with whom Justice Brennan joins, dissenting.
This case concerns the evidentiary threshold that a claimant of black lung benefits must meet to invoke the interim presumption of total disability due to pneumoconiosis under 20 CFR § 727.203(a) (1987). The Director of the Office of Workers’ Compensation (Director) interprets the regulation to require that a claimant prove by a preponderance of the evidence one of the four medical requirements listed in § 727.203(a) to trigger the presumption. The Court of Appeals, en banc, rejected the Director’s proffered interpretation and held that the presumption is invoked once the claimant has presented a single item of evidence meeting one of the medical requirements — that is, “one positive x-ray, one qualifying set of ventilatory or blood gas studies, or one physician’s opinion.” Stapleton v. Westmoreland Coal Co., 785 F. 2d 424, 436 (CA4 1986). The Court chooses to embrace the Director’s view. Because I believe that the Director’s interpretation of the regulation contravenes its plain language and creates a regulatory scheme that is unnecessarily complex and internally inconsistent, I dissent.
I
The language and structure of the regulation provide the most compelling evidence for rejecting the Director’s interpretation. The regulation sets up two evidentiary stages: *162part (a), the presumption-invocation stage, and part (b), the rebuttal stage. Part (a) provides that a “miner who engaged in coal mine employment for at least 10 years will be presumed to be totally disabled due to pneumoconiosis” as long as the miner meets any one of four medical requirements. § 727.203(a). Part (b) provides that the presumption is rebutted if the Administrative Law Judge (ALJ) determines that the claimant is doing or could do his usual coal mine work or comparable gainful work; that the disability “did not arise in whole or in part out of coal mine employment”; or that the “evidence establishes that the miner does not, or did not, have pneumoconiosis.” § 727.203(b).
Under § (a)(1), the presumption is triggered when “[a] chest [X ray], biopsy, or autopsy establishes the existence of pneumoconiosis.” Under § (a)(4), the presumption is triggered when “[o]ther medical evidence, including the documented opinion of a physician exercising reasoned medical judgment, establishes the presence of a totally disabling respiratory or pulmonary impairment.” The regulation does not refer to “the X-ray evidence taken as a whole,” or “the weight of documented opinions of physicians” — the inquiry mandated by the Court’s interpretation — but rather to an X ray, or a physician’s opinion. Moreover, the regulation does not provide that a claimant who presents a single qualifying piece of evidence may be presumed to be totally disabled because of pneumoconiosis, but rather provides that such a claimant “will be presumed” totally disabled because of the disease. The regulation thus expressly commands that a single qualifying X ray or a single documented physician’s opinion will trigger the presumption.1
*163The Court attempts to evade the plain meaning of the regulation by placing overriding emphasis on the regulation’s use of the word “establishes.” The Court interprets “establish” to mean that a claimant must prove each of the medical requirements by a preponderance of the evidence.2 Under this interpretation, an ALJ must weigh conflicting like-kind evidence before invoking the presumption. A more natural reading of “establish” in the context of the presumption-invocation stage, however, is simply that the ALJ must determine whether a claimant has come forward with an X ray, a set of ventilatory or blood gas studies, or a physician’s opinion that meets the requirements and standards of the regulation. In other words, an ALJ determines whether the single item of evidence “establishes the existence of pneumoconiosis” or “establishes the presence of a totally disabling respiratory or pulmonary impairment” by referring to the strict reliability and authenticity requirements of the regulations, see, e. g., 20 CFR §§410.428, 727.206 (1987), and, with regard to ventilatory and blood gas studies, to the qualifying standards set out in § 727.203(a).
