    UNITED STATES COURT OF APPEALS                       April 28, 2014
    Elisabeth A. Shumaker
    TENTH CIRCUIT                             Clerk of Court
    UNITED STATES OF AMERICA,
    Plaintiff - Appellee,
    v.                                                          No. 12-5154
    STANLEY EUGENE HILL,
    Defendant - Appellant.
    Appeal from the United States District Court
    for the Northern District of Oklahoma
    (D.C. No. 4:11-CR-00179-GKF-2)
    Howard A. Pincus, Assistant Federal Public Defender (Warren R. Williamson, Federal
    Public Defender, Interim, with him on the briefs), Office of the Federal Public Defender,
    Denver, Colorado, for the Defendant-Appellant.
    Joel-lyn Alicia McCormick, Assistant United States Attorney (Danny C. Williams, Sr.,
    United States Attorney and Leena Alam, Assistant United States Attorney, with her on
    the briefs), Office of the United States Attorney, Tulsa, Oklahoma, for the Plaintiff-
    Appellee.
    Before LUCERO, MURPHY, and MATHESON, Circuit Judges.
    LUCERO, Circuit Judge.
    Stanley Hill appeals following his conviction on several charges related to the
    robbery of a bank. During trial, Charles Jones, a special agent with the Federal Bureau of
    Investigation (“FBI”), testified as an expert. Agent Jones stated that he was trained in
    “special tactics and ways to identify [] deception in statements and truths in statements”
    and that in his opinion, many of Stanley’s1 answers were not worthy of credence and
    “[did] not make sense.” Jones claimed that Stanley displayed evasive behaviors
    “common among the criminal element to keep law enforcement at bay” during an
    interrogation. When asked about Stanley’s statement that he would rather die than face
    charges, Jones testified, “Never in my career have I seen that with an innocent person.”
    And when the prosecutor asked about Stanley’s repeated invocations of God in support of
    his truthfulness, Jones stated, “My training has shown me, and more[ ]so my experience
    in all these interviews, when people start bringing faith into validating [] their statements,
    that they’re deceptive. Those are deceptive statements.”
    Stanley did not contemporaneously object to the admission of this evidence.
    Nevertheless, we conclude the court plainly erred in admitting this testimony and, in light
    of the relative weakness of the government’s overall case, that it affected Stanley’s
    substantial rights. We further conclude that this is one of the exceptional cases in which
    we exercise our discretion to notice the plain error because failing to do so would
    seriously undermine the fairness, integrity, or public reputation of judicial proceedings.
    1
    Because this case involves several individuals with the surname Hill, we will
    refer to these individuals by their first names.
    -2-
    Exercising jurisdiction under 
    28 U.S.C. § 1291
    , we reverse.
    I
    Stanley Hill and his brother, Vernon Hill, were charged with bank robbery with a
    dangerous weapon and use of a firearm during and in relation to a crime of violence
    following the November 5, 2011, robbery of an Arvest Bank in Tulsa, Oklahoma.
    Stanley was also charged with being a felon in possession of a firearm. Following a jury
    trial, Vernon was convicted of both charges against him. The jury deadlocked on the
    charges against Stanley, and the district court declared a mistrial.
    The government proceeded at the first trial upon the theory that Vernon and
    Stanley were the two men who physically entered and robbed the bank. After
    discovering additional evidence, however, the government came to believe that Vernon
    and another Hill brother, Dejuan, were the two individuals who entered the bank while
    Stanley acted as a getaway driver. Stanley and Dejuan were charged by superseding
    indictment with a Hobbs Act violation, bank robbery with a dangerous weapon, and use
    of a firearm during and in relation to a crime of violence. Stanley was also charged a
    second time with being a felon in possession of a firearm.
    A
    At the second trial, the government adduced testimony from several individuals
    who witnessed the robbery. A customer of the bank testified that she observed two men
    wearing hoodies and gloves standing near the bank just before the robbery, one of whom
    was talking on a cell phone. Bank employees testified that two masked men entered the
    -3-
    bank shortly after it opened, ordered the occupants to the floor, and demanded money.
    One of the robbers brandished a firearm, identified by one of the employees as a 9mm.
    The thieves put money into a tan-colored pillow case and fled west on foot. A witness
    who lived near the bank testified that he saw two men running down the street carrying a
    white bag shortly after the robbery.
    The stolen cash included “bait bills” containing a GPS tracking device. This
    device automatically activated upon being removed from a teller’s drawer and sent a
    notification to a Tulsa police sergeant at 8:30 or 8:31 a.m. The speed of the tracking
    device indicated it was being transported in a vehicle. Within a few minutes, it came to a
    stop in the vicinity of 1100 East Pine in Tulsa. A Tulsa police officer testified that within
    five to ten minutes of being notified by dispatch of the bank robbery, he arrived in that
    area and focused on a residence at 1107 East Pine (the “East Pine residence”). Shortly
    after he arrived on the scene, the officer reported that a black Nissan Altima left the
    driveway of that home. The driver was later identified as Dejuan. No other person
    entered or exited the home during the officer’s surveillance. The tracking device
    indicated that it was located within the East Pine residence.
    After approximately two hours, officers observed Stanley exiting the home and
    took him into custody. A short time later, Vernon came outside and was also detained.
    The brothers were transported separately to a Tulsa police station. Stanley identified
    himself as Daniel Hill and provided a birth date and social security number. After pulling
    up a picture of Daniel Hill and noticing the absence of a tattoo on Stanley’s arm, a
    -4-
    detective confronted Stanley about his identity. Stanley began crying and asked to speak
    with the detective in a room out of Vernon’s earshot. After being moved to a separate
    interview room, Stanley identified himself truthfully.
    Tulsa Police Corporal Christopher Stout and Agent Jones conducted a videotaped
    interview with Stanley, which was played for the jury. In his interview, Stanley stated
    that he woke up at his girlfriend Whitney Landrum’s home although he claimed not to
    know her address. He said that after waking up he went to the home at 1107 East Pine,
    which Stanley identified as belonging to his father, Stanley Battle. He claimed that no
    one was present when he arrived at roughly 6:00 a.m., and that he watched television and
    fell asleep in the living room shortly thereafter. Stanley said that he planned to watch his
    stepsister, who was going to be dropped off at the home by her mother sometime that
    morning. He was unsure of the step-sister’s exact age, stating that she was about 11, and
    said he did not know the name of her mother. Stanley did not know the exact time she
    was going to be dropped off.
    Stanley claimed he was awoken by the house phone several hours later and was
    informed there were police outside. He went outside and was taken into custody.
    Stanley claimed that an officer told him that another man had left the house, and said that
    he did not hear anyone come or go while he was sleeping. He stated in the interview that
    he was not sleeping very deeply, and would have known if anyone else was in the home.
    After hearing Stanley’s story, the interviewers challenged his version of the facts.
