In July 1986 the Occupational Health and Safety Administration of the Department of Labor (“OSHA” or the “Secretary”) issued revised standards governing workers’ exposure to asbestos, tremolite, anthophyllite and actinolite (together referred to simply as “asbestos”) pursuant to its authority under § 6(f) of the Occupational Safety and Health Act of 1970 (the “Act”), 29 U.S.C. § 655(f) (1982). 51 Fed.Reg. 22,612 et seq. (1986), codified at 29 C.F.R. § 1910.1001 (1987) (general industry standards) and 29 C.F.R. § 1926.58 (1987) (construction industry standards).
In these consolidated appeals, the Asbestos Information Association/North America and other petitioners representing industry (collectively referred to as the “AIA”) attack the Secretary’s findings that asbestos under the prior permissible exposure level (“PEL”) caused significant risk, question the technological feasibility of the *312new PEL, challenge the rule against the spraying of asbestos-containing products, and assert that the record does not support OSHA’s decision not to require smoking-related programs and controls. Two unions, the American Federation of Labor and Congress of Industrial Organizations (“AFL-CIO”) and the Building and Construction Trades Department, AFL-CIO (“BCTD”), maintain that a lower PEL is technologically feasible for most industry subgroups, and argue that a short term exposure limit (“STEL”) is feasible and would further reduce health risks. BCTD further claims that the Secretary acted unlawfully in refusing to adopt their recommendations relating to many subsidiary aspects of the rule, including “action levels,” compliance methods, monitoring, reporting, warnings, medical surveillance, questionnaires, records, hygiene, and respirator use.
We uphold the Secretary’s findings concerning significant risk and the feasibility of the new PEL, and find many of BCTD’s challenges to details within the various provisions to be without merit. We cannot, however, find substantial evidence supporting several of the agency’s conclusions: its categorical ban on the spraying of asbestos-containing products and its decisions not to adopt a lower PEL for certain major industrial subgroups, a STEL, smoking-related regulations, and some of BCTD’s suggested safety measures. We therefore remand the case to the agency for reconsideration of these issues.
I. Background
Asbestos is a generic term applied to a number of naturally occurring hydrated, fibrous, silicate minerals. Final Regulatory Impact and Regulatory Flexibility Analysis of the Revised Asbestos Standard (OSHA Office of Regulatory Analysis, June 11, 1986) (“Impact Analysis”), Joint Appendix (“J.A.”) at 1541, 1542. It is characterized by flexibility, high tensile strength, and resistance to heat and other destructive conditions. Id. Asbestos is classified into different grades by fiber length and resilience. Id. at 1543. The longest and strongest fibers are generally used to make textiles, electrical insulation, and pharmaceutical and beverage filters. Id. Medium length fibers are used in the production of asbestos cement (A/C) pipe, A/C sheet, clutch facings, brake linings, asbestos paper, packings, gaskets and pipe coverings. Id. The shorter fibers are used primarily as reinforcers in plastics, floor tiles, coatings and compounds, special papers and roofing felts. Id.
Unfortunately, asbestos creates very serious health hazards. During the production of asbestos products, and in their later handling and use, tiny asbestos fibers are released into the air. Impact Analysis, J.A. at 1552. They are also released when old asbestos-based insulation is removed or disturbed in the demolition or repair of buildings and ships. Id. at 1562, 1565, 1567. When inhaled, the fibers settle in the lungs and often travel to other organs.
Inhalation of the fibers causes, or is at least causally related to, a number of serious and often fatal diseases. These include lung cancer, asbestosis (a chronic and disabling lung disease), mesothelioma (cancer of the lining of the chest cavity or abdomen), and gastrointestinal cancer. 51 Fed.Reg. at 22,615/3. It is also associated with an increased risk of cancer of the esophagus, stomach, colon, kidneys, larynx, pharynx, and buccal cavity. Id. at 22,616/1. In the preamble to the standards here at issue, the Secretary emphasized that “OSHA is aware of no instance in which exposure to a toxic substance has more clearly demonstrated detrimental health effects on humans than has asbestos exposure.” Id. at 22,615/3.
In June 1972 OSHA issued a standard which established an immediate PEL of 5 fibers per cubic centimeter (f/cc) and a ceiling of 10 f/cc. (A cubic centimeter is one-millionth the size of a cubic meter. Thus 5 f/cc means 5,000,000 fibers per cubic meter. The convention appears to be to express the limit in f/cc, even though this conjures up a far less vivid image than does the measurement in fibers per meter.) The rule also provided for reduction of the PEL to 2 f/cc as of July 1976. We substantially upheld the standard in Industri*313al Union Department, AFL-CIO v. Hodgson, 499 F.2d 467 (D.C.Cir.1974). This initial standard was based primarily on evidence of a causal relationship between asbestos inhalation and asbestosis. See 51 Fed.Reg. at 22,614/2.
In 1975, based on new evidence of the carcinogenic properties of asbestos and advances in the technology of monitoring and protection, OSHA proposed a new rule that would reduce the PEL to 0.5 f/cc and the ceiling limit to 5 f/cc. 40 Fed. Reg. 47,652 (1975). The proposal was based on the premise that if (as the agency believed) there was no level of asbestos in the workplace that would be absolutely safe, it was authorized and required to set a PEL as low as was economically and technologically feasible. 51 Fed.Reg. at 22,614/2. The Supreme Court rejected this concept of absolute safety in Industrial Union Department, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 639-52, 100 S.Ct. 2844, 2862-69, 65 L.Ed.2d 1010 (1980) (“Benzene ”), in favor of the view that the Secretary may promulgate standards only “where a significant risk of harm exists.” Id. at 652, 100 S.Ct. at 2869. The 1975 proposal seems to have died.
In April 1984, acting in part in response to recommendations of the National Institute for Occupational Safety and Health (NIOSH), see NIOSH, Revised Recommended Asbestos Standard (Dec. 1976), J.A. at 119, and a NIOSH/OSHA task force, 51 Fed.Reg. at 22,614/3, OSHA issued a notice of proposed rulemaking suggesting adoption of either a 0.5 f/cc or a 0.2 f/cc PEL, 49 Fed.Reg. 14,116/2 (1984), and raising the question of whether a 0.1 f/cc PEL might be feasible. Id. at 14,117/1. OSHA held public hearings and received written submissions before and after the hearings. The process generated a record with over 340 exhibits and approximately 55,000 pages. As a result of its analysis, OSHA made a finding of a “significant risk” at the then-prevailing standard of 2 f/cc, id. at 22,647/3, a finding that is a prerequisite to more stringent regulation under the plurality opinion in Benzene, 448 U.S. at 614-15, 100 S.Ct. at 2849-50. See also American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490, 505-06 n. 25, 101 S.Ct. 2478, 2488-89 n. 25, 69 L.Ed.2d 185 (1981) (“Cotton Dust ”) (majority opinion seeming to approve significant-risk criterion). It estimated that at the PEL of 2 f/cc asbestos use generated a lifetime excess risk per 1000 workers of 64 excess deaths from asbestos-related cancers, and an excess incidence of 50 cases of asbestosis. 51 Fed.Reg. at 22,644/3. Although it found that reducing the PEL to 0.2 f/cc would reduce the risk of cancer and the risk of asbestosis tenfold, id. at 22,644, Table 6, it concluded that the risks remaining at 0.2 f/cc were still significant. Id. at 22,648/1.
