This drug conspiracy case presents two principal issues. The first is whether the district court erred in ordering defense counsel not to talk to his client during an overnight recess. The second is whether the district court abused its discretion in excluding expert testimony about the defendant’s subnormal intelligence. We reverse.
FACTS
Twin brothers, Eduardo and Ricardo Sandoval-Mendoza, were convicted of conspiring to sell methamphetamine. Eduardo Sandoval-Mendoza sold the drugs. He argues that the government entrapped him as a matter of law. He also argues that the district court erroneously excluded medical evidence of an enormous brain tumor that made him especially vulnerable to entrapment. Ricardo Sandoval-Mendoza argues that the government presented insufficient evidence to convict him. We treat Ricardo Sandoval-Mendoza’s appeal separately in an unpublished disposition.
A family friend named “Marcos” introduced Eduardo Sandoval-Mendoza to “Tony” in February of 2000. Marcos and Tony were government informants. Sandoval-Mendoza sold them about 12 pounds of methamphetamine in three separate deals.
Sandoval-Mendoza testified. He admitted selling the drugs, but claimed the government entrapped him. He claimed the government informants knew a large brain tumor rendered him especially susceptible to suggestion and preyed upon his weakness. The tumor was diagnosed in 1992, eight years before the methamphetamine sales. At first, Sandoval-Mendoza took medication to shrink the tumor, but quit because of the side effects. He resumed only when his doctor told him the tumor would kill him without the medicine.
Sandoval-Mendoza was depressed. He was worried about dying and about providing for his wife and five children. And he was worried about the impotence his tumor caused. He talked about his problems with Marcos, his sister’s boyfriend. Sandoval-Mendoza testified that Marcos told him he made $5,000 to $10,000 a week selling drugs. Marcos and Tony suggested that Sandoval-Mendoza sell drugs to make some money to support his family after he died.
Sandoval-Mendoza testified that he refused to sell drugs for several months, lacking both the experience and the inclination. But eventually he caved in, making three sales to Marcos and Tony. He testified that he sold them drugs only because they used his depression and fear to persuade him, and that he never sold drugs to anyone else.
Sandoval-Mendoza’s account is not entirely credible. On wiretap recordings he sounds suspiciously like an experienced drug dealer, not a neophyte. He testified that a relative, a fugitive drug dealer in Mexico, told him what to say and how to portray himself. This relative also connected him with suppliers in Los Angeles.
Neither of the government informants took the stand to contradict Sandoval-Mendoza’s entrapment defense or offer an alternative explanation. Tapes of their conversations with Sandoval-Mendoza came into evidence, but the government did not put them on the stand. And the informants had an incentive to entrap Sandoval-Mendoza. The government paid them money for their assistance as well as *648offering benefits in their own criminal cases.
To bolster his entrapment defense, Sandoval-Mendoza sought to introduce expert testimony explaining that his large brain tumor damaged his intelligence, memory, and judgment, making him especially susceptible to suggestion. Sandoval-Mendoza’s lawyer’s theme for the jury was “thou shalt not put a stumbling block before the blind.”1 His defense theory was that the government improperly induced a sick and suggestible man to sell drugs.
The district court admitted some evidence in support of Sandoval-Mendoza’s defense, permitting his sister and ex-wife to testify that his tumor made him forgetful. But it excluded all the defense expert testimony. The defense had a neuropsy-chologist and a neurologist ready to testify that the brain tumor did indeed impair Eduardo’s intellect and judgment. After an in camera Dauberb2 hearing, the district court excluded the expert testimony, partly because it did not demonstrate the tumor caused suggestibility and partly because it would be long and confusing. Sandoval-Mendoza argues that this error requires reversal.
The government cross examined Sandoval-Mendoza over the course of two days, spanning a morning recess, a lunch recess, an overnight recess, and another recess on the second day. The district court ordered Sandoval-Mendoza and his lawyer not to communicate with each other during the recesses regarding Sandoval-Mendoza’s testimony, including the overnight recess. The district court allowed communication on other matters, “just not concerning his testimony.” Once cross examination was over, the prohibition was lifted.
Sandoval-Mendoza was convicted and sentenced to 235 months. He appeals, claiming entrapment as a matter of law, error in excluding his expert witnesses’ testimony, error in limiting his consultation with counsel, and error on other grounds.