    They explained that a bank has been robbed and that material from the bank was found in
    -5-
    the East Pine residence. They suggested that if he was the only one in the house, he was
    likely the bank robber. Stanley then acknowledged that his brother exited the home after
    he did, but said he did not know anyone else was in the home. He denied involvement in
    the bank robbery. When officers expressed disbelief, he repeatedly stated that he would
    swear on the Bible, and swore to God that he was telling the truth. He also stated that he
    did not want to live and that he would rather die than face charges.
    Meanwhile, Tulsa Police obtained a search warrant for the East Pine residence.
    Officers discovered a tan-colored pillowcase full of cash in the bottom drawer of the
    oven, including the tracking device taken from the Arvest Bank. An officer testified that
    “the drawer was very full of pots and pans” and that he could hear officers “struggling to
    get that drawer open with pots and pans in the way” from the living room. The living
    room was located fifteen to eighteen feet from the kitchen. Officers also found a Glock
    .45 caliber pistol, a pair of black pants, a black ski mask, and two pairs of black gloves in
    a bedroom of the house. They discovered mail addressed to Vernon and Battle.
    Corporal Stout testified about the difficulties he encountered attempting to contact
    Landrum. He stated that Landrum repeatedly hung up on him, and that Landrum resisted
    contact when served with a subpoena. He also learned that Landrum drove a black 2004
    Nissan Maxima. Landrum was subpoenaed, asserted her Fifth Amendment rights, and
    was granted immunity. She testified that Stanley is her boyfriend and the father of two of
    her children. Landrum further testified that Stanley left her home early in the morning on
    the day of the robbery and that she went to the East Pine residence sometime around 8:00
    -6-
    or 9:00 a.m. to pick him up. By the time she arrived in her Nissan, police were already
    on the scene.
    The prosecution also introduced two phone calls made by Stanley from jail. In a
    discussion between Stanley and Landrum about bond money, Landrum relays that
    Vernon referred to Stanley as “the weak link of the group” and that he suggested Stanley
    be bailed out first if possible. In another call between Stanley and his father, Battle warns
    Stanley not to say “nothing to do with what they holding y’all for . . . don’t say where
    you was.” Battle notes that the police might suggest other people had provided
    information, and says, “you know we’re family, we stick together, so you don’t have to
    worry about the police talking about who told what.”
    B
    A significant portion of the government’s case involved cell phone evidence.
    Neither Vernon nor Stanley was carrying a cell phone when they were taken to the police
    station. Landrum testified that her cell phone number ended in “3860” and that she was
    not sure where her phone was at the time of the robbery. She also stated that she spoke
    with Vernon and Dejuan by phone regularly. A phone company employee testified that
    Vernon subscribed to a cell phone number ending in “9204,” and that Landrum
    subscribed to two different numbers: the number ending in 3860, and another phone
    number ending in “1576.” Landrum told investigators that she did not know which
    phone Stanley was using on the day of the robbery, but she was certain it was not
    Vernon’s phone. Stanley provided the 1576 number as a contact number to officers on
    -7-
    the day of the robbery. Dejuan also listed the 1576 number as his contact information
    when visiting his brothers in jail.
    Another special agent with the FBI testified about cell phone tower data. He
    explained that cell phone towers typically include three separate receivers, each of which
    provides cellular service to a wedge-shaped “sector” emanating from the tower. Each
    sector encompasses 120 degrees, meaning that a tower provides 360 degrees of coverage.
    Phone companies maintain data on the tower and sector to which a cell phone connects
    when making or receiving a call. The agent testified that in densely populated areas, cell
    phone towers are designed to have some overlap such that a phone might connect to two
    different towers from the same location, but that the data provides a “general area” from
    which a call was placed or received. The agent prepared several exhibits showing the
    locations of various towers in the Tulsa region, superimposed with highlighted areas
    depicting an approximate range for each relevant tower sector. He stated that the angles
    of the sectors depicted in the maps were “definitive” but that the distance from the tower
    was a “general depiction” based on his training and experience. The maps show that
    Arvest Bank is located to the southeast of the East Pine residence, and that Landrum’s
    home is between the two, not far from the bank.
    The cell tower data showed that Vernon’s phone, ending in 9204, contacted the
    -8-
    tower nearest Arvest Bank on the evening prior to the robbery.2 It was located in the
    sector that includes Landrum’s home at 6:47 a.m. on November 5, the day of the robbery.
    The phone then received a call while located near the East Pine residence at 7:09 a.m. At
    8:19 and 8:30, the phone contacted two towers near Arvest Bank. Seconds before 8:31,
    and just before the robbery, Vernon’s phone made a call that contacted the tower nearest
    Arvest Bank. At 8:53 and 9:07, Vernon’s phone was again located in the vicinity of the
    East Pine residence.
    The phone number ending in 3860 also provided detailed information. That phone
    received a call from Vernon’s phone at 6:47 a.m. and it too was located near Landrum’s
    home. At 7:15, it received a call from the 1576 phone and was still located in the vicinity
    of Landrum’s home. At 8:19, Vernon’s phone called the 3860 number, and both phones
    contacted the same tower near the Arvest Bank. The two calls made from Vernon’s
    phone at 8:30 were both to the 3860 number, and the two phones contacted the same
    towers: the first call connected the phones through a tower near Arvest Bank, and the
    second call went through the tower nearest the bank. At 9:04, the 3860 phone connected
    to a tower located southwest of the East Pine residence. The 3860 phone exchanged
    several calls with Vernon’s phone from several different towers, apparently while
    travelling in a southerly direction.
    The data on the phone ending in 1576 was more limited. It contacted a tower near
    2
    The government’s evidence indicated that the area just west of Arvest Bank was
    served by up to three separate towers.
    -9-
    Arvest Bank the night before the robbery. Between 7:09 and 7:14 a.m. the morning of
    the robbery, that phone exchanged several calls with Vernon’s phone while both were
    located near the East Pine residence. At 7:15 and 7:16, the 1576 number was near the
    East Pine residence when it exchanged calls with the 3860 number, which as noted
    above, was in the vicinity of the Landrum home. At 8:07, the 1576 number called an
    unknown phone number from the vicinity of the Landrum home, and was turned off
    shortly thereafter.
    Special Agent Jones provided testimony as the case agent on the government’s
    theory of the case. Jones testified that he “believe[d] Stanley Hill was the getaway driver
    waiting for Dejuan and Vernon to exit the bank and get in the vehicle as he drove them to
    1107 E. Pine after the bank robbery.” With respect to the cell phone data, Jones opined
    that Dejuan was carrying the 1576 phone and Vernon the 9204 phone on the night before
    the robbery, and were “casing” the bank when those phones contacted the tower nearest
    the bank on the evening of November 4. On the morning of the robbery, Jones believed
    that Stanley had the 3860 phone, and was using it to communicate with Vernon and
    Dejuan through Vernon’s phone. Jones stated that he believed Dejuan was carrying the
    1576 phone on November 5, but had it switched off during the time of the robbery.