The revised standards specify that the new 0.2 PEL is to be achieved primarily through engineering and work practice controls. 29 C.F.R. §§ 1910.1001(f), 1926.-58(g). Where it is not “feasible” for an employer to meet the PEL through these measures, it must resort to supplementary respirator use. 29 C.F.R. §§ 1910.-1001(f)(1)(iii), 1926.58(g)(1)(ii). The standards also include various ancillary requirements, such as exposure monitoring, medical examinations, labelling, protective clothing, designated regulated areas, clean change rooms, decontamination areas, separate lunchrooms or lunch areas, and employee training. 29 C.F.R. §§ 1910.1001(d), (e), (hHn); 1926.58(d), (e), (f), (i)-(m). The regulations establish an “action level” of 0.1 f/cc. If exposures exceed this level, employers must initiate monitoring, medical surveillance, and employee information and training programs. 51 Fed.Reg. at 22,679/3. The more rigorous programs, such as establishment of regulated areas, and provision of respirators, protective clothing, and hygiene facilities, are triggered by exposures at or above the PEL. 29 C.F.R. §§ 1910.1001(e), (f), (h) & (i), 1926.58(e), (g) & (j).
Section 6(b)(5) of the Act provides that when the Secretary chooses to regulate toxic materials he must
set the standard which most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impair*314ment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.
29 U.S.C. § 655(b)(5). In Cotton Dust the Court held that the “feasibility” standard of § 6(b)(5) does not require the Secretary to balance cost and benefit in defining a standard, and clearly manifested the Court’s belief that the Act did not permit the Secretary to do so. 452 U.S. at 506-22, 101 S.Ct. at 2489-97.
The Act provides for review under the substantial evidence standard. 29 U.S.C. § 655(f). This has a somewhat special meaning as applied to essentially legislative judgments made in informal rulemaking, such as the agency’s decisions here on “significant risk” and “feasibility.” Recognizing both OSHA’s superior technological capacity and its broad legislative mandate, the courts have not required the agency to “support its findings ... with anything approaching scientific certainty.” Benzene, 448 U.S. at 656, 100 S.Ct. at 2871. They have instead tailored their review to the nature of the judgment, requiring OSHA to “identify relevant factual evidence, to explain the logic and the policies underlying any legislative choice, to state candidly any assumptions on which it relies, and to present its reasons for rejecting significant contrary evidence and argument.” United Steelworkers of America v. Marshall, 647 F.2d 1189, 1207 (D.C.Cir.1980) (“Steelworkers ”), cert. denied, 453 U.S. 913, 101 S.Ct. 3148, 69 L.Ed.2d 997 (1981).
We tackle industry challenges first, then the unions’. The only real overlap is the general issue of evaluating agency explanations for rejecting proposed measures for further reducing risk. See especially part II.E infra.
II. Industry Challenges
A. Significant Risk
OSHA selected the epidemiological studies it regarded as the most relevant and reliable, and then, using the assumption that the dose-response relationship was linear, generated a table describing the relationship between exposure to varying levels of airborne asbestos and contraction of various asbestos-related diseases. 51 Fed. Reg. at 22,644. As noted above, it found that the existing standards of 2 f/cc carried an excess cancer mortality risk of 64 deaths per 1000 asbestos workers, assuming a 45-year working lifetime. Id. at 22,-644, Table 6. It found this excess risk significant. 51 Fed.Reg. at 22,647/3. Even under the assumption of a 20-year working life, it found the risk at the existing PEL to be 44 extra cancer deaths per 1000 (which it also found significant). Id. It believed that a reduction in the PEL to 0.2 f/cc would decrease the excess mortality risk and asbestosis risk by 90% under both 45-year and 20-year exposure assumptions. Id.
OSHA acknowledged that a significant risk of contracting asbestos-related diseases would remain at a PEL of 0.2 f/cc or even 0.1 f/cc. Id. at 22,648/1. It explicitly stated that it chose the 0.2 level not because it would eliminate significant risk, but rather because it was the lowest feasible level. Id.
AIA attacks the finding of significant risk as unsupported by substantial evidence, identifying four errors, each basically one of analysis. We address them seriatim.
1. Exposure duration. AIA claims that because the median duration of employment in representative industry sectors is approximately five years, the Secretary erred in assuming a 45-year work-life in performing his risk assessment. Although the assumption may well be unrealistic, it appears to conform to the intent of Congress. For toxic substances, § 6(b)(5) requires the Secretary to select the standard designed to assure that “no employee will suffer material impairment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.” 29 U.S.C. § 655(b)(5) (emphasis added). Even if it is only the rare worker who stays with asbestos-related tasks for 45 years, that *315worker would face a 64/1000 excess risk of contracting cancer; Congress clearly authorized OSHA to protect such a worker. (Whether OSHA might adopt more realistic assumptions in the face of the statute is an issue we need not address.) We note that even if OSHA assumed an exposure duration of only one year, its tables indicate an excess cancer mortality risk of three deaths per 1000 workers at the 2 PEL. 51 Fed.Reg. at 22,644, Table 6. We doubt if a court could find that risk insignificant as a matter of law. See Benzene, 448 U.S. at 655, 100 S.Ct. at 2870-71 (“if the odds are one in a thousand that regular inhalation of gasoline vapors that are 2% benzene will be fatal, a reasonable person might well consider the risk significant____”).
2. Inclusion of smokers. AIA claims that because smoking and asbestos have synergistic effects (i.e., the cancer risks of a smoking worker exposed to asbestos are greater than the mere sum of the risks of smoking and asbestos), see 51 Fed.Reg. at 22,616/1, OSHA’s reliance on studies that included a high proportion of smokers led it to overestimate the risks posed by asbestos. The claim is not that OSHA failed to control for smoking; it is that the agency considered smokers’ incremental risks from asbestos and should not have done so. But § 6(b)(5) calls on OSHA to set standards such that “no employee” will experience the forbidden level of risk. We understand the employers’ aggravation that they are being forced to bear part of the burden imposed by employees’ decisions to smoke, but we do not think that at this stage of American history smokers can be regarded as so far beyond the pale as to require OSHA to disregard them in computing the risks of asbestos.
