Judge JOSÉ A. CABRANES concurs in part and dissents in part in a separate opinion.
MESKILL, Circuit Judge:
In this appeal we consider whether plaintiff-appellee Marilyn J. Bartlett has a disability within the meaning of the Americans with Disabilities Act of 1990(ADA). Bartlett has been diagnosed with dyslexia, a learning impairment. However, not every impairment constitutes a disability under the ADA.
“Disability” is defined to include “a physical or mental impairment that substantially limits one or more of the major life activities” of an individual. ADA § 3(2)(A), 42 U.S.C. § 12102(2)(A). The district court found that Bartlett was not substantially limited with respect to major life activities such as reading or learning, but that she was substantially limited with respect to the major life activity of working. Therefore the court held that the defendant-appellant New York State Board of Law Examiners (the Board) was required to provide Bartlett with reasonable accommodations on the New York State Bar Examination. See Bartlett v. New York State Bd. of Law Examiners, 970 F.Supp. 1094 (S.D.N.Y.1997) (.Bartlett I); see also Bartlett v. New York State Bd. of Law Examiners, 2 F.Supp.2d 388 (S.D.N.Y.1997) (Bartlett II) (denying motion for post-judgment relief).
The defendants appealed. We affirmed in part, vacated in part and remanded. See Bartlett v. New York State Bd. of Law Examiners, 156 F.3d 321 (2d Cir.1998) (Bartlett III). We agreed that Bartlett was disabled, but on different grounds. In particular, in determining whether Bartlett was substantially limited with respect to reading, we held that the district court should not have taken into account Bartlett’s ability to “self-accommodate,” stating that Bartlett’s “history of self-accommodations, while allowing her to achieve roughly average reading skills (on some measures) when compared to the general population, do not take her outside of the protective provisions of the ADA.” Id. at 329 (internal quotation marks omitted). Because we concluded that Bartlett was substantially limited with respect to reading, we did not consider whether she also was substantially limited with respect to working. Id. We vacated and remanded solely as to the proper measure of compensatory damages. See id. at 331-32. A petition for rehearing, with a suggestion for rehearing en banc, was denied.
The Supreme Court granted certiorari and vacated and remanded in light of Sutton v. United Air Lines, 527 U.S. 471, 119 S.Ct. 2139, 144 L.Ed.2d 450 (1999), Murphy v. United Parcel Serv., 527 U.S. 516, 119 S.Ct. 2133, 144 L.Ed.2d 484 (1999), and Albertson’s, Inc. v. Kirkingburg, 527 U.S. 555, 119 S.Ct. 2162, 144 L.Ed.2d 518 (1999). See New York State Bd. of Law Examiners v. Bartlett, 527 U.S. 1031-32, 119 S.Ct. 2388, 144 L.Ed.2d 790 (1999) (mem.). We now hold that, although the district court properly considered corrective or mitigating measures such as Bartlett’s self-accommodations, see, e. g., Al-bertson’s, 527 U.S. at 565-66, 119 S.Ct. 2162, it nevertheless applied the wrong legal standard when it found that Bartlett was not substantially limited with respect to reading because she has “roughly average reading skills (on some measures) when compared to the general population.” Bartlett I, 970 F.Supp. at 1120. It is not enough that Bartlett has average skills on “some” measures if her skills are below average on other measures to an extent that her ability to read is substantially limited. In fact, the district court found [75]*75that Bartlett reads “slowly, haltingly, and laboriously.” Id. at 1099; see also id. (“She simply does not read in the manner of an average person.”). Therefore, we remand for the district court to determine, in the first instance, whether Bartlett is substantially limited in the major life activity of reading by her slow reading speed, or by any other “conditions, manner, or duration” that limits her reading “in comparison to most people.” See 28 C.F.R. Pt. 35, App. A, § 35.104 (1999).
We also disagree with the district court’s analysis of whether Bartlett was substantially limited with respect to the major life activity of working. The district court held that “[i]f plaintiff’s disability prevents her from competing on a level playing field with other bar examination applicants, then her disability has implicated the major life activity of working.” Bartlett I, 970 F.Supp. at 1121. However, it is not enough for a plaintiff to prove that an impairment “implicates” a major life activity — she is required to prove that the impairment “substantially limits” that activity. In this case, it has not been shown that Bartlett’s inability to practice law results from her reading impairment, rather than from other factors that might prevent her from passing the bar. Therefore, we remand for the district court to determine, if necessary, whether it is Bartlett’s impairment, rather than factors such as her education, experience or innate ability, that “substantially limits” her ability to work.
BACKGROUND
After a 21 day trial, the district court found the following relevant facts. Bartlett has a cognitive disorder that impairs her ability to read. Despite her limitation, she has earned a Ph.D. in Educational Administration from New York University, a law degree from Vermont Law School, and has met all prerequisites to sit for the New York State Bar Examination. The Board is a state entity charged with testing and licensing applicants seeking admission to the New York State Bar.