ANALYSIS
I. Entrapment.
The jury instruction required the government to prove beyond a reasonable doubt that Sandoval-Mendoza was not entrapped. The jury decided that Sandoval-Mendoza was not entrapped. But Sandoval-Mendoza argues that he was entrapped as a matter of law. We review de novo.3 We “will not disturb the jury’s finding unless, viewing the evidence in the government’s favor, no reasonable jury could have concluded that the government disproved the elements of the entrapment defense.” 4
Entrapment has two elements: “government inducement of the crime and the absence of predisposition on the part of the defendant.”5 Inducement is “any government conduct creating a substantial risk that an otherwise law-abiding citizen would commit an offense.”6 We assume Sandoval-Mendoza proved inducement because the government did not dispute that its informants proposed the drug sales.
*649Even so, Sandoval-Mendoza does not establish absence of predisposition as a matter of law. “Where the Government has induced an individual to break the law and the defense of entrapment is at issue, as it was in this case, the prosecution must prove beyond reasonable doubt that the defendant was disposed to commit the criminal act prior to first being approached by Government agents.”7 The government presented evidence Sandoval-Mendoza was predisposed to sell drugs, including wiretap recordings of him talking as though he were an experienced drug dealer. Offering to buy drugs from a drug dealer is not entrapment, even if the government “sets the dealer up” by providing an informant pretending to be a customer, because the dealer is already predisposed to sell.8
In order to prove he was entrapped as a matter of law, Sandoval-Mendoza must “point to undisputed evidence making it patently clear that an otherwise innocent person was induced to commit the illegal act by trickery, persuasion, or fraud of a government agent,”9 He argues that his testimony that he resisted the pressure to sell drugs to the government informants and never sold drugs to anyone else was undisputed because he testified and the informants did not. But the jury did not have to believe Sandoval-Mendoza. Un-contradicted testimony is not necessarily undisputed evidence.10 Jurors may reject uncontradicted testimony when cross examination, other evidence, or their own common sense and ordinary experience convince them the testimony is probably false. “Even perfectly plausible allegations can be disbelieved if they occur during the course of a generally implausible account.”11 The wiretap recordings in which Sandoval-Mendoza pretended to be or really was an experienced drug dealer belie his testimony. The jury could have believed Sandoval-Mendoza’s drug dealer relative coached him and he was just pretending. But it didn’t have to.
The jury could have found that Sandoval-Mendoza was entrapped. But its conclusion to the contrary was supported by enough evidence to meet the Jackson12 standard. This is not a case like Jacobson, 13 where the government failed to present any evidence of predisposition at *650all. Entrapment was properly left to the jury.
II. The order limiting attorney-client discussion.
Sandoval-Mendoza testified over the course of three days. The government’s cross examination spanned a morning recess, a lunch recess, an overnight recess, and another morning recess the following day. The district court instructed Sandoval-Mendoza and his lawyer not to discuss his testimony during any of the recesses, but permitted them to discuss anything else. Sandoval-Mendoza’s lawyer objected and was overruled. When the cross examination ended, the district court permitted Sandoval-Mendoza and his lawyer to discuss his testimony before redirect.
Sandoval-Mendoza argues that the district court’s order prohibiting him from discussing his testimony with his lawyer during the recesses amounted to a structural error under Geders v. United States14 and Perry v. Leeke. 15 Perry and Geders reach opposite conclusions based on different facts. In Geders, the trial court prohibited all communication between the defendant and his lawyer during an overnight recess between direct and cross examination. The Supreme Court held that this prohibition required reversal because it deprived the defendant of his Sixth Amendment right to counsel.16 In Perry, the trial court prohibited all communication between the defendant and his lawyer during a fifteen minute recess between direct and cross examination. The Supreme Court held that this prohibition did not violate the Sixth Amendment.17 Perry distinguished Geders, on the ground that “the normal consultation between attorney and client that occurs during an overnight recess would encompass matters that go beyond the content of the defendant’s own testimony — matters that the defendant does have a constitutional right to discuss with his lawyer, such as the availability of other witnesses, trial tactics, or even the possibility of negotiating a plea bargain.”18
The facts of this case fall in the middle. The district court instructed Sandoval-Mendoza’s lawyer, “You can communicate. Just not concerning cross, his testimony, now that he’s on cross-examination, unless that’s concluded. That doesn’t mean you can’t talk with your client at all, just not concerning his testimony.” On the one hand, the district court prohibited communication between Sandoval-Mendoza and his lawyer during an overnight recess, suggesting Geders controls. On the other hand, the district court only prohibited Sandoval-Mendoza and his lawyer from discussing his testimony, rather than anything at all, suggesting Perry controls.