As we understand the methodology, the inclusion of smokers within the sample correctly compares the risk that a heterogeneous group of workers normally faces in the workplace with the risk that this same group faces when working with asbestos. This follows from the finding that while smokers face a greater absolute risk from exposure to asbestos than do nonsmokers, asbestos posed the same relative risk to the two groups. 51 Fed.Reg. at 22,650/1. Thus, suppose that lung cancer risks are as follows:
Incidence per 1000 workers
Non-smoker, no asbestos X
Smoker, no asbestos 10X
Non-smoker, asbestos 2X
Smoker, asbestos 20X
If there were 4000 workers, 1000 in each of these four cells, the total lung cancer deaths will be 33X, compared to 22X for 4000 workers of whom none was exposed to asbestos but of whom half were smokers. Excess deaths from asbestos would be 11X/4000. It is clear that the deaths from smoking workers exposed to asbestos swamp the number that would result if no one smoked. (Excess deaths from asbestos if no one smoked would be 8X minus 4X, or 4X.) Yet, assuming OSHA is within the law in its decision not to banish smokers from consideration, there is no methodological error.
As in the case of exposure duration, acceptance of the AIA’s smoking argument would not undermine the finding of significant risk. Smoking has not been linked to mesothelioma, and the Secretary found a risk of 1.64 excess mesothelioma deaths per 1000 even at the new 0.2 f/cc PEL. Id. at 22,650/2. This he expressly found significant, id., and we could not probably overturn that judgment as a matter of law.
3. Disregard of gap between compliance levels and the PEL. When estimating excess mortality at a particular PEL, OSHA assumes a constant particulate concentration level equal to that PEL. AIA argues that in order for a firm to ensure that its peak levels are within a PEL, its average particulate levels must be well below that limit. For example, one of its experts concluded that to achieve compliance with any given PEL, “employers generally aim to achieve average airborne asbestos levels between one-half and one-quarter of the PEL____” J.A. at 1047. They assert, therefore, that an accurate risk assessment ought to measure the danger posed not by a proposed ceiling concentration, but rather by the average concentration that such a ceiling would bring about in the real world.
*316We can imagine circumstances where the claim would be successful. If compliance with a given standard entailed average exposure levels that were only a fraction of that level, calculation of significant risk at the nominal rather than the real average would strike one as arbitrary and capricious. But that seems not to be this case. First, the agency asserts that AIA has greatly exaggerated the proportion of uncontrollable exposure level fluctuation. See OSHA Brief at 63-65. But see AIA Reply Brief at 13 n. 5. If OSHA is correct, then AIA’s suggestion as to the likely ratio between a PEL and the resulting average exposure level will be similarly exaggerated. Second, even if we accept the view that average exposures will be one-half the ceiling, the new PEL of 0.2 would entail an average exposure of 0.1, with an estimated excess risk of three deaths per 1000 workers, which could well be found “significant.”
We note that this claim, unlike the first two, does not founder on any legal obstacle. While the Act plainly allows (and possibly mandates) use of unrealistic exposure durations and inclusion of smokers, there is no legal basis for totally disregarding a gap between real-world average exposure and nominal legal ceilings. AIA loses here on the facts. The Secretary will doubtless want to make sure that this potential risk of arbitrariness does not become actual in the future; a step toward that goal would be for him to make his own calculations of the relation between legal ceilings and implied real exposures.
4. Loading the dice. Finally, AIA disputes many of the Secretary’s technical assumptions, and charges that “OSHA’s potency assessment repeatedly resolved hotly disputed scientific issues in favor of higher potency estimates.” AIA Brief at 60. For example, it challenges OSHA’s adoption of a linear dose-response model for extrapolating risk at low exposure levels, OSHA’s conclusions on the applicability of various lung cancer risk studies, and OSHA’s decision to pool amphibole and chrysotile studies in developing a mesothelioma risk model. See id. at 25-26.
In assessing these claims, it is clear that our function is not to decide what assumptions or findings we would make were we in the Secretary’s position. Cf. Baltimore Gas and Electric Co. v. NRDC, 462 U.S. 87, 105, 103 S.Ct. 2246, 2256, 76 L.Ed.2d 437 (1983) (“It is not our task to determine what decision we, as Commissioners, would have reached [on policy judgment as to correct treatment of nuclear waste disposal in Nuclear Regulatory Commission’s individual licensing decisions].”). When called upon to review technical determinations on matters to which the agency lays claim to special expertise, the courts are at their most deferential. Id. at 103, 103 S.Ct. at 2255. Our primary duty is merely to scrutinize the record to ensure that the Secretary has made his findings of fact on the basis of substantial evidence and has provided a reasoned explanation for his policy assumptions and conclusions. Steelworkers, 647 F.2d at 1207.
Here, we find that in each area of contention the Secretary has identified the studies and authorities on which he has relied, and has described in some detail his reasons for choosing between competing alternatives. See, e.g., 51 Fed.Reg. at 22,-632-38 (explaining his choice among various lung cancer epidemiological studies and his decision to adopt a linear dose-response model); id. at 22,638-42 (explaining the derivation of his mesothelioma risk projections and his reason for using risk estimates from studies of mixed amphibole/chrysotile exposures as well as single-fiber type exposure studies). We conclude that the Secretary has adequate support for his scientific assumptions and findings.
On a somewhat different plane, AIA contends that the Secretary’s risk analyses were consistently marred by a pronounced bias towards worker safety. AIA Brief at 60. Even were we to accept this contention as true, no relief would be forthcoming. The Supreme Court has made clear that “so long as they are supported by a body of reputable scientific thought, [OSHA] is free to use conservative assumptions in interpreting the data with respect to carcin*317ogens, risking error on the side of overprotection rather than under protection.” Benzene, 448 U.S. at 656,100 S.Ct. at 2871. Of course such leaning towards safety may sometimes have the perverse effect of increasing rather than decreasing risk, see, e.g., Nichols & Zeckhauser, “The Perils of Prudence,” Regulation 13 (Nov/Dec 1986); see also Public Citizen v. Young, 831 F.2d 1108, 1113 (D.C.Cir.1987); see generally A. Wildavsky, Searching for Safety (forthcoming), but AIA has demonstrated neither the existence of such an effect nor (what would be necessary for us to reverse) that it exceeds what Congress was ready to accept.