Since 1991, Bartlett has taken the bar examination five times. On at least three and possibly four separate occasions, she applied as a reading disabled candidate to take the bar examination with accommodations. She requested accommodations for the July 1991, February 1993 and July 1993 examinations. Bartlett did not seek accommodations for the February 1992 bar examination, and the record is unclear as to whether she sought accommodations for the July 1992 exam. With respect to the July 1992 exam, the district court found that “[Bartlett] claims she [applied for accommodations], but the Board has no record of the request.” Bartlett I, 970 F.Supp. at 1102.
Bartlett sought unlimited or extended time to take the test and permission to tape record her essays and to circle her multiple choice answers in the test booklet rather than completing the answer sheet. The Board denied her request each time, contending that her application does not support a diagnosis of a reading disability or dyslexia. In total, Bartlett has taken the examination four times without accommodations and has yet to pass. On July 20, 1993, after the Board denied her most recent application for accommodations, she commenced this action in the district court alleging, among other things, violations of titles II and III of the ADA, 42 U.S.C. § 12101 et seq., and § 504 of the Rehabilitation Act, 29 U.S.C. § 794. In her complaint, she sought, among other things, injunctive relief in the form of reasonable testing accommodations and compensatory and punitive damages.
On July 26, 1993 — a mere two days before the July 1993 bar examination — the parties entered into a stipulation. Under its terms, Bartlett received some, but not all, of the accommodations she sought, including time-and-a-half for the New York portion of the test and the use of an aman[76]*76uensis1 to read the test questions and to record her responses. In addition, the Board allowed Bartlett to mark the answers to the multiple choice portion of the examination in the test booklet rather than on a computerized answer sheet. However, the parties agreed that if Bartlett passed the examination, the results would not be certified unless she prevailed in this lawsuit. Despite the accommodations, Bartlett failed the examination.
Prior to July 1993, the Board denied Bartlett’s requested accommodations because its expert on learning disabilities, Dr. Frank R. Vellutino, did not believe that Bartlett had dyslexia or a reading disability. Dr. Vellutino’s opinion was grounded primarily on Bartlett’s performance on two subtests of the Woodcock Reading Mastery Test — Revised (the Woodcock), a battery of tests commonly employed to assess learning disabilities.
The two subtests at issue are the Woodcock “Word Attack” and ‘Word Identification.” According to Dr. Vellutino, these tests are designed to measure a subject’s “[w]ord [^Identification and phonetic decoding or word analysis skills (ability to ‘sound out’ a word).” Bartlett I, 970 F.Supp. at 1112 (internal quotation marks omitted). Specifically, the “Word Attack” subtest requires the subject to sound out 45 nonsense words of varying complexity. The “Word Identification” subtest, on the other hand, measures a subject’s ability to identify 106 real words in isolation that range from a simple “is” to the more difficult “zymolysis.” Both tests are untimed and the scores do not reflect incorrect tries that precede a correct answer. Because “the incidence of learning disabil-it[ies] in the population is estimated at between 5% and 20%,:” id., Dr. Vellutino concluded that a 30% cutoff is reasonably certain to capture all disabled applicants. Accordingly, he generally recommended against providing accommodations to applicants who performed above the 30th percentile,- although he would occasionally “give applicants the benefit of doubt” if either their Word Attack or Word Identification scores were below 30% or 1 or 2 percentage points above. Id.
When Bartlett applied for accommodations for the July 1991 bar examination, she submitted Woodcock Word Attack and Word Identification scores that were above the 30th percentile, and her request for accommodations was denied. When she applied for accommodations for the July 1993 bar examination, she submitted a new evaluation from a clinical psychologist, Dr. Richard F. Heath, indicating at least one Word Attack score below the 30th percentile. Nevertheless, Dr. Vellutino did not give Bartlett “the benefit of doubt” or recommend accommodations because he considered that score to be “an anomaly.” Id.
At trial, Bartlett challenged Dr. Velluti-no’s opinion. She presented expert testimony and other evidence that her reading disability could not be measured solely by the Woodcock. Bartlett’s lead expert witness, Dr. Rosa A. Hagin, opined that Bartlett was learning disabled, placing “considerable emphasis” on Bartlett’s performance on the Diagnostic Reading Test (DRT), which “demonstrat[ed] plaintiffs slow rate of reading.” Id. at 1109. On the DRT, when compared to college freshmen, Bartlett’s reading rate of 195 words per minute, timed, placed her in the 4th percentile, while her reading rate of 156 words per minute, untimed, placed her below the 1st percentile. Id. at 1110. Dr. Hagin concluded that “plaintiff does not read in the same condition, manner or duration of the average adult reader in that plaintiff does not read with the auto-maticity or speed of an average reader.” Id.
