dissenting.
Gonzales testified that he has great difficulty reading highway signs while driving, that in stores he feels compelled to sign credit-card receipts blindly because reading them would take him so long that it would make other customers angry, and that one of his attorneys had to read to him line by line the complaint they prepared on his behalf in the present case so that he could understand it. Assuming that he was telling the truth (and there is no suggestion to the contrary), then I cannot understand why he should not be considered “disabled” within the meaning of the ADA.
I.
A primary basis for the district court’s conclusion that Gonzales does not have a reading disability is that he was able to get reasonably good grades through his first two years of college, and achieve average standardized test scores, without formal disability accommodations. If Gonzales’s claim was that “getting reasonably good grades in school” and “achieving average standardized test scores” were major life activities within the meaning of the ADA, and that those activities were the ones in which he was substantially impaired, the district court might have had a point. But that was not his claim, and, in any event, getting reasonably good grades in school and achieving average standardized test scores are not recognized as “major life activities.” Instead, Gonzales claims that he is disabled in the major life activity of reading, which the NBME does not dispute is a major life activity under the ADA. See, e.g., Bartlett v. New York State Board of Law Examiners, 156 F.3d 321, 324 (2d Cir.1998), vacated on other grounds, 527 U.S. 1031, 119 S.Ct. 2388, 144 L.Ed.2d 790 (1999); Sweet v. Electronic Data Systems, Inc., No. 95 Civ. 3987(MBM), 1996 WL 204471, at *6 (S.D.N.Y. Apr.26, 1996) (“[Rjeading itself is a major life activity independent of the major life activity of seeing.”); but see Hileman v. City of Dallas, 115 F.3d 352, 355 n. 4 (5th Cir.1997) (expressing doubt that reading is a “major life activity” for purposes of the Rehabilitation Act).
*633Gonzales’s reasonably good performance in high school and his first two years of college does not foreclose a finding that he has a reading disability. Cf. Andrew Weis, Jumping to Conclusions in “Jumping the Queue,” 51 Stan. L. Rev. 183, 203 (1998) (book review) (observing that learning-disabled students often are caught in a “Catch-22” situation in that “if we excel in some tasks, then we must not possess a disability, but if we fail in other areas, then we must be just lazy, careless, or inattentive.”). Indeed, Gonzales offered a plausible explanation for how he was able to get good grades without formal accommodations despite having a severe reading disability. He testified that he received very significant informal accommodations in high school (e.g., his teachers permitted him to redo unsatisfactory assignments and turn in untimed extra-credit projects). Gonzales also testified that in college he was able to get fairly good grades without formal accommodations simply by employing strategies that allowed him to get by while doing the bare minimum of reading; for example, by tape-recording lectures and having his friends read their lecture notes to him. If getting reasonably good grades were the bottom line, then I would agree that Gonzales is not disabled. Gonzales, after all, is clearly not substantially worse off than the average person when it comes to getting reasonably good grades.
But a person’s ability to get good grades is not the bottom line. If it were, then a student’s ability to achieve reasonably high marks (by whatever means) without formal accommodations would, as a practical matter, foreclose a finding that he has a reading disability — a conclusion that I believe is incorrect. No one, for example, would argue that a blind student who is able to get good grades in college with the help of friends and an appropriately configured personal computer — but without any formal accommodation from the school— would not be considered “disabled” for the purposes of the ADA. Similarly, if a student with severe reading difficulties can get reasonably high marks in school even though it takes him three times as long as the average person to read the required course materials, it would make little sense to say that he does not have a disability in reading. One might say that he is overcoming his disability as far as getting good grades is concerned, but his method or manner of reading would still be substantially limited as compared to the average person.
I recognize that the Supreme Court in its 1999 trilogy of ADA cases, Murphy v. United Parcel Service, Inc., 527 U.S. 516, 119 S.Ct. 2133, 144 L.Ed.2d 484 (1999), Sutton v. United Air Lines, Inc., 527 U.S. 471, 119 S.Ct. 2139, 144 L.Ed.2d 450 (1999), and Albertson’s, Inc. v. Kirkingburg, 527 U.S. 555, 119 S.Ct. 2162, 144 L.Ed.2d 518 (1999), broadly held that an individual’s ability to self-accommodate or self-correct through medication or treatment must be taken into account in ascertaining whether that person is “disabled” within the meaning of the ADA: The Supreme Court thus held that a hypertensive mechanic whose high blood pressure was controlled by medication was not disabled, see Murphy, 527 U.S. at 518-19, 119 S.Ct. 2133, and that severely myopic prospective airline pilots whose eyesight with glasses was normal were also not disabled, see Sutton, 527 U.S. at 488-89, 119 S.Ct. 2139.