    According to Jones, Dejuan obtained the 3860 phone shortly after the robbery, which
    explained the southward movement of that phone during the time that Stanley and
    Vernon were known to be inside the East Pine residence.
    -10-
    C
    Jones also testified about the interrogation of Stanley. He stated that he had
    attended “two specialized courses in interrogation and interviews, including the Reid
    school, which is a higher-level school of interrogation and interviewing.” He explained:
    [T]he Reid school is designed to – as an interview process and interrogation
    process; part of that is psychological as well. It’s much like your five-year-
    old children and how you can break down a story or you understand what’s
    going on during the process of that interview.
    In the Reid school, you’re trained on some special tactics and ways
    to identify on deception in statements and truths in statements. That school
    is a sought-after school for investigators and interviewers because of the
    caliber of that training you do get towards that endeavor.
    Jones further stated that he had conducted over a thousand interviews as an FBI agent.
    The prosecution asked, “in reference to [his] earlier testimony regarding [his]
    training and experience in interrogating and interviewing,” what Jones “based on [his]
    training and experience” took from the interrogation “as to [Stanley’s] truthfulness.”
    Jones responded:
    [T]he most difficult thing to tell the difference in is partial truths, . . .
    something that’s partly true, that’s a lot harder to detect than a flat-out lie or
    a convicting [sic] truth.
    So during the course of that interview, we were able to, as trained
    eyes, pick out that this isn’t – these are partial truths, at best. And several
    of those are – they’re shown through things that are not purposely said or
    done by the interviewee. They are responses that occur naturally, that’s a
    psychological thing that happens, that we don’t control.
    For example, in this case, and I’ve seen it in other interviews, a
    mumbling of something that they don’t want to talk about. You may say, I
    was at the grocery store at three or whatever or whatever, and you will go
    away from the question and just discount that as something you don’t need
    to know, Mr. Police Officer. And there was much of that going on
    throughout the interview, for whatever, or whatever, and whatever with Mr.
    -11-
    Stanley Hill’s interview, occurred on a continuous basis, just avoiding – it’s
    a way to avoid the question without just flat out saying, I’m not going to
    talk to you.
    The prosecutor then asked, “In reference to the substance of the responses that
    were provided . . . how does that factor into your observations of whether he’s being
    truthful or not?” She provided Stanley’s claim that he planned to babysit his step-sister at
    the East Pine home as an example of “the substance of responses.” Jones answered that
    Stanley’s version of events “does not connect [the] dots,” “does not make sense,” and
    was “not something that [he] viewed as reasonable.” Specifically, Jones doubted that if
    Stanley was going to be “responsible for a child,” he would immediately fall asleep “and
    never w[a]ke up while somehow bank robbery money got stuffed in the oven drawer of
    your house, and then the bank robbers ran away before the police could get there, and you
    didn’t hear anything, but you were waiting on somebody to arrive in this unlocked house
    in north Tulsa.”
    Jones then identified several factors that contributed to his opinion that Stanley
    was being untruthful during the interview. He noted that after Stanley was told that
    police found items connected to the bank robbery in the East Pine residence, Stanley’s
    story ha[d] to change a little bit. And prior to that, I wasn’t sleeping that
    hard. After that, “to my knowledge,” “to my knowledge.”
    I can’t question his knowledge. I cannot say, I know what you
    knew. But he could evade the question by saying, “well, to my
    knowledge,” because that’s something I cannot corroborate. That is a move
    that is common among the criminal element to keep law enforcement at bay
    and not be able to determine the actual facts of what happened.
    Jones also stated that Stanley’s assertions that he “had no will to live” were
    -12-
    indicative of guilt, testifying: “I have not seen, in my experience, an innocent person
    willing to die because they were talking with police officers and FBI agents. Never in
    my career have I seen that with an innocent person.” He continued: “I also don’t
    reasonably believe an innocent person would want to die because they were being talked
    to by police officers. It doesn’t make sense to me.” The prosecutor asked if, “in [Jones’]
    experience, has it been a demonstration of consciousness of guilt that an individual will
    want to die rather than tell the truth.” Jones responded, “In my experience, sometimes
    people believe death would be better than a long-term prison sentence.”
    The prosecution also asked how Jones viewed Stanley’s “call on his faith or
    swearing to God” during the course of the interview. Jones testified:
    Beyond my own religious feelings towards what he was saying, the training
    that I’ve received, that is a common way that somebody with guilt will
    want to validate the story they’re telling you. They can’t validate it with
    facts, so they hope they can get you to believe them, because they’re trying
    to validate their story through a supposed belief.
    He may be a God-fearing man, I do not know that, but the truth is
    the truth. You do not have to back the truth. When I’m asked a question, is
    the car blue, the car is blue. I don’t have to swear to God. I do not have to
    bring religion into that statement. The truth is the truth.
    My training has shown me, and more[ ]so my experience in all these
    interviews, when people start bringing faith into validating of their
    statements, that they’re deceptive. Those are deceptive statements.
    Defense counsel did not make any objections during Jones’ testimony about the
    truthfulness of Stanley’s statements. Jones was the final witness at trial. During closing
    argument, the prosecutor referred the jury back to Jones’ testimony, stating that Jones “in
    scrutinizing this interview with Stanley Hill . . . has to figure out what’s truthful in this
    -13-
    interview, what’s he trying to hide.”
    D
    The jurors were instructed that they are “the judges of the facts” and that they are
    “the sole judges of the credibility or ‘believability’ of each witness.” The instructions
    specifically discussed the witness “who expressed opinions concerning cell site tower
    analysis” and Jones, “who expressed opinions regarding investigation of conspiracies and
    robberies.” This instruction stated that “scientific, technical, or other specialized
    knowledge may assist the jury in understanding the evidence or in determining a fact in
    issue.” It also explained that “[a] witness who has such knowledge, skill, experience,
    training or education may testify and state an opinion concerning such matters.” The jury
    was cautioned that it was “not required to accept such an opinion” and was instructed to
    give the opinion testimony “as much weight as you think it deserves, considering the
    education and experience of the witness, the soundness of the reasons given for the
    opinion, and other evidence.” With respect to Stanley’s out-of-court statement, the jury
    was instructed to consider various factors in “determining whether [this] statement is
    reliable and credible.”
    The jury deliberated for several hours. After submitting two questions to the
    judge, the jury returned a verdict of guilty on all counts. The district court sentenced
    Stanley to 154 months’ imprisonment. He timely appealed.