The Court argues that § (a)(1) cannot have been intended to refer to a single item of evidence because an X ray is probative only when it is interpreted by a qualified expert. The Court reasons that because the presumption is invoked by an X ray, and not by an expert’s reading, an ALJ may have to consider different interpretations of the same X ray. From this premise, the Court concludes that “[j]ust as the ALJ must weigh conflicting interpretations of the same X ray . . . , there would seem to be no reason why he must ignore *164all X rays in a series except one.” Ante, at 148-149 (footnote omitted). No reason except the regulatory language. Whatever the merit of the Court’s conclusion that conflicting readings of the same X ray must be weighed prior to invoking the presumption (for that question is not before the Court), the regulation’s plain language requires that the presumption be invoked when a single X ray is read only as positive. In addition, contrary to the Court’s characterization, additional X rays that indicate the absence of pneumoconiosis are not “ignored”; they are fully considered by the ALJ during the rebuttal stage, when all evidence against the presumed existence of pneumoconiosis is brought to bear.3
The Court of Appeals ruling that a single qualifying test or medical opinion is sufficient to invoke the presumption is further supported by the comments that the Secretary of Labor issued in connection with the final promulgation of the regulations. In addressing the standard of rebuttal, the Secretary stated:
*165“[T]he Department cannot, as has been requested by some, look for the single item of evidence which would qualify a claimant on the basis of the interim presumption, and ignore other previously obtained evidence. This does not mean that the single item of evidence which establishes the presumption is overcome by a single item of evidence which rebuts the presumption.” Notice of Final Rulemaking under the Black Lung Benefits Reform Act of 1977, 43 Fed. Reg. 36826 (1978) (hereinafter Notice of Rulemaking) (emphasis added).
The Director’s current position conflicts with this strong evidence of regulatory intent.
Another compelling reason to reject the Director’s interpretation is that it conflicts with the requirement in part (b), the rebuttal section, that in “adjudicating a claim under this subpart, all relevant medical evidence shall be considered.” § 727.203(b). The Director’s interpretation turns the regulation on its head, requiring that all relevant medical evidence be submitted and weighed at the invocation stage, but severely restricting the consideration of medical evidence during the rebuttal stage. In the Director’s view, the presumption is triggered only when the weight of evidence in one of the categories of medical evidence in part (a) proves the fact specified in that category. For instance, the “fact proved” under § (a)(1) is the existence of pneumoconiosis, and under § (a)(4) it is the presence of a totally disabling respiratory or pulmonary impairment. Because these facts have been proved in the invocation stage, relitigation in the rebuttal stage through like-kind evidence is foreclosed. Brief for Federal Respondent 14-15. Of course, nonmedical evidence could be presented in these instances, but this hardly conforms to the mandate that “all relevant medical evidence” be considered in the rebuttal stage.
The Court argues that the placement of the “all relevant medical evidence” requirement was inexact, and that the regulation requires only that all relevant medical evidence be *166considered at some point in the evaluation of a claim, whether it be during the invocation or rebuttal stage. Yet if the Secretary intended that the “all relevant medical evidence” language apply to both stages of the evaluation process, it is remarkable that he placed the language in the introduction to the rebuttal section. It would have been a simple matter, if such were the Secretary’s intent, to place the “all relevant medical evidence” language at the beginning of § 727.200. I see no reason to assume such inartful drafting. Moreover, comments by the Secretary accompanying the final promulgation of the regulations conflict with the Court’s interpretation. The Secretary stated: “The many comments which urge that all relevant evidence should not be considered in rebutting the interim presumption must also be rejected. . . . [T]he Social Security regulations . . . similarly do not limit the evidence which can be considered in rebutting the interim presumption. ” Notice of Rulemaking 36826 (emphasis added). These comments demonstrate that the Secretary understood the language to apply directly to the rebuttal section.
In addition, the Director’s approach renders virtually useless one of four grounds for rebuttal in part (b). Under § (b)(4), eligibility for benefits is rebutted if all relevant medical evidence establishes that the miner does not have pneumoconiosis. Yet in the Director’s view all relevant medical evidence has already been considered at the presumption-invocation stage. It is only when the evidence presented during the invocation stage is mismatched, as for example when the claimant produces qualifying blood-gas-studies evidence and the mine operator produces negative X rays, that the presumption may be triggered and rebuttal evidence under § (b)(4) is available that has not already been weighed. Moreover, as the Director acknowledges, if a claimant invokes the presumption under § (a)(1) by a preponderance of X-ray, biopsy, or autopsy evidence, as a practical matter there is no further evidence that the coal mine operator could submit to rebut the presumed existence of pneumoconiosis *167because such § (a)(1) evidence is the most reliable method of diagnosing the disease. See Brief for Federal Respondent 24, n. 22. The Court counters that “[njothing in the regulation requires each rebuttal subsection to be fully available in each case.” Ante, at 150. This is of course true. Yet it is extraordinary that the regulation would intend to make the rebuttal stage an often useless exercise with respect to the central aspect of a valid claim: whether a miner suffers from pneumoconiosis.