Notably, the Supreme Court also held that a truck-driver with amblyopia, an uncorrectable condition that essentially left him with sight in only one eye, might not be “disabled” for the purposes of the ADA if his brain and body had adapted sufficiently so that his ability to see was not substantially impaired in comparison to the average person. See Kirkingburg, 527 U.S. at 565-66, 119 S.Ct. 2162. The majority’s analysis in the present case relies on Kirkingbwrg, reasoning that Gonzales has learned to self-accommodate in a similar fashion. I respectfully disagree.
In plain although admittedly unscientific terms, Gonzales’s claim is that the part of his brain responsible for decoding written *634language is not wired the same as, and functions substantially worse than, that of the average person, even though in other respects his mental faculties are significantly better than average. If, despite this faulty “wiring,” Gonzales had been able to adapt so that his ability to read was substantially no worse than that of the average person, then Gonzales would not be considered to have a reading disability under Kirkingburg’s rationale. But that is not Gonzales’s claim. Rather, Gonzales has asserted that despite his best efforts to work around his problem, he is still not able to read nearly as well as the average person. I do not believe that working around a reading impairment by pursuing strategies in school that minimize the necessity for reading is the type of self-accommodation the Supreme Court had in mind in Murphy, Sutton, or Kirkingburg.
II.
In addition, the district court’s conclusion that Gonzales had no substantial likelihood of success on the merits is seriously flawed because it is based on the finding that Gonzales does not have a disability within the meaning of the ADA. I do not believe that this finding is supported by the record before us.
Of the experts who testified or submitted reports to the district court, the only two who ever met Gonzales were Drs. Ulrey and Giordani. They both concluded that he had a very significant reading disability. The only expert who opined that Gonzales did not have a disability, or at least had not satisfied her that he had a disability, was Dr. Flanagan. Yet Dr. Flanagan never examined or even met with Gonzales, even though under the guidelines set forth by the American Psychiatric Institute’s Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (4th ed.1994), which the NBME accepts as the appropriate criteria for diagnosing learning disabilities, an interview or clinical examination is essential to making a competent diagnosis of a learning disability. See DSM-IV, p. xxiii (“The specific diagnostic criteria included in DSM-IV are meant to serve as guidelines to be informed by clinical judgment and are not meant to be used in a cookbook fashion”). In other contexts, we accord substantial deference to the medical opinions of treating physicians. See, e.g., Walker v. Secretary of Health & Human Svcs., 980 F.2d 1066, 1070 (6th Cir.1992) (social security disability benefits case). The same should be true for the opinion of an examining psychologist who has diagnosed a disorder that cannot be reliably detected through test scores alone.
Dr. Flanagan conceded that she did not diagnose Gonzales. She explained that she is an academician, not a clinical psychologist competent to diagnose learning disabilities in specific individuals. Instead, her opinion was based solely on second-guessing the conclusions reached by Drs. Ulrey and Giordani, who had examined Gonzales.
In particular, Dr. Flanagan testified that based on her review of the data submitted to the NBME by Drs. Ulrey and Giordani, she could find “absolutely no evidence” of an impairment. She also testified, however, that her methodology for assessing whether persons have learning disabilities was only a “theoretical model that has not been subjected to rigorous empirical analysis for the purposes of diagnosis and treatment.” This leads me to question whether her testimony would satisfy the “gatekeeper” requirements of Kumho Tire Co. v. Carmichael, 526 U.S. 137, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999), and Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 589, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993). Furthermore, to the extent Dr. Flanagan concluded that there was no indication that Gonzales had any learning disability, that conclusion was not shared by Dr. Litchford, the NBME’s other expert witness. Dr. Litchford testified that the results of the tests administered by Dr. Giordani supported a finding of a writing disorder, although Dr. Litchford *635opined that he did not believe that performance on the NBME would be affected by a writing disability.