On July 3, 1997, the district court issued its opinion and order. After a thorough [77]*77and painstaking discussion of the evidence, the district court found fatal infirmities in Dr. Vellutino’s reliance on the Woodcock and the Board’s subsequent rejection of Bartlett’s claim of disability. Specifically, the court found (a) the Woodcock could not measure Bartlett’s lack of “automaticity,” i.e., her ability to recognize a printed word and read it accurately and immediately without thinking; (b) the Woodcock was not timed and thus could not measure the slowness of reading — an important characteristic of adult dyslexics like Bartlett, who had demonstrated a reading rate comparable to the bottom fourth percentile of college freshman or worse; (c) the Woodcock was designed principally to assess children and did not have enough items in the difficult range; and (d) Bartlett’s Woodcock results exhibited discrepancies, revealing high reading comprehension scores in comparison to low, but average, Word Attack and Word Identification scores. See id. at 1114. Furthermore, the district court found that Dr. Vellutino’s use of a 30th percentile cutoff was arbitrary and flawed because other studies demonstrated that one third of adults with dyslexia scored above that percentile on similar tests. See id.
In sum, the court agreed with Bartlett’s experts that “a reading disability is not quantifiable merely in test scores.... [Djiagnosing a learning disability requires clinical judgment.” Id. In this regard, the district court found that Bartlett’s low “test scores on the Woodcock, combined with clinical observations of her [slow and halting] manner of reading amply support a conclusion that she has an automaticity and a reading rate problem.” Id.; see also id. at 1107. Moreover, the court agreed with Bartlett’s experts that her “earlier work as a school teacher where phonics were stressed allowed [her] to develop ‘self-accommodations’ that account for her ability to spell better and to perform better on word identity and word attack tests than would be expected of a reading disabled person.” Id. at 1109; see also id. at 1120.
The district court, however, did not find that Bartlett is substantially limited in the major life activities of reading or learning, reasoning that her “history of self-accommodation has allowed her to achieve ... roughly average reading skills (on some measures) when compared to the general population.” Id. Rather, the court, relying on regulations promulgated under Title I of the ADA, held that Bartlett is disabled in her ability to “work” because her reading rate compared unfavorably with “persons of comparable training, skills, and abilities.” Id. at 1Í21 (internal quotation marks omitted). Specifically, the court concluded that Bartlett’s inability to compete on the bar examination constituted a work disability, stating:,
If plaintiffs disability prevents her from competing on a level playing field with other bar examination applicants, then her disability has implicated the major life activity of working because if she is not given a chance to compete fairly on what is essentially an employment test, she is necessarily precluded from potential employment in that field. In this sense, the bar examination clearly implicates the major life activity of work ing.
Id. The court then concluded, inter alia, that Bartlett is disabled within the meaning of the ADA and § 504 of the Rehabilitation Act, id. at 1126, and that the Board’s failure to accommodate her constituted violations of those statutes.
As a remedy for the violations found, the court ordered injunctive relief in the form of reasonable testing accommodations including double time in taking the examination over four days, the use of-a computer, permission to circle multiple choice answers in the examination booklet, and large print on both the New York State and Multistate Bar Exam. Id. at 1153. The court also awarded compensatory damages in the amount of $12,500 for fees paid and the cost of review courses taken [78]*78in connection with the five bar examinations that Bartlett failed. Id. at 1152.
On July 14, 1997, the Board moved for relief from the judgment or, in the alternative, to amend the judgment, pursuant to Fed.R.Civ.P. 59(e) and 60(b). When that motion was denied, see Bartlett II, 2 F.Supp.2d at 396, the Board appealed. We identified the legal issues presented by the appeal to be:
(1) whether the district court erred in refusing to defer to the Board’s determination that Dr. Bartlett is not disabled; (2) whether the district court erred in concluding that Dr. Bartlett is disabled under the ADA and the Rehabilitation Act in her ability to work and thus entitled to accommodations in taking the New York State Bar Examination; (3) whether the district court erred in concluding that the Board is subject to the strictures of the Rehabilitation Act; and (4) whether the district court erred in awarding Dr. Bartlett compensatory damages in the amount of $12,500 from the Board for fees paid in connection with the five bar examinations that she failed.
Bartlett III, 156 F.3d at 324. Because the Supreme Court’s decisions in Sutton, Murphy and Albertson’s have no bearing on issues (1) and (3), we adhere to our earlier resolution of those issues. In other words, for the reasons stated in Bartlett III, we hold that the Board is not entitled to deference on the question of whether Bartlett suffers from a disability under the ADA and the Rehabilitation Act, see id. at 327 (“The Board has no expertise in assessing learning disabilities.”), and we hold that the Board is subject to the strictures of the Rehabilitation Act because it is a recipient of federal funds, see id. at 329-30. We also adhere to our earlier determination that Bartlett is entitled to compensatory damages if her rights under the ADA were violated, see id. at 331 (finding that exclusive reliance on two Woodcock sub-tests constituted “deliberate indifference to a strong likelihood of violating ... federally protected rights”), but we provide additional guidance on the question of the proper measure of compensatory damages. We turn now to whether Bartlett suffers from a statutorily cognizable disability and, if so, the proper measure of compensatory damages.
DISCUSSION
I. Whether Bartlett Has A Disability
Title II, subtitle A of the ADA prohibits discrimination by public entities, such as the Board, on the basis of disability:
Subject to the provisions of this title, no qualified individual with a disability shall, by reason of such disability, be excluded from participation in or be denied the benefits of the services, programs, or activities of a public entity, or be subjected to discrimination by any such entity.