The core issue is whether prohibiting a defendant and his lawyer from discussing his testimony during an overnight recess violates the Sixth Amendment. Two other circuits have addressed this question. In Mudd v. United States, the District of Columbia Circuit held a similar prohibition *651unconstitutional under Geders. 19 And in United States v. Santos, the Seventh Circuit held a similar prohibition unconstitutional under Perry. 20
This is a difficult question. Cross examination best exposes the truth when a witness must answer questions unaided. Coaching may vitiate its value. But it is hard to see how a defendant and his lawyer can communicate without implicit coaching. The Seventh Circuit suggests that “the judge may instruct the lawyer not to coach his client” but may not prohibit discussion of the client’s testimony.21 But that is not a workable rule, because coaching is implicit in any discussion of a defendant’s testimony, even if the defendant’s lawyer tries his best to avoid coaching.22
We conclude that any overnight ban on communication falls on the Geders side of the line and violates the Sixth Amendment. That seems the fairer reading of Perry, which only permitted prohibitions on communication between a defendant and his lawyer during a “brief recess.”23 Perry recognized a defendant has a “constitutional right” to discuss matters other than his own testimony with his lawyer, “such as the availability of other witnesses, trial tactics, or even the possibility of negotiating a plea bargain,” during an overnight recess.24 And it conceded that “such discussions will inevitably include some consideration of defendant’s ongoing testimony.”25 Indeed, it is hard to see how a defendant’s lawyer could ask him for the name of a witness who could corroborate his testimony or advise him to change his plea after disastrous testimony, subjects Perry expressly says a defendant has a right to discuss with his lawyer during an overnight recess, without discussing the testimony itself.
Thus, we conclude that trial courts may prohibit all communication between a defendant and his lawyer during a brief recess before or during cross-examination, but may not restrict communications during an overnight recess.26 This simple rule is consistent with the reasoning of Geders and Perry. And it has several other advantages. First, it is easy to understand and apply. Second, it dispenses with pretense. Jurors can believe that a defendant did not communicate with his lawyer during a bathroom break. But only a lawyer more wedded to words than common sense can believe that a defendant communicated with his lawyer during an overnight recess *652without at least implicitly discussing his testimony.
Third, as Geders explains, prosecutors and judges can address the coaching problem without prohibiting communication during an overnight recess. The trial court may “exercise reasonable control” over the order and timing of direct and cross examination in order to “make the interrogation and presentation effective for the ascertainment of the truth.”27 For instance, it “may direct that the examination of the witness continue without interruption until completed” or “arrange the sequence of testimony so that direct— and cross-examination of a witness will be completed without interruption.”28 If a defense lawyer strings out direct examination until the usual time for the evening recess, the court can delay the recess and give the prosecutor a few minutes to get in some cross examination.
Thus, we conclude that the district court erred in prohibiting Sandoval-Mendoza and his lawyer from discussing his testimony during an overnight recess. But does the error require reversal? While Geders implies it does,29 Geders, a 1976 decision, preceded many recent Supreme Court decisions requiring prejudice as well as constitutional error for reversal.30 Still, under the recent Supreme Court decision in United States v. Gonzales-Lopez, 31 if an error is structural, prejudice is irrelevant. We need not decide whether or not an overnight prohibition of communications regarding the defendant’s testimony is structural error, because another error, described below, independently requires reversal.
III. The Excluded Expert Witness Testimony.
‘We review the district court’s decision to exclude expert witness testimony for abuse of discretion.”32
Sandoval-Mendoza wanted to present expert testimony concerning his mental condition and susceptibility to suggestion. He offered two expert witnesses: Dr. Michael Shore, a neuropsyehologist, and Dr. J. Richard Mendius, a neurologist. The prosecutors also proposed expert witnesses: Dr. Ronald H. Roberts, a neuropsychologist, and Dr. Richard Cuneo, a neurologist. The district court held a Daubert 33 hearing on whether any of these expert witnesses would be permitted to testify.
Defense witness Michael Shore, Ph.D., is a psychologist with extensive clinical and teaching experience in neuropsycholo-gy, focusing on the rehabilitation of patients suffering from brain damage caused by strokes, tumors, and other causes. He testified that Sandoval-Mendoza suffered from an unusually large pituitary tumor measuring 2x2x3 centimeters, about the size of an apricot, when diagnosed. The tumor compressed Sandoval-Mendoza’s frontal lobe, temporal lobe, and thalamus, probably causing damage. Medication eventually shrank *653Sandoval-Mendoza’s tumor to some degree, but could not reverse any brain damage.
The relationship between brain damage and cognitive impairment is well-documented. Tumors like Sandoval-Mendoza’s may affect mental condition in two ways. First, damage to the pituitary gland may affect thyroid production, causing mood disorders, including depression. Second, direct damage to the frontal lobe, temporal lobe, and thalamus may affect memory, decision-making, judgment, mental flexibility, and overall intellectual capacity. In particular, damage to the frontal lobe often affects concentration, focus, learning, memory, decision-making, reasoning, judgment, and problem-solving, according to Dr. Shore.