B. Technological Feasibility
The AIA maintains that a PEL of 0.2 f/cc is not technologically feasible. First, it objects to the methodology of measurement. The new rule adopts for the first time a protocol intended to standardize measurement of airborne asbestos. The agency’s aim is to eliminate variations in measurement due to (for example) different sampling techniques, equipment, mounting and clearing procedures, and rules for counting fibers. AIA argues that as the evidence of feasibility was derived from studies conducted without the new protocol, the results of those studies cannot be reliable indicators of what can actually be attained when workplaces are monitored under the new protocol. It believes that OSHA ought to have established comparability between that method and those used in the feasibility studies on which OSHA relied. Specifically, it argues that this should have been established by “side-by-side” testing, i.e., by conducting tests under the new protocol and under others (the ones used when generating the feasibility data), in the same place and at the same time.
We do not believe the absence of such testing undermines the existence of substantial evidence for the feasibility of the 0.2 f/cc PEL. First, although side-by-side comparisons may be appropriate when OSHA changes protocols, we note that here OSHA has merely adopted a universal protocol where none had previously prevailed. Second, so far as appears, the new protocol is not markedly different from one formulated by NIOSH in 1974, 51 Fed.Reg. at 22,686/1, nor from the protocol formerly used in OSHA laboratories, nor indeed from a protocol propounded by AIA itself, see OSHA Brief at 72-75. Third, in order to guard against random sampling error, the agency has determined to add the full value of the standard sampling error under the protocol to its measurements in determining compliance. 51 Fed.Reg. at 22,654/2. Thus, although the PEL is 0.2 f/cc, OSHA will assert a violation only if it measures a level in excess of 0.25 f/cc. This practice should offset any unfairness that may exist because of OSHA’s standardization of previously variable measurement practices. Finally and vitally, while OSHA maintains that the use of the new method will not result in significantly increased fiber counts, see OSHA Brief at 73 & n. 35, AIA has pointed to nothing that suggests otherwise. It is hard to believe that the claim has substance when neither common sense nor the claim’s propounder offers anything to support such a belief.
The AIA makes two attacks independent of the measurement problem. It concedes that “asbestos exposures in manufacturing operations utilizing the most effective engineering and work practice controls will sometimes be below 0.1 f/cc” and “will be below 0.2 f/cc ‘most of the time.’ ”. AIA Reply Brief at 12. But it argues that (1) compliance in some sectors will require reliance on respirators, and that (2) given uncontrolled random fluctuations and an enforcement policy that punishes employers for isolated counts above the PEL, a standard that is only attainable “most of the time” is not truly feasible.
1. Need for respirators in some sectors. OSHA has determined that the 0.2 PEL is feasible without regular use of respirators for manufacturing processes involving more than 90% of the relevant manufacturing workforce. 51 Fed.Reg. at 22,-654, Table 11. As noted, AIA concedes that concentrations can be below 0.2 f/cc “most of the time.”
*318We believe this to comply with the standards of feasibility laid down in Steelworkers. There we observed that OSHA
must prove a reasonable possibility that the typical firm will be able to develop and install engineering and work practice controls that can meet the PEL in most of its operations____ Insufficient proof of technological feasibility for a few isolated operations within an industry, or even OSHA’s concession that respirators will be necessary in a few such operations, will not undermine this general presumption in favor of feasibility.”
647 F.2d at 1272.
As to operations that cannot be brought into full compliance with the PEL through engineering and work practices, OSHA has provided that the employer must use them to attain the lowest achievable levels and supplement them with respirator use. 29 C.F.R. §§ 1910.1001(f)(1)(ii), 1926.58(g)(1)(ii). This seems to comply with our conclusion in Steelworkers that in these refractory operations firms must use engineering and work practice methods “to the extent feasible.” 647 F.2d at 1272. Here, in fact, OSHA has identified certain processes as ones in which it is not presumptively possible to attain 0.2 f/cc through engineering controls. For these the agency has set a 0.5 f/cc PEL until such time as it believes the lower 0.2 f/cc level can be attained. 29 C.F.R. § 1910.1001(f)(1)(iii). This seems to go beyond what Steelworkers required.
2. Random uncontrollable fluctuations. AIA’s complaint here is that firms for which compliance is generally feasible, and that are indeed normally in compliance, will be unfairly subject to sanctions when a random and uncontrollable upward fluctuation coincides with an inspector’s visit. OSHA conceded in the preamble that an employer whose processes were beset by legitimately uncontrollable random fluctuations might find it infeasible to meet the PEL 100% of the time.1 It has adopted a policy to account for this situation. An employer using state-of-the-art work practices and engineering controls can, when found in apparent violation, seek to show the inspector that the one-day measurement was unrepresentively high. Upon such a showing,
OSHA may reinspect the workplace and measure the employees’ exposure or may decide not to issue a citation, unless OSHA has reason to believe that there are circumstances within the employer’s control to account for the high exposure measurement.
51 Fed.Reg. at 22,653/3.
The AIA attacks this proposed enforcement policy as relying on too much human discretion and as defeating § 6(b)(5)’s requirement that standards be expressed in terms of “objective criteria.” That requirement, however, is only conditional: objective criteria are to be used “[w]herever practicable.” OSHA has considered the adoption of “objective criteria,” but has explicitly found that route to be impracticable on the ground that it would not be able to bring to bear on the outcome the OSHA inspector’s “observations about the quality of the employer’s sampling and analytic program and the asbestos control, housekeeping, and training programs which OSHA believes are equally important in showing why fluctuations occur.” 51 Fed.Reg. at 22,653-54. Moreover, if OSHA is permitted to adopt a standard that some employers assuredly will not be able to meet some of the time, with the employers limited to challenging feasibility at the enforcement stage, see Steelworkers, 647 F.2d at 1273 & n. 125, it is difficult to find legal objection in an OSHA enforcement policy that attempts to take account of particular feasibility difficulties at the pre-citation stage.
Industry also raises a second, slightly baffling, objection to OSHA’s enforcement policy. It asserts that the employer’s right to claim individual infeasibility in an enforcement proceeding
*319only protects employers against a citation for violating the requirement that the PEL be achieved by use of ‘feasible’ engineering and work practice controls. Even if successful on this defense, the employer can, and usually will, be separately cited for violating the PEL. See, e.g., GAF Corp., 9 O.S.H.C. (BNA) 1451 (1981); Texaco, Inc., 8 O.S.H.C. (BNA) 1758 (1980).
AIA Reply Brief at 15 (emphasis in original).
It is difficult for us to understand exactly what the petitioners are driving at. Texaco, Inc. may provide a clue. There the employer was exonerated from non-compliance with required engineering and work practices on grounds of infeasibility, but was then cited for reliance on inadequate respirators. Perhaps AIA’s contention is that if OSHA cites employers for failure to provide respirators on those days where the exposure level uncontrollably fluctuates above the PEL, then all firms whose exposure level may ever fluctuate above the PEL must, to avoid citation, either institute continuous monitoring or provide satisfactory respirators 100% of the time. (In fact we doubt if firms facing a remote risk of citation would respond with measures whose cost vastly exceeded the expected value2 of the fines and other losses that citation might bring.) But OSHA has not demonstrated an intention to enforce its regulations in this manner; if that should prove to be its policy, firms may challenge it at the enforcement stage. See Steelworkers, 647 F.2d at 1270.