ADA § 202, 42 U.S.C. § 12132. Regulations promulgated by the Attorney General (Department of Justice regulations) pursuant to delegated authority, see ADA § 204, 42 U.S.C. § 12134, specifically provide that “[a] public entity may not administer a licensing or certification program in a manner that subjects qualified individuals with disabilities to discrimination on the basis of disability.” 28 C.F.R. § 35.130(b)(6) (1999).2
[79]*79An individual has a disability under the ADA if, inter alia, that individual has “a physical or mental impairment that substantially limits one or more of the major life activities of such individual.” ADA § 3(2)(A), 42 U.S.C. § 12102(2)(A). Thus, to decide whether an individual has a disability under the ADA, a three-step inquiry is called for: first, we must decide whether that individual suffers from a “physical or mental” impairment; second, we must determine whether the life activity on which the individual relies amounts to a “major” life activity; and third, we must ask whether the specified impairment “substantially limits” that major life activity. See Bragdon v. Abbott, 524 U.S. 624, 631, 118 S.Ct. 2196, 141 L.Ed.2d 540 (1998).
Our analysis is guided by the Department of Justice regulations. Although “no agency has been delegated authority to interpret the term ‘disability,’ ” Sutton, 527 U.S. at 479, 119 S.Ct. 2139, the agency authorized to issue regulations implementing specific provisions of the ADA is entitled to “great deference” on the meaning of “disability” as used in those provisions. See Muller v. Costello, 187 F.3d 298, 312 & n. 5 (2d Cir.1999) (according “great deference” to Equal Employment Opportunity Commission’s interpretation of “disability” in employment discrimination case under title I). As noted previously, the Attorney General has authority to issue regulations implementing title II, subtitle A, under which this case arises, so the Department of Justice regulations interpreting “disability” are entitled to great deference.
A.Physical or Mental Impairment
With respect to the first step of our analysis, the regulations provide that the phrase “physical or mental impairment” means “[a]ny mental or psychological disorder such as mental retardation, organic brain syndrome, emotional or mental illness, and specific learning disabilities.” 28 C.F.R. § 35.104 (1999). The Board contends that Bartlett does not suffer from a specific learning disability but is merely a slow reader with otherwise average reading skills. The district court rejected the Board’s argument:
[A] reading disability is not quantifiable merely in test scores.... By its very nature, diagnosing a learning disability requires clinical judgment. Clinicians need to examine a patient to ensure that low or disparate scores are not the result of low intelligence, or emotional or other social problems.... [A]s much as the Board would like to find an easy test discriminator for a reading disability in its applicants, such a test does not exist.
Bartlett I, 970 F.Supp. at 1114. The district court’s conclusion that Bartlett suffers from a mental impairment, in this case “an automaticity and a reading rate problem,” and not from “low intelligence, or emotional or other social problems,” is not clearly erroneous. See id.; see also Bartlett II, 2 F.Supp.2d at 394 n. 4.
B.Major Life Activities
Next, we must decide whether the life activities allegedly implicated by Bartlett’s impairment are “major” life activities. “The term ‘major life activit[y],’ by its ordinary and natural meaning, directs us to distinguish between life activities of greater and lesser significance.” Reeves v. Johnson Controls World Servs., 140 F.3d 144, 151 (2d Cir.1998) (alteration in original); see Bragdon, 524 U.S. at 638, 118 S.Ct. 2196. The question is whether the life activity is “major” as contemplated by the ADA, not whether the life activity is particularly important to the plaintiff. Reeves, 140 F.3d at 151-52. The regulations list life activities that are “major life activities per se,” id. at 152, including “caring for one’s self, performing manual tasks, walking, seeing, hearing, speaking, breathing, learning, and working,” 28 C.F.R. § 35.104 (1999). “[T]his list ... is [80]*80meant to be illustrative and not exclusive.” Reeves, 140 F.3d at 150.
In this case, Bartlett claimed that her impairment limits her with respect to the major life activities of learning, reading, writing, studying, test-taking and working. See Bartlett I, 970 F.Supp. at 1117. However, the district court, and the parties on appeal, focused primarily on the life activities of reading and working. Therefore, we do not decide whether life activities such as studying and test-taking are “major.” It is sufficient for us to hold that “reading” and “working” are major life activities. See Bartlett III, 156 F.3d at 328 n. 3 (reading); EEOC v. R.J. Gallagher Co., 181 F.3d 645, 654-55 (5th Cir.1999) (working). Indeed, the Board presents no arguments to the contrary.
C. Substantial Limitation
Thus, we reach the central question presented by this appeal: whether Bartlett’s impairment “substantially limits” her with respect to the major life activities of reading or working. Whether an individual is substantially limited with respect to a major life activity is a mixed question of law and fact. See, e. g., Bridges v. City of Bossier, 92 F.3d 329, 333 (5th Cir.1996). We therefore review this aspect of the district court’s judgment de novo. See, e.g., Hirschfeld v. Spanakos, 104 F.3d 16, 19 (2d Cir.1997).