    II
    Stanley advances a single argument on appeal: that the district court erred in
    -14-
    permitting Jones to provide expert opinion testimony as to Stanley’s credibility. Because
    Stanley did not object below, we review only for plain error. See United States v. Frost,
    
    684 F.3d 963
    , 971 (10th Cir. 2012). “Plain error occurs when there is (1) error, (2) that is
    plain, which (3) affects substantial rights, and which (4) seriously affects the fairness,
    integrity, or public reputation of judicial proceedings.” 
    Id.
     (quotation omitted); see also
    Fed. R. Crim. P. 52(b) (“A plain error that affects substantial rights may be considered
    even though it was not brought to the court’s attention.”).
    A
    “Error is ‘plain’ if it is obvious or clear, i.e., if it is contrary to well-settled law.”
    United States v. Edgar, 
    348 F.3d 867
    , 871 (10th Cir. 2003) (quotation omitted). “In
    general, for an error to be contrary to well-settled law, either the Supreme Court or this
    court must have addressed the issue.” United States v. DeChristopher, 
    695 F.3d 1082
    ,
    1091 (10th Cir. 2012) (quotation omitted). However, in certain circumstances, the
    “weight of authority from other circuits” may make an error plain even absent a holding
    from this court or the Supreme Court. United States v. Hardwell, 
    80 F.3d 1471
    , 1484
    (10th Cir. 1996). We conclude that the admission of Jones’ opinion testimony was error,
    and that it was plain.
    As a general matter, expert opinion testimony may be introduced at trial if it “will
    help the trier of fact to understand the evidence or to determine a fact in issue.” Fed. R.
    Evid. 702(a). An expert may provide opinion testimony if: “the testimony is based upon
    sufficient facts or data” and “is the product of reliable principles and methods,” and “the
    -15-
    expert has reliably applied the principles and methods to the facts of the case.” Fed. R.
    Evid. 702(b)-(d). “The touchstone of admissibility under Rule 702 is the helpfulness of
    the evidence to the trier of fact.” United States v. Rangel-Arreola, 
    991 F.2d 1519
    , 1524
    (10th Cir. 1993).
    Stanley does not argue that Jones was unqualified to offer the opinion he provided,
    but instead that the subject matter of his testimony—the credibility of another person—
    may not be addressed by an expert testifying under Rule 702. We agree. As this court
    made clear in United States v. Toledo, 
    985 F.2d 1462
     (10th Cir. 1993), “[t]he credibility
    of witnesses is generally not an appropriate subject for expert testimony.” 
    Id. at 1470
    .
    There are several reasons for the prohibition against expert testimony on other witness’
    credibility. Such testimony: (1) “usurps a critical function of the jury”; (2) “is not
    helpful to the jury, which can make its own determination of credibility”; and (3) when
    provided by “impressively qualified experts on the credibility of other witnesses is
    prejudicial and unduly influences the jury.” 
    Id.
     (citations omitted).
    In Toledo, we considered testimony from two psychiatrists who had examined the
    victim of a kidnapping and sexual assault. One stated that “[n]o matter how crazy people
    are or how psychotic people are, we, as therapists, try to look for that window of
    normalcy; and there were certain little windows of normalcy that things she told me, I
    knew were very true.” 
    Id. at 1469
     (emphasis omitted). Another testified, “In my opinion,
    it appears that what occurred was: One, she was off her medication; [and] two, she
    underwent emotional trauma that contributed to her psychotic symptoms.” 
    Id.
     (emphasis
    -16-
    omitted). The same witness, when asked to distinguish between “what’s delusional from
    what is accurate history as reported by the patient,” responded that “delusions are fixed
    false beliefs which have no basis in reality and histories are – well, if it makes sense, then
    its [sic] not a delusion.” 
    Id.
     He then concluded that portions of the victim’s story “were
    consistent with a high likelihood that [an abduction] occurred.” 
    Id.
     (emphasis omitted).
    In considering the admissibility of this testimony, we discussed at length an Eighth
    Circuit case, United States v. Azure, 
    801 F.2d 336
     (8th Cir. 1986), in which the court
    reversed a conviction after a pediatrician testified that the victim “was believable and that
    he could ‘see no reason why she would not be telling the truth. . . .’” 
    Id. at 339
    . The
    district court concluded that this statement “was admissible under Fed. R. Evid. 702 as an
    expert opinion”; the Eighth Circuit disagreed. 
    Id. at 339, 340-41
    . The Toledo panel
    focused on the Eighth Circuit’s statement that an expert might permissibly testify “about
    a child’s ability to separate truth from fantasy, [or] by summarizing the medical evidence
    and expressing his opinion as to whether it was consistent with [the victim’s] story that
    she was sexually abused.” 
    985 F.2d at 1470
     (quoting Azure, 
    801 F.2d at 340
    ). We also
    noted another Eighth Circuit decision that upheld a conviction in a case involving a child
    psychologist’s testimony “that it was not unusual for a child victim of sexual abuse to
    link two traumatic events together, even though they did not actually occur at the same
    time” and that “‘whether or not they were contingent in time they both occurred.’” 
    Id.
    (quoting United States v. Provost, 
    875 F.2d 172
    , 176 (8th Cir. 1989) (emphasis omitted)).
    In light of these two cases, we held that “the admissibility of the testimony in question
    -17-
    here presents a close question” and thus denied relief under the plain error standard. 
    Id.
    In United States v. Charley, 
    189 F.3d 1251
     (10th Cir. 1999), we held that a district
    court did not err in admitting a pediatrician’s testimony that sexual abuse would provide a
    unifying diagnosis for several of the victim’s physical and emotional problems, 
    id. at 1263-64
    , but reached the opposite conclusion as to another expert’s statement of “her
    unconditional opinion that each of the girls was in fact sexually abused.” 
    Id. at 1266
    . To
    the extent the opinion was “based on crediting the girls’ account, whether disclosed to her
    or others,” it amounted to an expert “essentially vouching for their truthfulness.” 
    Id. at 1267
    . We ruled that “expert testimony which does nothing but vouch for the credibility
    of another witness encroaches upon the jury’s vital and exclusive function to make
    credibility determinations, and therefore does not ‘assist the trier of fact’ as required by
    Rule 702.” 
    Id.
     However, we determined that the admission of the expert’s opinion was
    harmless “[i]n light of the strength of the properly admitted testimony.” 
    Id. at 1272
    .
    In United States v. Velarde, 
    214 F.3d 1204
     (10th Cir. 2000), the same expert at
    issue in Charley testified that if certain behavioral symptoms were present, she would be
    “comfortable forming a diagnosis of child sexual abuse” and that she “would base that
    diagnosis on the child’s statements about what had happened to them.” 
    214 F.3d at 1209
    .