C. Cost-Effectiveness and “Reasonably Necessary” Standards
OSHA standards for toxic substances are governed not only by § 6(b)(5) but also by § 3(8) of the Act, defining occupational safety and health standards as ones which require “adoption or use of one or more practices, means, methods, operations, or processes, reasonably necessary or appropriate to provide safe or healthful employment and places of employment.” 29 U.S. C. § 652(8) (emphasis added). The “reasonably necessary” language was the basis in Benzene for the Court’s requirement of a finding of a “significant risk” as a precondition to OSHA’s imposition of more drastic standards. 448 U.S. at 642, 100 S.Ct. at 2864. In Cotton Dust, however, the Court made clear that, in light of the feasibility language in § 6(b)(5), “reasonably necessary” in § 3(8) did not require cost-benefit analysis. 452 U.S. at 513, 101 S.Ct. at 2492.
The AIA now invokes this language in support of a requirement of cost-effectiveness, i.e., a requirement that OSHA must, when confronted with two alternative routes to a given level of safety, choose the less expensive one. More specifically, AIA asserts that a 0.2 f/cc PEL is not “reasonably necessary” because a 0.5 f/cc PEL, combined with various supplemental methods, would produce better results at a lower cost. AIA Reply Brief at 24.
Cost-effectiveness certainly appears to have a powerful claim. As between saving lives expensively and thriftily, it would seem curmudgeonly and wasteful in the extreme (i.e., not “reasonably necessary”) to insist on the former. We need not, however, resolve the issue. The Secretary’s rule will not reduce the dangers below the level of risk that he reasonably deems “significant” (he finds such risk even at concentrations of 0.1 f/cc), and the AIA has not shown that their alternative package would do so. Therefore, except insofar as there might be inconsistencies between elements of the Secretary’s and AIA’s packages, it is his duty to keep adding measures so long as they afford benefit and are feasible, up to the point where he no longer finds significant risk. Accordingly, we will turn shortly to AIA’s proposed measures — originally advanced as partial substitutes but seemingly compatible as supplements — and the Secretary’s duty to consider the reasonableness of adopting them.
D. OSHA’s Ban on the Spraying of Asbestos Products
In the revised regulations, OSHA bans the spraying of products containing asbes*320tos. 29 C.F.R. § 1910.1001(f)(1)(vi), 1926.-58(g)(2)(iii). Its reason for adopting the ban is not protection of workers actually spraying the materials — it concedes that they may be adequately protected by respirators and protective clothing — but rather its perception of potential danger to bystanders. 51 Fed.Reg. 22,716/2. It supports this perception with the observation that both the EPA and the State of California have banned the spraying of asbestos-containing products. Id.
We sidestep the AIA’s argument that the ban did not emerge as a “logical outgrowth” of the originally noticed proposal, cf. South Terminal Corp. v. EPA, 504 F.2d 646, 659 (1st Cir.1974), and thereby violated the procedural requirements of notice-and-comment rulemaking, 5 U.S.C. § 553 (1982). The support for the ban plainly fails to meet the “substantial evidence” standard imposed by § 6(f) of the Act, 29 U.S.C. § 655(f). OSHA rests its case on (1) its failure to find anything in the 55,000-page record exonerating such practices and (2) bans imposed by the EPA and the State of California.
These supports do not hold up under examination. First, the AIA points to evidence in the record indicating that modem process of encapsulating spray-on asbestos products ensures that asbestos fibers are not released on application.3 OSHA says nothing to refute these claims. Second, both EPA and California specifically exempt encapsulated asbestos products from their spraying prohibitions, so their bans provide no support at all for OSHA’s far more sweeping one. The ban cannot stand.
E. Failure to Issue a Smoking-Control Regulation
Uncontested evidence shows that the lung cancer risk posed by asbestos exposure is ten times greater for smokers than for nonsmokers. Chronic Hazard Advisory Panel on Asbestos, Report to the U.S. Consumer Product Safety Commission at II-131 to II—132 (July 1983), J.A. at 293-94. OSHA itself noted in both its proposal and its final regulations that asbestos exposure “acts synergistically with cigarette smoking to multiply the risk of developing lung cancer.” 51 Fed.Reg. at 22,616/1; 49 Fed. Reg. at 14,118/2. In response to this relationship, OSHA included limited smoking control provisions in its final regulations. It banned smoking in all areas where the levels of airborne asbestos levels exceed the PEL, 29 C.F.R. §§ 1910.1001(e)(5), 1926.58(e)(5), and required employers to inform employees of the relationship between smoking and asbestos exposure. 29 C.F.R. §§ 1910.1001(j)(5)(iii)(B); 1926.-58(k)(3)(iii)(C). However, in response to AIA’s suggestion in a posthearing brief that OSHA adopt additional provisions to reduce smoking-related asbestos risks, such as prohibiting employers from hiring smokers and forcing employers to provide smoking cessation programs, OSHA stated that it had “made a determination, based on policy considerations,” not to expand this aspect of its standards. 51 Fed.Reg. at 22,700/2. Industry challenges OSHA’s failure to require more vigorous measures to reduce smoking-related asbestos risks.
The proposal raises the issue of what circumstances require OSHA to justify its refusal to adopt a proposal for tighter standards. Our starting point is § 6(b)(5) of the Act, which provides that OSHA “shall set the standard which most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impairment of health....” 29 U.S.C. § 655(b)(5) (emphasis added). In Public Citizen Health Research Group v. Tyson, 796 F.2d 1479 (D.C.Cir.1986), this court applied that language to a claim that OSHA should have promulgated a short-term emissions limita*321tion (STEL). The court held that “[i]f in fact a STEL would further reduce a significant health risk and is feasible to implement, then the OSH Act compels the agency to adopt it (barring alternative avenues to the same result).” Public Citizen, 796 F.2d at 1505 (emphasis in original). We take the principle to be applicable to any proposal.
Clearly, however, courts may not apply the principle in a way that would deprive OSHA of the “almost unlimited discretion” the statute affords it “to devise means to achieve the congressionally mandated goal.” Steelworkers, 647 F.2d at 1230. If the courts were to force the Secretary to produce substantial evidence from the record to support his choice in every case of a difference, no matter how minor, between the rules he adopts and the proposals of every party to the rulemaking, his discretion would be illusory and the administrative and judicial review processes interminable. The Act simply does not' require the agency to respond in detail to entire shopping lists of minute criticisms.