The Department of Justice regulations do not define the phrase “substantially limits,” but the preamble to the regulations provides: “A person is considered an individual with a disability ... when the individual’s important life activities are restricted as to the conditions, manner, or duration under which they can be performed in comparison to most people.” 28 C.F.R. Pt. 35, App. A § 35.104 (1999) (emphasis added). Thus, a person who has lost a leg but who is able to walk or run with the aid of a prosthetic limb is still considered disabled if that person is substantially limited by any “conditions, manner, or duration” that limits his or her ability to walk or run “in comparison to most people'.” See id.
The Supreme Court acknowledged this point in Sutton. In Sutton, the Supreme Court held that “if a person is taking measures to correct for, or mitigate, a physical or mental impairment, the effects of those measures — both positive and negative — must be taken into account when judging whether that person is ‘substantially limited’ in a major life activity.” Sutton, 527 U.S. at 482, 119 S.Ct. 2139; see also Murphy, 527 U.S. at 521, 119 S.Ct. 2133; Albertson’s, 527 U.S. at 565-66, 119 S.Ct. 2162. The Court noted, however, that the use of corrective devices or mitigating measures “does not, by itself, relieve one’s disability.” Sutton, 527 U.S. at 488,119 S.Ct. 2139.
For example, individuals who use prosthetic limbs or wheelchairs may be mobile and capable of functioning in society but still be disabled because of a substantial limitation on their ability to walk or run. The same may be true of individuals who take medicine to lessen the symptoms of an impairment so that they can function but nevertheless remain substantially limited.
Id. Thus, we must account for Bartlett’s self-accommodations in determining whether she is disabled, but the fact that she is able' to self-accommodate does not itself determine whether she is disabled. “[T]hat determination depends on whether the limitations [Bartlett] actually faces are in fact substantially limiting.” Id.
Bartlett is required to show that any limitations “are in fact substantial,” not amounting to only a “mere difference” in “conditions, manner, or duration.” See Albertson’s, 527 U.S. at 564-65, 119 S.Ct. 2162. The burden is not necessarily an onerous one, and in fact, certain limitations will “ordinarily” qualify as disabilities. See id. at 567, 119 S.Ct. 2162 (agreeing that monocular vision will ordinarily be a disability under the ADA); see also Heilweil v. Mount Sinai Hosp., 32 F.3d 718, [81]*81722 (2d Cir.1994) (“Because the [Rehabilitation] Act is a remedial statute, it and the regulations promulgated under it are to be construed broadly.”).
1. Substantial Limitation With Respect to Reading
In Bartlett III, we observed that “Bartlett suffers from a lack of automaticity and a phonological processing defect that significantly restricts her ability to identify timely and decode the written word, that is, to read as compared to the manner and conditions under which the average person in the general population can read or learn.” Bartlett III, 156 F.3d at 329. We concluded that Bartlett was substantially limited with respect to reading, but our conclusion was premised in part on a mistaken belief that Bartlett’s self-accommodations should not have been considered when evaluating her condition. See id. (“ ‘A disability should be assessed without regard to the availability of mitigating measures, such as reasonable accommodations or auxiliary aids.’ ” (quoting H.R.Rep. No. 101-485(11), at 52 (1990), reprinted in 1990 U.S.C.C.A.N. 303, 334)).
Although the district court properly accounted for Bartlett’s self-accommodations, we cannot rely on its conclusion that Bartlett is not substantially limited with respect to reading because the court did not apply the correct legal standard. In particular, the court relied on its finding that Bartlett had achieved “ ‘roughly average reading skills (on some measures) when compared to the general population.’ ” Id. at 326 (emphasis added) (quoting Bartlett I, 970 F.Supp. at 1120). It is not enough that Bartlett has average skills on “some” measures if her skills are below average on other measures to an extent that her ability to read is substantially limited. Compare Gonzales v. National Bd. of Medical Examiners, 225 F.3d 620, 627 (6th Cir.2000) (quoting expert testimony that “in all areas that were assessed, ... [Gonzales] performed in the average to superior range” (emphasis added and internal quotation marks omitted)). Slow reading speed is clearly a condition or manner that can present a substantial limitation — unlike, perhaps, wearing contact lenses. Cf. Sutton, 527 U.S. at 488, 119 S.Ct. 2139 (noting that petitioners did not claim that the use of contact lenses was itself substantially limiting). Alternatively, Bartlett’s slow reading rate may be viewed as a “negative side effect[] suffered by an individual resulting from the use of mitigating measures.” See id. at 484,119 S.Ct. 2139.