    We declined to reach the issue of whether this testimony would have been admissible, 
    id. at 1211
    , but noted the expert’s assertion “that she would base her diagnosis of child
    sexual abuse on the child’s statements about what had happened to them, appears to be
    impermissible vouching for [the victim’s] credibility,” 
    id.
     at 1211 n.6 (quotation
    -18-
    omitted). We reached the same conclusion as to a second expert’s testimony that she
    found no evidence suggesting the victim “was subject to either lying or overexaggerated
    fantasizing in her life.” 
    Id. at 1211
    . The Velarde panel rejected a harmless error
    argument and remanded for a new trial. 
    Id. at 1212
    .
    In United States v. Samara, 
    643 F.2d 701
     (10th Cir. 1981), which we cited in
    Toledo, 
    985 F.2d at 1470
    , we rejected a claim that an expert should have been permitted
    to state an opinion that other witnesses were not credible. The government supported its
    charges of income tax violations in that case with testimony from numerous witnesses
    who paid legal fees to the defendant. Samara, 
    643 F.2d at 702
    . The government
    summarized this evidence in an exhibit that showed a substantial amount of unreported
    gross income. 
    Id.
     The defendant sought to challenge that summary with his own expert
    who prepared a competing exhibit “purporting to show that certain items should be
    deleted from the government’s showing of gross receipts.” 
    Id. at 705
    . This expert based
    the deletions on “witness credibility because of felony convictions and lack of
    documentation.” 
    Id.
     We affirmed the district court’s exclusion of the exhibit and related
    testimony “on the ground that credibility was for determination by the jury, not by a
    defense witness” because “[a]n expert may not go so far as to usurp the exclusive
    function of the jury to weigh the evidence and determine credibility.” 
    Id.
     (quotation
    omitted).
    We have repeatedly affirmed the exclusion of expert evidence suggesting that a
    defendant was likely to make a false confession. See United States v. Benally, 541 F.3d
    -19-
    990, 995 (10th Cir. 2008); United States v. Adams, 
    271 F.3d 1236
    , 1246 (10th Cir. 2001).
    And we denied habeas relief, under a de novo standard of review, on a claim that a state
    court erred in excluding expert testimony stating certain child witnesses were not
    credible. Gilson v. Sirmons, 
    520 F.3d 1196
    , 1243 (10th Cir. 2008). Each of these rulings
    was based on the theory that the credibility of another is not an appropriate subject for
    expert opinion testimony.
    It appears our sibling circuits that have considered this issue have uniformly
    agreed. See Engesser v. Dooley, 
    457 F.3d 731
    , 736 (8th Cir. 2006) (“An expert may not
    opine on another witness’s credibility.”); Nimely v. City of New York, 
    414 F.3d 381
    , 398
    (2d Cir. 2005) (“[T]his court, echoed by our sister circuits, has consistently held that
    expert opinions that constitute evaluations of witness credibility, even when such
    evaluations are rooted in scientific or technical expertise, are inadmissible under Rule
    702.”); United States v. Vest, 
    116 F.3d 1179
    , 1185 (7th Cir. 1997) (“Credibility is not a
    proper subject for expert testimony; the jury does not need an expert to tell it whom to
    believe, and the expert’s stamp of approval on a particular witness’ testimony may
    unduly influence the jury.” (quotations omitted)); United States v. Gonzalez-Maldonado,
    
    115 F.3d 9
    , 16 (1st Cir. 1997) (“An expert’s opinion that another witness is lying or
    telling the truth is ordinarily inadmissible pursuant to Rule 702 because the opinion
    exceeds the scope of the expert’s specialized knowledge and therefore merely informs the
    jury that it should reach a particular conclusion.” (quotation omitted)); United States v.
    Beasley, 
    72 F.3d 1518
    , 1528 (11th Cir. 1996) (“Absent unusual circumstances, expert
    -20-
    medical testimony concerning the truthfulness or credibility of a witness is inadmissible
    . . . because it invades the jury’s province to make credibility determinations.”); United
    States v. Rivera, 
    43 F.3d 1291
    , 1295 (9th Cir. 1995) (“[A]n expert witness is not
    permitted to testify specifically to a witness’ credibility or to testify in such a manner as
    to improperly buttress a witness’ credibility.” (quotation and alteration omitted)); United
    States v. Dorsey, 
    45 F.3d 809
    , 815 (4th Cir. 1995) (“[E]xpert testimony can be properly
    excluded if it is introduced merely to cast doubt on the credibility of other eyewitnesses,
    since the evaluation of a witness’s credibility is a determination usually within the jury’s
    exclusive purview.”).
    Although the government acknowledges this court’s jurisprudence on this point, it
    contends that the case at bar is distinguishable in two crucial respects. First, the
    government argues that the foregoing Tenth Circuit cases establish a rule that it is
    impermissible for an expert to vouch for the credibility of a witness, but say nothing
    about an expert’s opinion that a defendant is not credible. We disagree. Our opinion in
    Toledo does not distinguish between pro- and anti-credibility opinions, stating instead
    that “[t]he credibility of witnesses is generally not an appropriate subject for expert
    testimony.” 
    985 F.2d at 1469
    . Moreover, the reasons noted for this rule are that such
    testimony “usurps a critical function of the jury,” “is not helpful to the jury, which can
    make its own determination of credibility,” and “the testimony of impressively qualified
    experts on the credibility of other witnesses is prejudicial and unduly influences the jury.”
    
    Id. at 1469-70
    . These rationales apply with equal force in the context of opinion
    -21-
    testimony that attacks credibility. Further, the government’s theory does not account for
    our holdings affirming the exclusion of expert testimony indicating that other witnesses
    or other out-of-court statements were not credible. See Benally, 541 F.3d at 995; Adams,
    
    271 F.3d at 1246
    ; Samara, 
    643 F.2d at 705
    .3
    The government also seeks to distinguish our circuit precedent as a factual matter,
    arguing that Stanley was not a witness at the trial and therefore the case at bar is not
    analogous to the above-cited cases. It is true that Stanley did not testify; his recorded
    out-of-court statements were played for the jury. But the Federal Rules of Evidence
    provide that statements by in-court witnesses and out-of-court declarants should be
    treated the same with respect to credibility: “the declarant’s credibility may be attacked,
    and then supported, by any evidence that would be admissible for those purposes if the
    declarant had testified as a witness.” Fed. R. Evid. 806. And our court has previously
    applied the rule that expert opinion testimony is not a permissible form of evidence as to
    a declarant’s credibility. In Bledsoe v. Bruce, 
    569 F.3d 1223
     (10th Cir. 2009), we
    considered whether counsel had been ineffective in failing to object to a detective’s
    testimony that he believed an out-of-court statement. 
    Id. at 1237
    . After quoting Toledo,
    and noting that the state court had determined counsel’s failure to object constituted
    deficient performance because “this was objectionable testimony that invaded the
    3
    In each of these cases, we reviewed the district court’s ruling for an abuse of
    discretion. A deferential standard of review is nonetheless sufficient to develop the
    “well-settled law,” Edgar, 
    348 F.3d at 871
     (quotation omitted), required to satisfy the
    second prong of plain-error review.