Moreover, proposals must be considered in their practical context. In some abstract sense a regulation can always be tightened. It will be rare that any given extra twist will push the cost up to the levels necessary for economic infeasibility; even a series of 20 or 50 proposals may not do so in the aggregate. And many imaginable tightenings will, standing alone, be technically feasible. But it does not follow from this that the statute mandates every proposed tightening not shown to be infeasible in either of these senses, just because it is theoretically capable of conferring a health benefit. OSHA’s enforcement and employers’ compliance duties both depend on ordinary human beings. Multiplication of requirements tax their abilities. (The present rules run some 57 pages in the Federal Register, 51 Fed.Reg. 22,733-90, not including the preamble.) At some point the confusion resulting from an extra burden is likely to outweigh any practical benefit.
Accordingly, we hold that a party challenging an OSHA standard must bear the burden of demonstrating that the variations it advocates will be feasible to implement and will provide more than a de minimis benefit for worker health. Cf. Steelworkers, 647 F.2d at 1207 (the Act requires that when making legislative-type policy decisions, the Secretary “present [his] reasons for rejecting significant contrary evidence and argument ”) (emphasis added); Alabama Power Co. v. Costle, 636 F.2d 323, 360-61 (D.C.Cir.1979). Naturally the force of the evidence and argument that OSHA must offer to defend its choice will vary with the force of the proponent’s evidence and argument. Cf. Stevenson v. Linens of the Week, 688 F.2d 93, 99 (D.C.Cir.1982) (“[T]he quantum and type of evidence required ... corresponds to the specificity of the claimant’s evidence and allegations.”) (internal quotation omitted).
OSHA here recognizes that asbestos exposure poses heightened risks to smokers. It makes no attempt to argue that the smoking-related provisions in its regulations will eliminate smoking-related asbestos hazards, or reduce them to insignificant levels. The AIA points to specific evidence in the record showing that smoking cessation programs would be both beneficial in terns of worker health and feasible. John M. Pinney, of the John F. Kennedy School of Government’s program on smoking behavior and policy, testified that such programs could achieve “long term success rates in excess of 30 percent.” J.A. 1078. He also listed the types of programs that he would recommend (e.g., smoking cessation activities on at least a quarterly basis; incentives to participate; making self-help materials available; and providing employees with periodic medical advice and health examinations to heighten individual motivation). Id. at 1078-79. Feasibility seems highly probable; Pinney estimated the average cost of a comprehensive program at about $200 per worker. Id. at 1079.
This evidence seems to us enough to trigger OSHA’s duty under Public Citizen to justify non-adoption of restrictions of the sort proposed. It clearly must defend any decision not to institute such regulations with more than unarticulated “policy considerations.” OSHA has not re*322vealed what these might be, but we note two points. First, OSHA regulations have not infrequently mandated special treatment of workers with higher risk propensities. For instance, it has required the removal of workers unable to wear respirators. See, e.g., 29 C.F.R. § 1910.1001(g)(3)(iv) (asbestos standard); 29 C.F.R. § 1910.1043 (cotton dust standard). Second, OSHA has even regulated matters of personal appearance. Thus, it has required that workers trim their beards to allow a good face-seal when using respirators. See 29 C.F.R. § 1910.134(e)(5)(i). Accordingly, glib references to possible intrusion on matters of personal taste will not support rejection of smoking control measures.
III. Union Challenges
A. Feasibility of a 0.1 f/cc PEL
Where a significant health risk exists, the Act requires that the agency adopt “the most stringent standard to protect against material health impairment, bounded only by technological and economic feasibility.” Cotton Dust, 452 U.S. at 503, 101 S.Ct. at 2487. As noted above, the agency acknowledges that a 0.2 f/cc PEL will not eliminate all significant risk, and has found that significant risk would continue to exist even at 0.1 f/cc. 51 Fed.Reg. at 22,648/1. Thus, OSHA’s decision to set the PEL at 0.2 f/cc was based solely on its view that industry could not attain a lower level. Id.
The unions assert that OSHA’s own findings demonstrate that a 0.1 f/cc level is achievable in operations affecting over 97 percent of the workers exposed to asbestos in general industry, and that its decision not to adopt a PEL of 0.1 f/cc is therefore unsupportable. Indeed, both union and industry petitioners point out that OSHA’s exposure data for the automotive brake and repair industry, which employs 541,998 workers, or 93 percent of all general industry workers exposed to airborne asbestos, 51 Fed.Reg. 22,654, Table 11, shows that average exposures in that industry are below 0.1 f/cc. 51 Fed.Reg. at 22,665, Table 22 (estimated current average exposure level is 0.06 f/cc in that industry). (OSHA also cites a study based on 47 observations between 1979 and 1984, finding a mean eight-hour time weighted average of 0.03 f/cc. Id. at 22,662/1.)
Rather than lower the PEL for the automotive repair sector, however, OSHA has merely recommended that employers “who wish to reduce their employees’ asbestos exposure during repair operations to levels below the new standard’s action level (0.1 f/cc)” employ certain control measures that are indisputably available (such as solvent mist or high efficiency particulate air (HEPA) vacuum systems). Appendix F, 29 C.F.R. § 1910.1001.
OSHA rested its decision not to mandate a 0.1 f/cc PEL on its finding that "most of the ‘best’ plant exposure data indicate that average exposures at many stations ... are in excess of 0.1 f/cc and cannot be reduced using current controls and practices,” 51 Fed.Reg. at 22,654/3. But the factual findings implicit in Table 11 obviously make this statement inapplicable to the automotive brake and clutch repair sector. In its brief here, OSHA appears to acknowledge the feasibility of 0.1 f/cc for that sector, but turns the fact around, arguing that there is no need to give weight, in a feasibility analysis, to “an operation that may be practically unaffected by [the Secretary's] feasibility resolution.” OSHA Brief at 79. But, quite apart from the argument being a mere post hoc rationalization of the agency’s counsel, see Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 419, 91 S.Ct. 814, 825, 28 L.Ed.2d 136 (1971), the argument turns feasibility analysis inside out: sectors where compliance with a given PEL is easy become arguments, in effect, for a less stringent PEL. The consequence is to deny OSHA protection to those in the automotive repair sector who work in shops with higher-than-average exposure levels. We see no basis for this in the Act.