In fact, the district court found that Bartlett “reads slowly, haltingly, and laboriously.” Bartlett I, 970 F.Supp. at 1099. “She simply does not read in the manner of an average person.” Id. The court credited the testimony of Dr. Hagin, Bartlett’s lead expert witness, see id. at 1110, 1114, who testified that Bartlett “does not read in the same condition, manner or duration of the average adult reader in that [she] does not read with the automaticity or speed of an average reader,” id. at 1110; see also id. at 1107 (Dr. Heath, another of Bartlett’s experts credited by the district court, “noted in his evaluation that ‘Dr. Bartlett decoded words slowly and without automaticity.’ ”). The district court’s factual finding, that Bartlett lacks automaticity and is a slow reader, is not clearly erroneous. However, it is not dispositive, because the ultimate question is whether Bartlett’s lack of automaticity and slow rate of reading amount to a substantial limitation in comparison to most people or only a “mere difference.” See Albertson’s, 527 H.S. at 565, 119 S.Ct. 2162. The evidence on the point is mixed. One of Bartlett’s own experts, Dr. Massad, had asked her to read a test passage aloud “to get a feel” for her reading rate but “didn’t see anything remarkable to report.” Bartlett I, 970 F.Supp. at 1105-06. Bartlett’s DRT test results, which show a reading rate in the fourth percentile or below as compared to college freshmen, are of limited value, because the proper [82]*82reference group is “most people,” not college freshmen. Therefore, we remand to the district court to determine, in the first instance, whether Bartlett is substantially limited in the major life activity of reading by her slow reading rate, or by any other “conditions, manner, or duration” that limits her reading “in comparison to most people.”
2. Substantial Limitation With Respect to Working
Although the district court held that Bartlett was not substantially limited with respect to reading, it held that Bartlett was disabled because she was substantially limited with respect to working. We cannot affirm the district court’s finding of disability on this alternative ground.
The district court applied the definition of “substantially limits” that was promulgated by the Equal Employment Opportunity Commission (EEOC), holding that “the EEOC’s interpretation of substantial limitation in the context of the major life activity of working is both a part of, and consistent with, the Department of Justice’s regulations.” Bartlett II, 2 F.Supp.2d at 390. The EEOC regulations provide that, with respect to the major life activity of working, the term “substantially limits” means “significantly restricted in the ability to perform either a class of jobs or a broad range of jobs in various classes as compared to the average person having comparable training, skills and abilities.” 29 C.F.R. § 1630.2(j)(3)(i) (1999). Thus, rather than comparing Bartlett to the average person in the general population, the district court compared her to “the average person having comparable training, skills, and abilities” in determining whether Bartlett was substantially limited with respect to working. Bartlett I, 970 F.Supp. at 1120-21. The district court then concluded that Bartlett was significantly restricted in the ability to perform the “class of jobs” involving the practice of law. Id. at 1121.
The Board argues that the district court erred in adopting the EEOC definition of “substantially limits” and that Bartlett should be considered disabled only if she is substantially limited compared to the average person in the general population. The Board also argues that Bartlett is not excluded from a class of jobs, but only from the job of “practicing attorney.” In addition, the National Board of Medical Examiners and the Federation of State Medical Boards of the United States of America, Inc., as amici curiae on the initial appeal in this case, argued that the district court erred in failing to require a causal “nexus” between Bartlett’s impairment and her purported substantial limitation with respect to working. We do not agree with the Board’s arguments, but we find merit in the position of the amici curiae.
The EEOC is one of three agencies authorized to issue regulations to implement the ADA. See Sutton, 527 U.S. at 478, 119 S.Ct. 2139. The EEOC is granted authority to issue regulations implementing the employment provisions of title I, rather than the title II provisions at issue here, see id., so it is not entitled to the same “great deference” that we accord regulations promulgated by the Attorney General. Nevertheless, the EEOC’s views “constitute a body of experience and informed judgment to which courts and litigants may properly resort for guidance.” See Bragdon, 524 U.S. at 642, 118 S.Ct. 2196 (internal quotation marks omitted). See generally Christensen v. Harris County, — U.S. -, 120 S.Ct. 1655, 1662-63, 146 L.Ed.2d 621 (2000) (discussing various degrees of deference to which an agency interpretation may be entitled).
The district court correctly recognized that, if permitted by the relevant regulations, the phrase “substantially limits” should be given a consistent meaning throughout the ADA, especially in light of the “cooperative spirit in which the regulations were promulgated.” Bartlett I, 970 F.Supp. at 390 (citing 1 Henry H. Perritt, [83]*83Jr., Americans With Disabilities Act Handbook § 1.9 (3d ed.1997)); see also Price v. National Bd. of Medical Examiners, 966 F.Supp. 419, 425 & n. 2 (S.D.W.Va.1997) (“Congress clearly intended for the term ‘disability’ (and, therefore, the phrase ‘substantially limits’) to have a uniform meaning throughout the ADA.”). But see Gonzales, 225 F.3d at 601-31.
As noted earlier, the Department of Justice regulations do not define “substantially limits.” The preamble to the regulations provides some guidance and specifies that impairments must be evaluated in comparison to “most people,” but it is ambiguous insofar as it “neglects to explain[] whether ‘most people’ refers to most people in the general population or to most people engaging in that particular life activity.” Bartlett I, 970 F.Supp. at 390 n. 2. We conclude, as the Department of Justice as amicus curiae argues, that the EEOC regulations defining “substantially limits” with respect to working are not inconsistent with the Department of Justice regulations. Cf. Auer v. Robbins, 519 U.S. 452, 461, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997) (holding agency’s interpretation of its own regulation to be “controlling unless plainly erroneous or inconsistent with the regulation” (internal quotation marks omitted)). Therefore the district court did not err in relying on the EEOC regulations.