Dr. Flanagan also never reviewed Gonzales’s second or third applications for testing accommodations while those applications were actually pending before the NBME, even though the subject of the present litigation was Gonzales’s third application for accommodations. Rather, she reviewed them only for the purpose of testifying on the NBME’s behalf in the present case. Gonzales’s attorney, in fact, made a motion in limine to limit Dr. Flanagan’s testimony to the subject of her initial report in which she criticized Dr. Ulrey’s report. The district court denied the motion, stating that “[t]his is merely a hearing, ... not a trial.”
Actually, because the “preliminary” in-junctive relief that the district court denied was the lion’s share of the relief Gonzales was seeking in this action, I believe that the district court should have considered more carefully what evidence would have been considered competent and admissible at trial. See 11A Chaeles Alan Weight, Arthur R. Miller & Mary Kay Kane: Federal Practice and Procedure § 2950, p. 241 (2d Cir. 1995 & Supp. 2000) (noting that when the district court consolidates the motion for preliminary injunc-tive relief with the trial on the merits, “in general the evidentiary rules applicable to trial should govern during a consolidated hearing” because the hearing “really is a trial on the merits”).
Although it appears doubtful that private plaintiffs suing under Title III of the ADA are entitled to a jury trial, see 42 U.S.C. § 12188 (apparently authorizing only injunctive relief and attorney’s fees in Title III suits brought by private plaintiffs); Abbott v. Bragdon, 882 F.Supp. 181, 182 (D.Me.1995) (suggesting that there is no right to a trial by jury in private plaintiffs’ suits under Title III), the district court is still required to rely only on admissible and reliable expert testimony, even while conducting a bench trial. This is true even though district courts conducting bench trials have substantial flexibility in admitting proffered expert testimony at the front end, and then deciding for themselves during the course of trial whether the evidence meets the requirements of Kumho Tire Co. and Daubert and deserves to be credited. See Ekotek Site PRP Committee v. Self, 1 F.Supp.2d 1282, 1296 n. 5 (D.Utah.1998) (concluding that district courts presiding over bench trials can decide questions of admissibility and reliability after the proffered evidence is presented at trial); Bradley v. Brown, 852 F.Supp. 690, 700 (N.D.Ind.) (granting a motion in limine to exclude unreliable expert evidence following the completion of a bench trial), aff'd, 42 F.3d 434 (7th Cir.1994). By failing to conduct a Daubert analysis as to Dr. Flanagan’s testimony (the district court’s eonclusory statement that it found the testimony of Dr. Flanagan “more persuasive” than that of Dr. Giordani does not constitute such an analysis), the district court effectively terminated Gonzales’s case on the merits without employing the evidentiary safeguards that district courts are required to apply when genuine issues of material fact are in dispute — i.e., at a hearing or trial.
In any event, Dr. Flanagan’s exclusive reliance on the test results and reports of Drs. Ulrey and Giordani is troublesome. She reasoned that because Gonzales performed well on tests that measured his cognitive skills, and because individuals’ cognitive processes are “the best predictors of reading achievement,” Gonzales’s reading achievement “ought to be within the average or better range of ability.” I am unable to accept this reasoning.
Correlation is the probability that two factors, such as cognitive ability and reading ability, will accompany each other. A strong positive correlation between cognitive ability and reading ability would mean that, given a large enough sample of people with strong cognitive ability, a predictably large percentage would also have strong reading ability. But Gonzales is a *636person, not a sample. Dr. Flanagan’s failure to examine or even meet with Gonzales leaves her stuck with relying on the argument that Gonzales’s reading ability ought to be good because most people with cognitive ability as strong as Gonzales’s also have good reading ability. This reasoning conflicts with, and leaves unexplained, Gonzales’s testimony (as well as that of Dr. Giordani) to the effect that Gonzales is a member of the minority of individuals with strong cognitive ability but with very weak reading ability.
Furthermore, Dr. Litchford’s testimony did little to support the NBME’s claim that Gonzales does not have a learning disability. Rather, Dr. Litchford testified that the results of the tests administered by Dr. Giordani supported a finding of a writing disorder — although he doubted that having a writing disorder would hinder one’s performance on the Step 1 exam. Dr. Litchford nevertheless forthrightly conceded that neither he nor the NBME had ever conducted any sort of empirical study to determine whether writing disabilities have any effect on how one performs on the NBME’s examinations.