    -22-
    province of the jury,” we held that the state court’s conclusion that the error did not cause
    prejudice was at least reasonable. Id. at 1237-38 (quotation omitted).
    The government also contends that because Stanley’s recorded statement was
    played for the jury, and because Jones identified specific aspects of Stanley’s statements
    that were indicative of untruthfulness, his testimony assisted the jury rather than usurping
    its role. The government cites United States v. Simpson, 
    7 F.3d 186
     (10th Cir. 1993), in
    which we stated that “[w]hen an expert merely states an opinion on an ultimate issue
    without adequately exploring the criteria upon which the opinion is based, the jury is
    provided with no independent means by which it can reach its own conclusion or give
    proper weight to the expert testimony.” 
    Id. at 188-89
    . We held that the district court did
    not abuse its discretion in excluding testimony from a banking expert that certain
    transactions amounted to “concealment of funds,” 
    id. at 188
    , but noted it was a “close
    question,” 
    id. at 189
    . The government also relies on United States v. Zepeda-Lopez, 
    478 F.3d 1213
     (10th Cir. 2007), in which we permitted an officer’s opinion that the voice on a
    tape belonged to the defendant because the officer had reviewed the tape many times and
    thus the “testimony was helpful to [the jury] in deciding whether Mr. Zepeda-Lopez
    appeared on the portion of the video tape played” at trial. 
    Id. at 1222
    . However, neither
    of these cases involved an expert offering his opinion of the veracity of another’s
    statements, which we have repeatedly held is “not an appropriate subject for expert
    testimony.” Toledo, 
    985 F.2d at 1470
    . It was the subject, rather than the particular
    manner, of Jones expert opinion testimony that was improper.
    -23-
    We must also reject the government’s characterization of Jones’ testimony as
    merely providing to the jury expert tools to permit them to make their own credibility
    determinations. First, much of the testimony asserted fairly obvious and commonly
    understood issues of veracity. For example, Jones identified as indicators of mendacity
    Stanley’s attempt to steer the conversation away from questions he did not want to
    answer and the changing of his story during the course of the interview. The reason the
    issue of credibility “belongs to the jury” is that jurors “are presumed to be fitted for it by
    their natural intelligence and their practical knowledge of men and the ways of men.”
    Aetna Life Ins. Co. v. Ward, 
    140 U.S. 76
    , 88 (1891). A jury does not need an expert to
    inform it that a witness who changes his story may be less than truthful.
    We acknowledge that in some narrowly circumscribed situations, expert testimony
    that touches on the issue of credibility might be properly admitted. For example, in
    United States v. Shay, 
    57 F.3d 126
     (1st Cir. 1995), the court held that a psychiatrist
    should have been permitted to testify that the defendant “suffered from a recognized
    mental disorder known as ‘pseudologia fantastica’” that caused him to tell false and self-
    aggrandizing stories. 
    Id. at 129, 133-34
    . And in United States v. Hall, 
    93 F.3d 1337
     (7th
    Cir. 1996), the court ordered a new trial following the exclusion of expert testimony
    suggesting that false confessions sometimes occur and that the defendant had a
    “personality disorder that makes him susceptible to suggestion.” 
    Id. at 1341
    . The court
    held that “[i]t was precisely because juries are unlikely to know that social scientists and
    psychologists have identified a personality disorder that will cause individuals to make
    -24-
    false confessions that the testimony would have assisted the jury in making its decision.”
    
    Id. at 1345
    . Similarly, our decision in Toledo dealt with psychiatric testimony regarding
    the nature of delusions, 
    985 F.2d at 1469
    , and analogized to child psychiatry issues
    outside the ken of a normal juror, 
    id. at 1470
    .
    As we explained in Adams, however, there is a wide gulf between that type of
    specialized psychiatric knowledge, which one would not expect a jury to possess, and
    testimony that merely asserts an opinion as to the veracity of an “explanation that a jury
    is capable of resolving without expert testimony.” 
    271 F.3d at
    1246 (citing Hall, 
    93 F.3d at
    1341 and Shay, 
    57 F.3d at 129-30
    ). Jones’ testimony plainly fell into the latter
    category, and thus impermissibly “‘encroache[d] upon the jury’s vital and exclusive
    function to make credibility determinations.’” 
    Id.
     (quoting Charley, 
    189 F.3d at 1267
    ).
    After being asked what he drew from the interview “as to [Stanley’s] truthfulness,” Jones
    stated that Stanley provided “partial truths” and continually sought to “avoid the
    question[s]” asked of him. The prosecutor followed with a clearly inappropriate
    question, inquiring, “In reference to the substance of responses that were provided . . .
    how does that factor into your observation of whether he’s being truthful or not?” Jones
    then summarized the story Stanley provided and stated: “It does not connect the dots.
    That does not make sense. That is not something that I viewed as reasonable.”
    This testimony plainly violated Rule 702 and our case law interpreting the rule.
    Even if Agent Jones arguably had “specialized knowledge,” Fed. R. Evid. 702(a), on the
    subject of interrogations, his testimony on Stanley’s credibility fails under Rule 702
    -25-
    because it “encroache[d] upon the jury’s vital and exclusive function to make credibility
    determinations, and therefore [did] not assist the trier of fact.” Charley, 
    189 F.3d at 1267
    (quotation omitted). He simply informed the jury that Stanley’s version of events was
    unworthy of belief based on his opinion of what is generally “reasonable.” We cannot
    excuse his statement as a witness veering into non-responsive commentary; the
    prosecutor specifically asked about the effect the “substance” of Stanley’s responses had
    on Jones’ “observation of whether [Stanley was] being truthful or not.”
    Moreover, this exchange was not an isolated aside. Jones’ testimony as to
    Stanley’s credibility continued with Jones claiming that Stanley’s references to his
    subjective knowledge “is a move that is common among the criminal element to keep law
    enforcement at bay and not be able to determine the actual facts of what happened.” The
    prosecutor mischaracterized Stanley’s statements in the video by asking whether it was “a
    demonstration of consciousness of guilt that an individual will want to die rather than tell
    the truth.”4 On the same topic of Stanley’s claim that he would rather die than be
    imprisoned, Jones stated, “Never in my career have I seen that with an innocent person.”
    He also asserted that “when people start bringing faith into validating of their statements,
    4
    A review of the video of the interview beginning at the 35 minute mark indicates
    that it was Corporal Stout, not Stanley, who referred to preferring death rather than telling
    the truth. Stout’s follow up question to Stanley’s statement that he would “rather die”
    was: “You’d rather die than tell us the truth?” Stanley’s response was, “I am telling the
    truth.” Both the context of the exchange and the actual video of the interview confirm
    that the questioning of Agent Jones at trial was premised on a misrepresentation of
    Stanley’s statements.