Next, the Board argues that Bartlett’s impairment excludes her from only “one type of job — practicing attorney.” However, we do not find clear error in the district court’s painstaking analysis of this issue. See Bartlett I, 970 F.Supp. at 1121-26. EEOC regulations provide that an individual is substantially limited with respect to working if she is “significantly restricted in the ability to perform either a class of jobs or a broad range of jobs in various classes.” 29 C.F.R. § 1630.2(j)(3)(i) (1999). The regulations further elaborate on what is meant by a “class of jobs,” id. § 1630.2(j)(3)(ii)(B), or a “broad range of jobs in various classes,” id. § 1630.2(j)(3)(ii)(C), limiting the inquiry to the geographical area “to which the individual has reasonable access,” id. § 1630.2(j)(3)(ii)(A). With respect to “a class of jobs,” a court must consider “[t]he job from which the individual has been disqualified because of an impairment, and the number and types of jobs utilizing similar training, knowledge, skills or abilities.” Id. § 1630.2(j)(3)(ii)(B).
The district court weighed “the number and types of jobs involving the practice of law in New York City alone, much less in the broader geographical market to which plaintiff has reasonable access,” and found that “[a]ll of these countless jobs' and opportunities are foreclosed to plaintiff.” Bartlett I, 970 F.Supp. at 1122; see also id. at 1123-26. Even though, as explained below, Bartlett has not shown that her exclusion from those “countless jobs” was a result of her impairment, we do not doubt that the many and varied jobs utilizing legal training constitute a “class of jobs.” Cf. Department of Labor, Dictionary of Occupational Titles 84-86 (4th ed.1991) (fisting 26 jobs under the heading “Occupations in Law and Jurisprudence” including twelve categories of lawyers).
We begin by identifying Bartlett’s pertinent “training, knowledge, skills or abilities,” and then we compare the jobs utilizing those qualifications with the jobs from which Bartlett is excluded and ask whether exclusion from the latter is a significant restriction with respect to the former. See, e.g., 29 C.F.R. § 1630.2(j)(3). To be “significant,” it is not necessary that Bartlett be excluded from every job for which she is qualified, see, e.g., Fjellestad v. Pizza Hut of America, 188 F.3d 944, 955 (8th Cir.1999) (“[T]he Act does not require a showing that absolutely no employment opportunities exist.”), but it is also not enough for her to be excluded from only “one type of job, a specialized job, or a particular job of choice,” Sutton, 527 U.S. [84]*84at 492, 119 S.Ct. 2139. See generally 29 C.F.R. Pt. 1630, App. § 1630.2(j). In this case, the pertinent qualification is that Bartlett completed law school and earned a law degree, and we ask whether Bartlett’s purported exclusion from the practice of law is a significant restriction relative to the class of jobs utilizing a law degree. Like the district court, we answer yes. Although the Board argues that certain jobs in the legal profession are not foreclosed to Bartlett, e.g., law professor or legal consultant, those jobs, if in fact available to her, are the exception and not the rule.
The Supreme Court’s decision in Sutton is distinguishable. In Sutton, the Supreme Court held that the position of “global airline pilot” was only a single job among “a number of other positions utilizing [flying] skills, such as -regional phot and pilot instructor to name a few.” Sutton, 527 U.S. at 493, 119 S.Ct. 2139. However, the number of lawyers practicing law, relative to the number of people holding a law degree, is surely larger in proportion than the number of global airline pilots relative to the number of people who hold licenses to fly. Similarly, in Murphy, the plaintiff was “generally employable as a mechanic,” but claimed that he was excluded from mechanic jobs that required driving a commercial motor vehicle. See Murphy, 527 U.S. at 524, 119 S.Ct. 2133. The number of mechanic jobs that require driving a commercial motor vehicle is presumably small, relative to the number of mechanic jobs generally, so the Court understandably concluded that the plaintiff “failed to show that he is regarded as unable to perform a class of jobs.” Id. at 525, 119 S.Ct. 2133. Thus, neither Sutton nor Murphy undermine our conclusion: if an impairment bars a person with a law degree from practicing law, then that impairment is a disability under the ADA.3
In the end, however, we disagree with the district court’s conclusion that Bartlett is substantially limited with respect to working. The court reasoned as follows:
If plaintiffs disability prevents her from competing on a level playing field with other bar examination applicants, then her disability has implicated the major life activity of working because if she is not given a chance to compete fairly on what is essentially an employment test, she is necessarily precluded from potential employment in that field. In this sense, the bar examination clearly implicates the major life activity of work ing.