The question of whether a certain form of accommodation (i.e., more testing time) is required or appropriate for a given disability is a question entirely separate from whether the individual has a “disability” in the first place. See Shepler v. Northwest Ohio Developmental Ctr., No. 99-3079, 2000 WL 191496, at *7 (6th Cir. Feb. 9, 2000) (unpublished) (Gilman, J., dissenting) (observing that the question of whether the plaintiff in an ADA case has a disability should be kept analytically separate from other questions, such as whether certain proposed accommodations are required by the ADA); Davidson v. Midelfort Clinic, Ltd., 133 F.3d 499, 508 (7th Cir.1998) (same); Matthew Diller, Judicial Backlash, the ADA, and the Civil Rights Model, 21 Berkeley J. Emp. & Lab. L. 19, 21-29, 49 (2000) (criticizing judicial opinions that deny relief under the ADA after bending over backwards to find plaintiffs not covered by the statute) (“After all, a finding that a person is protected by the ADA only leads to the central question of whether the [defendant] has improperly discriminated against the individual.”). In view of the concession by Dr. Litchford (whose testimony the district court credited) that the results of the tests administered by Dr. Giordani supported a finding of a writing disorder, the district court’s conclusion that Gonzales did not have any disability for the purposes of the ADA appears untenable.
There might well have been other reasons why Gonzales was not likely to succeed on the merits. Indeed, there might have been reasons weighing against granting injunctive relief even if Gonzales had established a likelihood of success on the merits. But the district court’s decision to deny injunctive relief was based on its finding that Gonzales did not have any disability within the meaning of the ADA (and thus had no chance of success on the merits), and that finding, on the basis of the record before us, appears to me to have been clearly erroneous.
III.
I also disagree with the majority’s statement that Gonzales does not “fit within Congress’s vision of the disabled population.” Maj. Op. at 629. Although it is indisputable that “[pjersons with minor trivial impairments, such as a simple infected finger are not impaired in a major life activity,” Maj. Op. at 632, I do not understand how the reading impairment described by Gonzales and Dr. Giordani can reasonably be compared to an infected finger.
Finally, I disagree with the majority’s assertion that recognizing serious reading impairments of the sort described by Gonzales as “disabilities” would stretch the ADA well beyond its intended purpose. Congress anticipated that its definition of “an individual with a disability” would apply to “some 43,000,000 Americans.” 42 U.S.C. § 12101(a)(1). Forty-three million *637is too small a number to include all of the people in the United States who require eyeglasses or contact lenses to see normally, see Sutton, 527 U.S. at 487, 119 S.Ct. 2139 (observing that counting persons who can see normally with corrective lenses as “disabled” would drive the total of “disabled” persons well over one hundred million, a far greater number than Congress intended be covered by the ADA), but forty-three million is still, a very large number — larger, I think, than the majority is willing to recognize.
IV.
I must confess that before delving into this case, I would have been skeptical that a person with a reading disability as serious as the one Gonzales insists he has would be able to survive the rigors of medical school and be capable of practicing medicine successfully afterward. Nevertheless, Gonzales has apparently satisfied the faculty of one of this country’s preeminent medical schools that he would make a very good physician, and I have little doubt that the faculty of the University of Michigan Medical School are better judges than I of who ought to be allowed to practice medicine.
For the reasons set forth above, I believe that the district court’s finding that Gonzales does not have a disability within the meaning of the ADA is not supported by the record as it presently exists. The only finding consistent with Gonzales’s testimony and the conclusions of the two experts who have met with and examined him is that he has a rather severe reading disability. Although the district court was obviously not required to credit Gonzales’s testimony or that of Dr. Giordani, the district court’s decision to credit the testimony of Drs. Flanagan and Litchford (particularly Dr. Flanagan) seems to me irreconcilable with the district court’s recognition of Dr. Giordani as a “competent and accomplished psychologist.” Gonzale[z] v. Nat’l Bd. of Medical Examiners, 60 F.Supp.2d 703, 708 (E.D.Mich.1999).
I would therefore vacate the order of the district court and remand this case for further proceedings. The district court could allow supplemental briefing or, in the exercise of its discretion, the taking of additional evidence on the subject of whether Gonzales is likely to succeed on the merits, and then, if necessary, address the other factors that are relevant in determining whether injunctive relief should be granted.