    -26-
    that they’re deceptive. Those are deceptive statements.”
    Contrary to the government’s claim, Jones testified flatly and repeatedly that, in
    his expert opinion, Stanley was dishonest during his interview. Because our court has
    clearly held that “credibility [i]s for determination by the jury,” and “[a]n expert may not
    go so far as to usurp the exclusive function of the jury to weigh the evidence and
    determine credibility,” Samara, 
    643 F.2d at 705
     (quotation omitted), the admission of
    this testimony meets the stringent standard of plain error. See United States v. Toro-
    Pelaez, 
    107 F.3d 819
    , 827 (10th Cir. 1997) (holding that “only a particularly egregious
    and obvious and substantial error” qualifies as plain (quotation omitted)).
    B
    This conclusion does not end our inquiry. To obtain relief, Stanley must also
    show “that the plain error affected his substantial rights.” United States v. Trujillo-
    Terrazas, 
    405 F.3d 814
    , 818 (10th Cir. 2005). That is, he must demonstrate “a reasonable
    probability that but for the error claimed, the result of the proceeding would have been
    different.” 
    Id. at 819
     (quotation and alteration omitted). “A reasonable probability is a
    probability sufficient to undermine confidence in the outcome.” United States v. Hasan,
    
    526 F.3d 653
    , 665 (10th Cir. 2008) (quotation omitted). “The reasonable-probability
    standard is not the same as, and should not be confused with, a requirement that a
    defendant prove by a preponderance of the evidence that but for error things would have
    been different.” United States v. Dominguez Benitez, 
    542 U.S. 74
    , 83 n.9 (2009)
    (quotation omitted).
    -27-
    The government’s case against Stanley was far from airtight. Its best evidence
    was Stanley’s presence, shortly after the crime occurred, in his father’s home where he
    claimed to have been sleeping in the living room. Officers discovered money stolen from
    the bank in the home’s kitchen, and clothing and a firearm consistent with those used in
    the robbery in a bedroom.5 In many circumstances, such evidence standing alone might
    qualify as sufficiently probative of guilt to conclude that the defendant’s substantial rights
    were unaffected by the improper admission of opinion testimony. In this case, however,
    the probative value of Stanley’s presence at the East Pine residence is somewhat undercut
    by the fact that Stanley’s two brothers were also present at the home just after the
    robbery. Vernon has already been convicted of charges relating to the robbery, and
    Dejuan was also charged. The government expressly proceeded upon the theory that
    those two brothers were the two individuals who entered the bank and that Stanley was
    the getaway driver. It would have been reasonable for the jury to conclude that Vernon
    and/or Dejuan robbed the bank and that Stanley simply had the bad luck to be present at
    his father’s home when police surrounded it, especially given the absence of evidence
    presented about the identity of the getaway driver. At Stanley’s first trial, the
    government proceeded on the theory that Stanley was present in the bank and the jury
    5
    The firearm discovered at the East Pine residence formed the basis of the felon in
    possession charge. The jury was instructed that mere presence was insufficient to
    establish possession, and that the government was required to prove some connection
    between Stanley and the firearm. Absent any evidence tying Stanley to the gun (other
    than the evidence suggesting he was involved in the bank robbery), we conclude that the
    firearm conviction must stand or fall with the other convictions.
    -28-
    deadlocked on the charges against him.
    The government draws our attention to the cell tower evidence, which was not
    introduced at the first trial. But the phone most closely tied to Stanley by the evidence,
    with a number ending in 1576, did not produce much in the way of incriminating
    information. That phone was registered to Landrum and was listed by both Stanley and
    Dejuan as a contact number. It was located near Arvest Bank the night before the
    robbery, but the tower servicing the bank covered a fairly large area of Tulsa and was
    very near Landrum’s home. On the morning of the robbery, the 1576 phone was located
    first near the East Pine residence, then near Landrum’s home, when it exchanged calls
    with the two other phones discussed at trial. At the time of the robbery, it was turned off.
    The phone registered to Vernon, and the other phone registered to Landrum, with
    a number ending in 3860, told a more compelling story. Those phones contained data
    strongly suggesting that they were used by a bank robber and a nearby getaway driver,
    and this theory was supported by eyewitness testimony. Jones testified that he believed
    Dejuan was carrying the 1576 phone and that Stanley was carrying the 3860 phone, but
    the prosecution did not cite any evidence supporting the assertion that Stanley had the
    3860 phone. Further, the 3860 phone was tracked moving away from the East Pine
    residence at the time Stanley and Vernon were holed up in that house. Jones professed
    his belief that Dejuan obtained that phone from Stanley just after the robbery, but there
    was no evidence advanced to support this supposition. The cell tower testimony was not
    significantly probative of Stanley’s involvement.
    -29-
    The government also points to the recorded phone calls Stanley made from jail. In
    one call, Landrum mentions a discussion in which Stanley was referred to as “the weak
    link of the group.” The government argues that this is evidence of conspiracy, but it may
    simply refer to Stanley’s fortitude in dealing with incarceration. In another call,
    Stanley’s father warns him not to say “nothing to do with what they holding y’all for . . .
    don’t say where you was.” Given that Stanley’s brother was jailed on the same charges,
    this statement is hardly definitive proof of Stanley’s involvement. Stanley’s father may
    simply be telling Stanley to exercise his Fifth Amendment rights in an effort to protect
    his children.
    In addition to the foregoing evidence, the government relies on Stanley’s lack of
    candor when apprehended. He falsely identified himself when arrested. When
    confronted, he became tearful and provided his true identity only after being moved away
    from his brother Vernon. And it is clear that Stanley behaved evasively in his interview
    with Jones. He claimed not to know Landrum’s address. And he stated that he was
    planning to watch his stepsister, but said he did not know the name of the girl’s mother.
    We must view Stanley’s evasiveness during the interview in light of the fact that he knew
    his brother was also seized from the home. The jury may well have concluded that
    Stanley was wary of saying anything if he knew or suspected that Vernon and/or Dejuan
    had committed a crime. Indeed, Stanley’s father later advised him to keep quiet,
    reminding Stanley, “you know we’re family, we stick together.”
    The government also argues Stanley’s claim that he was sleeping through the
    -30-
    relevant events is implausible. It points to an officer’s testimony that the oven drawer in
    which the robbery proceeds were secreted was loaded with pots and pans, and that he
    could hear other officers struggling to open the drawer from the room in which Stanley
    claimed to be sleeping. Of course, an individual stashing robbery proceeds would likely
    attempt to keep quiet, whereas a group of officers executing a search warrant would have
    no need for stealth. And as noted above, a jury’s finding that Stanley attempted to avoid
    disclosing his full knowledge during the interview does not mandate a conclusion that he
    conspired to rob Arvest Bank.6
    In light of this evidence, we conclude that there is a reasonable probability that but
    for Jones’ improper testimony, the result of Stanley’s trial would have been different.