II
In addition to running afoul of the regulatory language and structure, the Director’s reading of the regulation creates a needlessly complex regulatory scheme that blurs the distinction between the presumption-invocation and rebuttal stages. Under the Director’s interpretation, when the weight of evidence in one of the medical-evidence categories invokes the presumption, then the same evidence cannot be considered during rebuttal to challenge the existence of the fact proved, but it may be considered if relevant to rebut one of the presumed elements of a valid claim for benefits. The Director’s approach subjects the ALJ to a mesmerizing swirl of evidentiary rules. If the presumption is invoked under § (a)(1), then X-ray evidence may not be considered regarding the existence of pneumoconiosis, but may be considered, if relevant, on the issues whether the miner is totally disabled or whether the disease arose from coal mine employment. Similarly, if the presumption is invoked under § (a)(4), then medical evidence, including physicians’ opinions, may not be considered on the issue of total disability, although it may be considered on the issues of the existence of pneumoconiosis and causation by coal mine employment. Finally, if the presumption is invoked using blood gas and ventilatory studies evidence under § (a)(2) and § (a)(3), the same evidence may be considered again regarding each rebuttal category, because the proved facts are not elements of a valid claim for benefits. The Director’s interpretation thus *168creates a procedural morass that could not have been intended by the regulation’s two-step inquiry.
By contrast, the Court of Appeals interpretation is marked by its simplicity. Under this approach the ALJ first determines whether the claimant has come forward with a qualifying medical test or physician’s opinion and, if so, proceeds to the rebuttal stage. At this point all relevant evidence must be considered, and the mine operators may rebut the presumed existence of pneumoconiosis, total disability, and causation by coal mine employment on the basis of all the grounds provided by § 727.203(b).
Ill
The Court’s willingness to accept the Director’s interpretation of the regulation is based, I believe, on a misperception of the problem Congress and the Department of Labor were trying to alleviate with the interim presumption. Pneumoconiosis is an elusive and progressive disease. Congress was deeply concerned about the difficulty of diagnosing pneumoconiosis and the dearth of medical-testing facilities available to miners. Testimony before congressional committees and by Members of Congress repeatedly emphasized the unreliability of negative test results. As this Court stated in Usery v. Turner Elkhorn Mining Co., 428 U. S. 1, 31-32, and n. 33 (1976), “Congress was presented with significant evidence demonstrating that X-ray testing that fails to disclose pneumoconiosis cannot be depended upon as a trustworthy indicator of the absence of the disease,” whereas there was no “authoritative indications that X-ray evidence of the presence of pneumoconiosis is untrustworthy. ”4 Juxtaposed with the difficulties in diagnosing pneumoconiosis was evi*169dence that the disease was rife among long-term coal miners. Congressman Paul Simon noted one study that found that autopsies of 400 coal miners with more than 20 years’ experience showed that 90-95% of them had pneumoconiosis. House Committee on Education and Labor, Black Lung Benefits Reform Act and Black Lung Benefits Revenue Act of 1977, 96th Cong., 282-283 (Comm. Print 1979). Testimony before the Senate estimated that as many as 50% of all coal miners will eventually become disabled from pneumoconiosis. See Hearings on S. 355 et al. before the Subcommittee on Labor of the Senate Committee on Labor and Public Welfare, 91st Cong., 1st Sess., pt. 2, pp. 641, 856 (1969) (statements of Dr. I. E. Buff and Dr. Leon Cander).