A battery of routine and widely-accepted tests showed Eduardo suffered from brain damage affecting his cognitive condition. Although an performance IQ test showed borderline mental retardation and a mental age of eleven, Dr. Shore concluded Sandoval-Mendoza’s school and work history were not consistent with retardation, and attributed the test results to his tumor. A classic nineteenth-century study showed frontal lobe damage causes a person to have “the passions of a man but the mind of a child,” increasing suggestibility. In Sandoval-Mendoza’s case, the brain damage apparently affected the “passions of a man” as well, because the tumor made him impotent.
Defense witness Dr. J. Richard Mendi-us, M.D., is a board-certified neurologist with additional expertise in clinical neurophysiology. He testified that a magnetic resonance imaging test showed Sandoval-Mendoza suffers from an unusually large pituitary tumor. When the tumor shrank after treatment, the frontal lobe herniated into the empty space. The tumor also caused atrophy of the inside of the left temporal lobe and penetrated a bone separating the pituitary gland from the brain stem. Brain damage of this kind tends to affect judgment, memory, and emotions connected to memory. A performance IQ test suggested a very low level of intellectual function. Both Shore and Mendius testified that they knew of no studies specifically linking brain damage of this kind with susceptibility to inducement to commit crimes. But they noted that it commonly causes disinhibition.
Prosecution witness Ronald H. Roberts, Ph.D., is a neuropsychologist who mainly testifies as an expert witness. Roberts agreed Sandoval-Mendoza suffered from an unusually large tumor. However, he believed Sandoval-Mendoza was deliberately under-performing on memory texts. He also believed the tumor did not significantly affect Sandoval-Mendoza’s performance IQ test result. Though he did not contradict the determinations from the MRI films that Sandoval-Mendoza had an extremely large tumor, he gave the opinion that on the memory tests, Sandoval-Mendoza was faking a worse memory than he really had.
Prosecution witness Dr. Richard Cuneo, M.D., is a neurologist. Cuneo agreed Sandoval-Mendoza suffered from an unusually large pituitary tumor near areas of the brain controlling behavior and cognition. But medical understanding of behavior and cognition is preliminary and inconclusive. Dr. Cuneo thought the studies presented were inadequate to show that Sandoval-Mendoza’s tumor and brain damage affected his behavior and cognition because the studies were retrospective and involve small samples. A patient of Dr. Cuneo’s had a similar tumor that neither caused brain damage nor affected behavior and cognition. Sandoval-Mendoza’s magnetic resonance imaging test did not show any brain damage, in Dr. Cuneo’s opinion. *654While some brain tumors may cause disin-hibition or greater susceptibility to influence, pituitary tumors do not, unless they are even larger than Sandoval-Mendoza’s. Dr. Cuneo conceded that Sandoval-Mendoza’s performance IQ test was borderline “retarded” and mentally retarded people are “known to be susceptible to the influence of others.”
After the Daubert hearing, the district court excluded the expert testimony as “not relevant to the entrapment defense” because it “does not tend to show either inducement or a lack of predisposition attributable to the tumor.” The court based its ruling on the expert testimony’s “lack of scientific validity” and “absence of ability to make a causal connection” between the tumor and inducement or predisposition. Alternatively, it concluded that the “probative value” of the expert testimony was “outweighed by the dangers of confusing the issues, misleading the jury, and creating undue delay,” and “would be extremely confusing to both the court and the jury” especially “given the fact that the defense witnesses will then be rebutted by government witnesses.”
As a consequence, only Sandoval-Mendoza himself, his ex-wife, and his sister could testify that his brain tumor made him forgetful and suggestible. The defense had another witness prepared to testify that Sandoval-Mendoza once drank his own urine sample, having forgotten what it was, but the witness disappeared after the prosecutor advised her that as an illegal alien she could be putting herself at risk of deportation. But jurors might well disregard the lay evidence that came in as biased and lacking scientific foundation, since they were deprived of medical evidence.