    Jones’ testimony directly contradicted Stanley’s “wrong place, wrong time” theory. He
    informed the jury that despite conducting more than a thousand interviews, he had
    “[n]ever in [his] career” seen “an innocent person” display the behaviors exhibited by
    Stanley. Jones opined: “I also don’t reasonably believe an innocent person would”
    behave as Stanley did. And he described Stanley’s reliance on subjective knowledge and
    appeals to religion as “common among the criminal element” and something done by
    “somebody with guilt.” In closing arguments, the government asked the jury to rely on
    Jones’ expertise: “Agent Jones, in scrutinizing this interview with Stanley Hill, he talks
    6
    We note that the jury was not instructed that it could convict based on a
    conspiracy to act as an accessory after the fact. See United States v. Lang, 
    364 F.3d 1210
    , 1223-24 (10th Cir. 2004) (discussing this theory), vacated on other grounds, 
    543 U.S. 1108
     (2005).
    -31-
    about how in his approach to interviewing and interrogation, sometimes he has to piece
    through, he has to figure out what’s truthful in this interview, what’s he trying to hide.”
    The government urges us to disregard the plainly erroneous admission of Jones’
    testimony because the jurors were instructed that they were “the judges of the facts” and
    “the sole judges of the credibility or ‘believability’ of each witness.” Jurors are presumed
    to follow their instructions. United States v. Almaraz, 
    306 F.3d 1031
    , 1037 (10th Cir.
    2002). But the instructions considered as a whole tend to exacerbate the erroneous
    admission of Jones’ expert opinion. They stated that Jones was permitted to express his
    opinions “regarding investigation of conspiracies and robberies” based on his
    “knowledge, skill, experience, training or education,” and that his testimony concerned
    “scientific, technical, or other specialized knowledge.”
    Although the jury was instructed that it was not required to accept the expert
    opinion testimony offered, the danger of such testimony on the subject of credibility is
    not that the jury will be misled into thinking the issue is outside its bailiwick, but that
    “the testimony of impressively qualified experts on the credibility of other witnesses is
    prejudicial and unduly influences the jury.” Toledo, 
    985 F.2d at 1470
    . During his
    testimony, Jones sought to burnish his credentials. He told the jury that he had attended
    “specialized courses” including “the Reid school, which is a higher-level school of
    interrogation and interviewing.” He stated that the Reid school trained him in “special
    tactics and ways to identify on deception in statements and truths in statements” and that
    the training was “sought-after . . . because of the caliber of that training you do get
    -32-
    towards that endeavor.”
    In light of the complete record before us, we conclude that Stanley has carried his
    burden of showing a reasonable probability that the result of his trial would have been
    different without the impermissible testimony. Although the prosecution presented a
    reasonable theory, the likelihood that the jury was unduly swayed by Jones’ improper
    testimony—and would not have found Stanley guilty beyond a reasonable doubt absent
    that testimony—is high enough to undermine our confidence in the result of the trial. See
    Hasan, 
    526 F.3d at 665
    .
    C
    Having concluded that Stanley has satisfied the first three prongs of plain-error
    review, we must consider whether to exercise our “discretion to correct the error if it
    seriously affects the fairness, integrity, or public reputation of judicial proceedings.”
    United States v. Begaye, 
    635 F.3d 456
    , 470 (10th Cir. 2011) (quotation omitted). “[A]
    plain error affecting substantial rights does not, without more, satisfy the . . . standard, for
    otherwise the discretion afforded by Rule 52(b) would be illusory.” United States v.
    Olano, 
    507 U.S. 725
    , 737 (1993).
    We hold that this case warrants reversal under this standard. As discussed supra,
    Stanley has shown a reasonable probability that but for the erroneously admitted
    testimony, he would not have been convicted. We have stated that the “key concern” in
    “considering whether to correct unobjected-to sentencing errors” at the fourth prong of
    plain error review is if “correct application of the sentencing laws would likely
    -33-
    significantly reduce the length of the sentence.” United States v. Brown, 
    316 F.3d 1151
    ,
    1161 (10th Cir. 2003); see also United States v. Cordery, 
    656 F.3d 1103
    , 1108 (10th Cir.
    2011) (applying this rule to correct a five-month sentencing discrepancy). We have also
    elected to reverse in several plain error cases in which the district court entered an
    impermissibly large restitution order. See United States v. Gordon, 
    480 F.3d 1205
    , 1212
    (10th Cir. 2007) (collecting cases and rejecting government’s argument on prong four).
    In contrast, we have declined to exercise our discretion in favor of defendants sentenced
    under a mandatory sentencing Guidelines regime who sought to challenge their sentences
    following the Supreme Court’s holding that the Guidelines were merely advisory. We
    held that the mere fact of mandatory application during a time when the law required it
    did not offend “core notions of justice.” United States v. Gonzalez-Huerta, 
    403 F.3d 727
    ,
    738-39 (10th Cir. 2005) (en banc).
    The consequences of ignoring the error in this case may be significantly more
    severe than the sentencing and restitution cases cited above; upon retrial without Jones’
    testimony (assuming the other evidence remains roughly similar), there is a reasonable
    probability that Stanley will not be convicted at all. However, the Supreme Court has
    indicated that this factor standing alone does not necessarily compel reversal. See Olano,
    
    507 U.S. at 737
    .
    It is the nature of the error at issue in this case that leads us to conclude that
    ignoring it would offend core notions of justice and seriously affect the fairness, integrity,
    and public reputation of judicial proceedings. A law enforcement agent informed the jury
    -34-
    that he was specially trained in ferreting out lies, and the instructions the jury received
    from the district court essentially confirmed that claim. The agent testified that in his
    expert opinion the defendant’s story was unworthy of belief and that the defendant’s
    evasive behavior demonstrated guilt. Such testimony is intolerable under our system of
    jurisprudence, which has long recognized jurors’ ability and sole responsibility to
    determine credibility. See Aetna Life Ins. Co., 
    140 U.S. at 88
    ; Samara, 
    643 F.2d at 705
    .
    Perhaps even more troublingly, Jones’ lengthy testimony was repeatedly incited by
    improper questions from the prosecutor, who inquired whether Jones believed the
    substance of Stanley’s statements and mischaracterized those statements in asking
    whether they indicated that Stanley was guilty.
    We cast no blame on the district court for the error that occurred in this case and
    warn the defense bar that reversal in the absence of contemporaneous objection is a rare
    exception rather than the rule. We nonetheless must conclude that Stanley has satisfied
    all four prongs of the plain error standard and is entitled to relief.
    III
    For the foregoing reasons, we REVERSE and REMAND with instructions to
    VACATE Stanley Hill’s convictions and sentence.
    -35-