At oral argument, OSHA counsel defended the Secretary’s decision not to set a lower PEL for this sector by invoking the Secretary’s discretion to decide what industries should be grouped together for regulatory purposes. Obviously, great num*323bers of minuscule industry subcategories may complicate enforcement, and we have elsewhere acknowledged OSHA’s discretion in the matter. See Hodgson, 499 F.2d at 481. But where the subcategory consists of 93 percent of affected workers, numbering over 500,000, this justification seems to completely disappear. Accordingly, we must remand the case for the Secretary to address the issue of disaggregating the general industry standard to afford workers the benefits of more stringent standards in areas where they are feasible.4
OSHA’s Tables 22 and 23 also indicate many smaller subcategories with average current concentrations well below 0.2 f/cc, several of them well below even 0.1 f/cc. 51 Fed.Reg. at 22,665. Moreover, the agency acknowledges that in many operations the 0.1 f/cc level “is already met on average by the best plants.” OSHA Brief at 79. It then invokes the problem of variability — heavily stressed by industry, as we have seen — as justifying the choice of 0.2 f/cc.
We think that in addressing this problem OSHA must provide a little more enlightenment. If its sole basis for selecting a uniform PEL of 0.2 f/cc is the administrative difficulty of selectively lower PELs, it should spell out the analysis. (We can imagine that the analysis would address such factors as the size of such sectors and the difficulties of drawing clear borders between them and ones where only 0.2 is attainable. Of course the expertise is OSHA’s, and it may well have other concerns.) To the extent that the administrative problems of excessive disaggregation are not the full answer, i.e., to the extent that it relies on such factors as excessive random fluctuations in exposure levels, it should make specific findings. As to such other factors, we note that: (1) OSHA’s own average figures presumably comprise the less advanced firms as well as the more advanced. Thus the figures in Tables 22 and 23 seem to suggest the attainability of the average levels reflected there. (2) If OSHA’s references to “the best plants” reflect a notion that their experience should be discounted, that is in plain contravention of the prevailing legal standard of feasibility. Steelworkers, 647 F.2d at 1264 (“OSHA can impose a standard which only the most technologically advanced plants in an industry have been able to achieve — even if only in some of their operations some of the time.”).5
B. Requirement of a STEL
All petitioners challenge the Secretary’s omission of a STEL, and OSHA has decided to reconsider its decision on this issue in light of this court’s decision in Public Citizen. OSHA reported to the court its belief that it could complete its reconsideration of this issue, on the existing record, by November 1987. The court hereby orders that this reconsideration be completed within 60 days of the issuance of the mandate in this case.
C. Miscellaneous BCTD Challenges
BCTD raises a host of objections to specific provisions of the construction standard. We review by the standard articulated in part II.E above (addressing AIA’s proposals of additional smoking controls).
1. Action level trigger issues. As noted above, the regulations are structured in *324such a way that some employer duties (e.g., exposure monitoring, employee training and information requirements) are triggered by exposures at or above the action level, while other duties (e.g., establishment of regulated areas, and provision of protective clothing and hygiene facilities) are not triggered until exposure levels reach the PEL. BCTD argues that it is feasible to trigger many of the latter duties at the action level rather than at the PEL, and that the Secretary therefore erred in not so providing. BCTD Brief at 41-42.
In the preamble, the agency explains that action levels “provide an appropriate cut-off point for many of the required compliance activities.” 51 Fed.Reg. at 22,680/1. It notes that this setting of compliance priorities “permits employers to concentrate their resources on those employees and workplace conditions with the potential for high asbestos exposures.” Id. at 22,707/2. BCTD in effect is arguing that OSHA’s rationale is unlawful because the agency has not justified its decision not to mandate greater duties at the action level on either infeasibility or laek-of-benefit grounds. We are skeptical of the agency’s asserted justification for similar reasons. However, as BCTD has failed to point to any evidence that triggering the regulated area requirements and other duties at the action level, rather than at the PEL, would result in a greater than de minimis incremental benefit, we do not find that the Secretary has abused his discretion.
2. Compliance Methods. BCTD argues that, like the general industry standard, the construction industry standard should allow employers to use only local exhaust ventilation or wet methods, and not any of the four other compliance methods approved by OSHA at 29 C.F.R. § 1926.58(g)(1)(i), to reduce exposure levels to the PEL. BCTD Brief at 42-44. BCTD points to evidence as to the effectiveness of its favored methods, but not as to any marked superiority of those methods over others OSHA chose to permit. Accordingly, we can find no evidence that the proposed restriction would have more than de minimis health benefits. There is obviously, moreover, some practical advantage in permitting employer choice and experimentation.
3. Respirator use and classification. BCTD contends that when exposure levels exceed the PEL and an employer is thereby forced to provide supplemental respirator protection, the regulations should mandate the use of the most protective respirator feasible rather than allowing the use of any respirator that can protect employees to the degree that they would be protected by exposure at 0.2 f/cc. BCTD Brief at 46-49. This would be likely almost always to require use of a “supplied-air” respirator for operations where the 0.2 PEL could not be attained.
OSHA noted in the rulemaking that the supplied-air respirator had hazards of its own—specifically tripping over the hose supplying the air. 51 Fed.Reg. at 22,719/2. We regard this sort of judgment as within OSHA’s discretion in the absence of evidence supporting the view that the incremental asbestos safety gains plainly exceed the incremental non-asbestos hazards.
One aspect of OSHA’s approach, however, fails to meet the basic standard of reasoned decision-making. The agency sets respirator requirements in terms of the asbestos level attained by the work practice and engineering requirements, so that the combination of measures will limit each worker’s exposure to the PEL. But OSHA’s statutory mandate is not framed as requiring it to ensure that all employees are protected at the level of its chosen PEL. It requires it to ensure that employers use all feasible methods to eliminate significant risk. Where the PEL is conceded to leave a significant health risk, as is the case here, we wonder why OSHA has concluded that protection at the PEL level should be regarded as enough. Our hunch is that OSHA’s policy is based on an implicit determination that universal enforcement of a 0.2 f/cc safety level is, for administrative efficiency reasons, the optimal enforcement strategy. But, as we noted above in part III.A, OSHA must explicitly define *325and justify such policies. As it has not done so, we include the respirator policy among the issues to be reconsidered on remand.
BCTD also points to evidence that the regulations credit various respirators with more effectiveness than the record supports. BCTD Brief at 49-50. OSHA does not dispute BCTD’s argument on the merits. It points out, however, that it is currently undertaking a comprehensive review of its general respirator standard, see 51 Fed.Reg. 38,593 (Oct. 27, 1986), and notes that upon the completion of this review it will make conforming changes in the asbestos standard where appropriate. OSHA Brief at 85. As the agency is already reevaluating the standard at issue, we can do no more than order that the agency complete such reconsideration and integrate its results with these regulations without undue delay.
5. “Initial monitoring" exemptions. The regulations provide two exceptions to their general requirement that all covered employers conduct initial asbestos exposure level monitoring. First, an employer can avoid initial monitoring if it can show “by means of objective data ... that the product or material containing asbestos, ... cannot release airborne fibers in concentrations exceeding the action level.” 29 C.F.R. § 1926.58(f)(2)(ii). BCTD objects that OSHA has not spelled out clearly enough what qualifies as “objective data.” BCTD Brief at 56-58. We think such drafting choices are plainly within OSHA’s discretion in the absence of a showing that a different approach would achieve more than de minimis health benefits.