Although the abuse of discretion standard of review is liberal, the district court’s decision to exclude the expert testimony creates a “definite and firm conviction that the district court committed a clear error of judgment.”34 Daubert makes the district court a gatekeeper, not a fact finder. When credible, qualified experts disagree, a criminal defendant is entitled to have the jury, not the judge, decide whether the government has proved its case.35
Federal Rule of Evidence 702 governs the admission of expert opinion testimony.36 Under Daubert37 and Kumho Tire, 38 only relevant and reliable expert opinion testimony is admissible. Expert opinion testimony is relevant if the knowledge underlying it has a “valid ... connection to the pertinent inquiry.”39 And it is reliable if the knowledge underlying it “has a reliable basis in the knowledge and experience of [the relevant] discipline.”40
*655Trial courts must exercise reasonable discretion in evaluating and in determining how to evaluate the relevance and reliability of expert opinion testimony.41 Daubert and Kumho Tire suggest factors trial courts may consider when evaluating the relevance and reliability of expert opinion testimony. For example, in evaluating the reliability of scientific expert opinion testimony, trial courts may consider: “(1) whether the scientific theory or technique can be (and has been) tested, (2) whether the theory or technique has been subjected to peer review and publication, (3) whether there is a known or potential error rate, and (4) whether the theory or technique is generally accepted in the relevant scientific community.”42 Of course, “there are many different kinds of experts, and many different kinds of expertise,” so these factors “may or may not be pertinent in assessing reliability, depending on the nature of the issue, the expert’s particular expertise, and the subject of his testimony.” 43
When evaluating specialized or technical expert opinion testimony, “the relevant reliability concerns may focus upon personal knowledge or experience.”44 Because medical expert opinion testimony “is based on specialized as distinguished from scientific knowledge, the Daubert factors are not intended to be exhaustive or unduly restrictive.”45 Under our decision in Sullivan v. United States Dep’t of the Navy, the district court “applied an inappropriately rigid Daubert standard to medical expert testimony” by not accepting what “a good [physician] would in determining what is reliable knowledge in the [medical] profession.”46 A trial court should admit medical expert testimony if physicians would accept it as useful and reliable. Utility to the jury of medical expert testimony should be determined by what physicians would accept as useful.
The district court concluded that the proposed medical expert opinion testimony was unreliable because it did not conclusively prove Sandoval-Mendoza’s brain tumor caused susceptibility to inducement or a lack of predisposition. But medical knowledge is often uncertain. The human body is complex, etiology is often uncertain, and ethical concerns often prevent double-blind studies calculated to establish statistical proof. This does not preclude the introduction of medical expert opinion testimony when medical knowledge “permits the assertion of a reasonable opinion.” 47
Predisposition or its absence is the focus of an entrapment defense.48 Therefore, *656medical expert opinion testimony showing that a medical condition renders a person unusually vulnerable to inducement is highly relevant to an entrapment defense. If it is adequately supported by medical expert opinion, it is admissible. Sandoval-Mendoza’s experts were well qualified and had sufficient expertise in the neurology of brain tumors and his particular case to be useful to the jury. The district court’s exclusion of medical expert opinion testimony prevented Sandoval-Mendoza from showing lack of predisposition, “and thereby deprived him of a fair opportunity to defend himself.”49 In this case, the foundation was sufficient. After hearing Drs. Mendius and Shore, the jury could have decided to disbelieve them. But Sandoval-Mendoza was entitled to have the jury decide upon their credibility, rather than the judge. As it was, the jury was left with nothing but unpersuasive lay evidence on a medical matter beyond what laymen could usefully testify about.
The district court excluded the medical expert opinion testimony alternatively in order to avoid “unfair prejudice, confusion of the issues, or misleading the jury, or by considerations of undue delay, waste of time, or needless presentation of cumulative evidence.”50 But there could be no “confusion of the issues” because predisposition was the issue. Sandoval-Mendoza’s only defense was entrapment, and entrapment came down to predisposition. And there was no risk of “misleading the jury.” The experts agreed Sandoval-Mendoza has an unusually large brain tumor. Their only disagreement was whether it caused susceptibility to inducement. The jury was capable of weighing the conflicting medical expert opinion testimony against the rest of the evidence presented and determining whether or not predisposition existed. As for “undue delay,” testimony would likely consume no more time than the Daubert hearing, and probably much less.
Without the medical expert opinion testimony, the real issue in dispute was hidden from the jury. It could not determine whether the government’s informants induced a vulnerable and suggestible man to break the law. The informants did not testify, so the jury could not evaluate the pressure they put on Sandoval-Mendoza. It could not evaluate the merits of Sandoval-Mendoza’s suggestibility, because the medical expert opinion testimony concerning the possibility his tumor or limited mental capacity made him susceptible to inducement was excluded. All the jury had was proof that Sandoval-Mendoza sold drugs, wiretap recordings in which he sounded like an experienced drug dealer, and a couple of lay witnesses testifying that he was addled by a brain tumor. Sandoval-Mendoza is entitled to present his case to the jury. For that, he deserves a new trial.
Because this error requires reversal, we need not reach Sandoval-Mendoza’s other claims.
REVERSED.