Bartleti I, 970 F.Supp. at 1121. However, it is not enough for a plaintiff to prove that an impairment “implicates” a major life activity — he or she must prove that it is the impairment that “substantially limits” the activity. In other words, the definition of “disability” (“a physical or mental impairment that substantially limits one or more ... major life activities”) encompasses the requirement that it be the impair[85]*85ment, and not some other factor or factors, that causes the substantial limitation; See Hainke v. Gleeson, Sklar, Sawyers & Cumpata LLP, 71 F.Supp.2d 885, 890 (N.D.Ill.1999) (finding no substantial limitation with respect to working absent'“any showing that [plaintiffs] chronic fatigue and sleeping problems were the cause of her tardiness”); see also Hill v. Metropolitan Atlanta Rapid Transit Auth., 77 F.Supp.2d 1291, 1296 (N.D.Ga.1999) (“No medical evidence was presented showing any kind of a link between [plaintiffs] impairments and tardiness.”). Compare Aquinas v. Federal Express Corp., 940 F.Supp. 73, 78 (S.D.N.Y.1996) (finding that poor work attendance was “unrelated to” plaintiffs medical condition) with Dutton v. Johnson County Bd. of County Comm’rs, 859 F.Supp. 498, 506 (D.Kan.1994) (finding that plaintiffs headaches “contributed significantly” to absenteeism).
In most cases, of course, the causation requirement is obviously met. If a person is impaired because he has lost a leg and is substantially limited with respect to the major life activity of walking, there is little room to doubt that the impairment is the cause of the substantial limitation. In this case, however, it has not been shown that Bartlett’s inability to practice law results from her reading impairment, rather than from other factors that might prevent her from passing the bar. Cf. Bartlett I, 970 F.Supp. at 1152 (denying damages for lost wages and benefits because “[a]s defendants correctly point out, ‘plaintiff has failed to prove that with accommodations she would have passed the Bar exam.’” (quoting letter submitted to court)).
However, Bartlett need not prove that she would have passed the bar examination “but for” the denial of accommodations, because even under the best of circumstances a well qualified candidate may not pass on any given sitting. In this case, notwithstanding Bartlett’s failure to pass the July, 1993 bar exam, Bartlett may be able to show that the denial of accommodations was a substantial factor preventing her from passing the exam. The July 1993 examination is an especially poor indicator, given that Bartlett was not granted accommodations until two days before the exam and apparently had no opportunity “to practice with her amanuensis, an accommodation she had never previously used.” See Bartlett I, 970 F.Supp. at 1104. We therefore remand for the district court to determine, if necessary, whether plaintiff has shown that it is her impairment, rather than factors such as her education, experience, or innate ability, that “substantially limits” her ability to work.
II. Compensatory Damages
In Bartlett I the district court awarded $12,500 in compensatory damages, $2,500 for each of the five times that the plaintiff took the bar exam. See id. at 1152. The Board argued that Bartlett was not entitled to compensation for three of the exams, two where Bartlett’s requests for accommodations were submitted late and one where accommodations were granted. The district court rejected the Board’s argument. With respect to the late-filed applications, the court found that the Board “consistently through the years considered untimely applications” and that the denials were clearly decisions on the merits. Bartlett II, 2 F.Supp.2d at 396 n. 7. With respect to the grant of accommodations, the court held that recovery was proper “given that the accommodations granted were neither those that the plaintiff requested nor those to which the. Court has deemed plaintiff was entitled.” Id.
On appeal we vacated the award, explaining that the Board was liable “only for bar examination expenses incurred where the Board denied accommodations because of illegal discrimination.” Bartlett III, 156 F.3d at 332. Because Bartlett did not seek accommodations for the February 1992 exam, we held that “the Board is not liable for damages arising from its failure to accommodate.” Id.
[86]*86We adhere to our earlier reasoning. In addition, we note that the district court should consider, if it finds a disability, whether the Board had enough information to determine that Bartlett was disabled. In other words, the district court may conclude that Bartlett is substantially limited by her slow rate of reading, but it may not be the case that the requests for accommodations and the accompanying reports provided to the Board reflected that limitation. See Bartlett I, 970 F.Supp. at 1137 (“[E]ven plaintiffs own experts in their evaluations did not address or identify plaintiffs reading problem with clarity.”). But see id. at 1107 (Dr. Heath “noted in his evaluation that ‘Dr. Bartlett decoded words slowly and without automaticity.’ ”). If Bartlett’s requests for accommodations did not provide enough information for the Board to determine that she was disabled, the Board would not be hable. See Heilweil, 32 F.3d at 725 (“[A]n employer is only responsible for employment decisions based on information available to it when it decides.”).
CONCLUSION
For the foregoing reasons', we affirm the district court insofar as it (1) declined to defer to the Board’s determination that Bartlett does not suffer from a disability, (2) found that the Board is subject to the strictures of the Rehabilitation Act, and (3) held that Bartlett is entitled to compensatory damages if her rights under the ADA were violated.
We vacate and remand as to (1) whether Bartlett has a disability under the ADA and the Rehabilitation Act, and (2) if so, the proper measure of compensatory damages. We leave it to the district court on remand to decide whether to allow the parties to submit further evidence or whether to resolve these questions on the existing record.