The Court recognizes that Congress was especially concerned with the difficulties miners face in showing they suffer from pneumoconiosis. The Court reasons, however, that “Congress intended that those long-term miners who can show that they are truly diseased should have to prove no more. But if a miner is not actually suffering from the type of ailment with which Congress was concerned, there is no justification for presuming that that miner is entitled to benefits.” Ante, at 158 (footnote omitted). Yet it is the difficulty in showing whether a miner is “truly diseased” that Congress found so troubling. No one disputes that the case file of a miner suffering from pneumoconiosis may include negative X rays, negative ventilatory studies, negative blood gas studies, and negative opinions by physicians. The interim presumption was designed to shift some of the risk of faulty test results from the miner to the employer. The evidence of high incidence of pneumoconiosis among long-term coal miners, coupled with the difficulties encountered in diagnosing the disease, gave the Department of Labor good reason for shifting this burden by presuming total disability due to pneumoconiosis based on findings of a single positive medical test or physician’s opinion. The Director’s current *170interpretation, which the Court today accepts, undermines that policy decision.
IV
The Court is correct that the agency’s interpretation of its own regulations is entitled to deference. See, e. g., Bowles v. Seminole Rock & Sand Co., 325 U. S. 410, 414 (1945). But deference has its bounds. It is not a license for an agency effectively to rewrite a regulation through interpretation. An agency must abide by its regulations as written until it rescinds or amends them. See United States v. Nixon, 418 U. S. 683, 695-696 (1974). The Director’s interpretation of the interim presumption is contrary to the plain language of the regulation, conflicts with comments of the Secretary accompanying the Anal promulgation of the regulation, and creates an unnecessarily complex regulatory scheme. Because I view the agency’s interpretation as plainly inconsistent with the regulatory language and history, I would not defer.
I accordingly dissent, and would affirm the judgment of the Court of Appeals.
With respect to the medical requirements under §§727.203(a)(2) and (a)(3), the regulation uses the plural, referring to “ventilatory studies” and “blood gas studies.” The use of the plural, however, only reflects the fact that a qualifying ventilatory and blood gas test consists of a set of many studies. See 20 CFR §§718.103, 718.105 (1987). As with § (a)(1) and § (a)(4), the presumption is triggered when a claimant presents a single set *163of ventilatory or blood gas studies that meet the qualifying standards set out in part (a).
Both the Director and the Court state that the agency has “long adhered to the ‘true doubt’ rule, which requires that the claimant prevail on those issues as to which the evidence is in equipoise.” Brief for Federal Respondent 16-17 (citation omitted). The Director has failed to bring to our attention, however, one instance in which the true-doubt rule actually has been applied by an ALJ in evaluating a miner’s claim.
The Court also argues that the Court of Appeals reads the invocation section “as though it merely required an X ray that ‘constitutes evidence of the presence of pneumoconiosis,’” and asserts that “[h]ad that been the Secretary’s intent, presumably he would have used that language as he did elsewhere to explain that meaning.” Ante, at 147. The Court notes that §718.102(e), a regulation defining standards required of X rays, refers to an X ray “that shall constitute evidence of the presence or absence of pneumoconiosis.” This argument arises from the Court’s unwillingness to recognize that the meaning of “establish” is informed by the requirements of the presumption-invocation stage. The Court is correct that in context of the presumption-invocation stage, a single item of evidence that “constitutes evidence of pneumoconiosis” will also “establish the existence of pneumoconiosis.” Given that the inquiry at the invocation stage is whether a single item of evidence is qualifying, the functional equivalence of these two terms is to be expected. As for the regulation cited by the Court defining X-ray standards, it guides the evaluation of all claims for black lung benefits, not just claims that fall under the interim regulations. That these standards speak of X rays that “constitute evidence of pneumoconiosis,” rather than X rays that “establish the existence of pneumoconiosis,” says nothing about whether these two terms have identical meanings in the context of the presumption-invocation stage.
For example, a study cited in a Senate Report found that “approximately 25 percent of a random sample of some 200 coal miners whose medical records based upon X-ray findings showed no coalworker’s pneumoconiosis were found on post mortem examination to have the disease.” S. Rep. No. 92-743, p. 12 (1972).