Second, the regulations allow an employer to use prior monitoring data to meet the regulation’s initial monitoring requirements where the monitoring occurred under conditions “closely resembling” those currently prevailing. 29 C.F.R. § 1926.58(f)(2)(iii). BCTD again argues that the rule should be drafted with greater specificity. BCTD Brief at 55-56. Again, on this record the choice is up to OSHA.
6. Periodic monitoring, (a) The regulations define a “regulated area” as “an area established by the employer to demarcate areas where airborne [asbestos concentrations] exceed or can reasonably be expected to exceed the [PEL],” 29 C.F.R. § 1926.58(b), and require employers to conduct daily “breathing zone” monitoring “that is representative of the exposure of each employee who is assigned to work within a regulated area.” Id. § 1926.58(f)(3). BCTD argues that while these breathing zone samples accurately determine employee exposure within the regulated areas, periodic environmental monitoring at the boundaries of regulated areas is necessary to define the areas’ boundaries and thus to protect employees working outside of them. BCTD Brief at 59-60.
OSHA explained in the preamble its belief that the concerns expressed by BCTD are adequately addressed by the regulations’ general monitoring provisions, §§ 1926.58(f)(1)(i) & (ii), which require breathing zone samples that are representative of the exposure of each employee. As employees outside of the regulated areas are covered by these general monitoring requirements, OSHA asserts that environmental monitoring is unnecessary. 51 Fed.Reg. at 22,714/2. BCTD points to nothing in the record undermining that judgment.
(b) The regulations allow employers to dispense with periodic monitoring within a regulated area if all employees working within that area are 'equipped with supplied-air respirators. 29 C.F.R. § 1926.58(f)(3). This monitoring exemption is explained by OSHA’s belief that “proper use of these respirators will assure that employees are not inhaling asbestos fibers above the PEL, regardless of the levels of airborne asbestos outside their breathing zone.” OSHA Brief at 89. BCTD argues that the possibility that an employee may have a defective respirator ought to preclude the Secretary from approving this exception. Given the fact that the standard mandates respirator maintenance and fit-testing, 29 C.F.R. §§ 1926.58(h)(3) & (4), the mere imperfection of respirators does not support an inference that eliminating *326this monitoring exception would result in more than a de minimis health gain.
(c) The regulations also allow a construction industry employer to terminate periodic monitoring if the results of monitoring demonstrate that exposures are below the action level. 29 C.F.R. § 1926.58(f)(4). Unlike the equivalent provision for general industry, however, the construction industry standard includes no requirement for resumption of monitoring when a change in workplace conditions may result in exposures above the action level. Compare 29 C.F.R. § 1910.1001(d)(5). OSHA replies only that this sort of resumption requirement is implicit in the current standard. OSHA Brief at 90. In view of the risk that an employer might infer the opposite from the contrast in language, we believe OSHA should clarify the regulation to conform to the intent expressed in its brief.
7. Medical program requirements. (a) The regulations require medical surveillance for employees subject to certain minimum exposure levels. In general industry, the trigger is any work at or above the action level. 29 C.F.R. § 1910.1001(i)(1)(i). In the construction industry, where worker turnover is often extremely high and much of the workforce is transient, see 51 Fed.Reg. at 22,727/3-28/1, the requirement applies to any worker who has been required to wear a negative pressure respirator or who has been exposed at or above the action level for 30 or more days per year. 29 C.F.R. § 1926.58(m)(1)(i). (The Secretary construes the standard as requiring aggregation of days with different employers for purposes of meeting the 30-day trigger. See OSHA Brief at 91 n. 40.) BCTD would expand the surveillance to all construction workers who are exposed to any airborne asbestos, regardless of levels or duration of exposure, thus covering even an employee exposed at the action level for one day. BCTD Brief at 60-62. Neither the record nor common sense suggests that the proposed expansion of medical surveillance would generate material gains in worker health.
(b) Because of the high turnover rate and the degree of variation between construction workplaces, OSHA concluded that “annual [chest] x-rays may not always be necessary,” 51 Fed.Reg. 22,729/2, and determined to leave the decision of whether a physician should administer a chest x-ray as part of the medical examination up to the discretion of the individual physician. BCTD disagrees and argues that OSHA should require mandatory annual chest x-rays. BCTD Brief at 62. It cites in support simply the fact that most witnesses recommend this approach. Conclusory statements of this sort are not enough.
(c) BCTD also asserts that the medical questionnaire that is required to be administered by a physician as part of the medical surveillance program, 29 C.F.R. § 1926.58, Appendix D, should include more questions that delve specifically into the nature and amount of the workers’ asbestos exposure. BCTD Brief at 63. The medical questionnaire adopted was in fact the only medical questionnaire proposed by anyone. See 51 Fed.Reg. at 22,729/3. BCTD’s opportunity to critique the questionnaire has passed.
(d) Finally, BCTD argues that although an examining physician ought to be directed to advise the employer about limitations in an employee’s job duties or use of respirators, the regulations should not permit the physician to divulge specific medical findings or diagnoses. BCTD Brief at 63-64. We made clear in Hodgson that the interest in guiding employers in their task of fitting specific work with specific workers justified medical surveillance methods that revealed “the results” of physical examinations to employers. 499 F.2d at 485. We believe this provision well within Hodgson’s view of OSHA’s authority to weigh the concerns of privacy against those of health. Although OSHA must be sensitive to the concern for confidentiality, as Hodgson indicated, we believe that OSHA’s choice here lay well within that decision’s view of the agency’s authority to balance health against privacy.
8. Warning and labeling requirements. (a) The regulation exempts from *327otherwise applicable labeling requirements those products in which asbestos is bonded or modified in such a way that no “reasonably foreseeable” use or handling will lead to release of airborne concentrations above the action level. 29 C.F.R. § 1926.58(k)(2)(vi)(A). BCTD claims the exemption is not supported by substantial evidence, but adduces neither evidence nor argument in support of a more stringent rule.
(b) The standard also exempts products containing less than 0.1% asbestos from the labeling requirements. 29 C.F.R. § 1926.58(k)(2)(vi)(B). BCTD argues that this exception is unjustified. BCTD Brief at 66. OSHA found no record evidence that narrowing the exemption would generate health benefits, 51 Fed.Reg. at 22,725/1, and BCTD points to none.
(c) OSHA’s labelling rule requires only warnings in English. 51 Fed.Reg. at 22,724/3. BCTD contends that this is insufficiently protective of non-English-speaking employees and urges that the necessary signs and labels should be “bilingual in the languages that predominate in the workforce area.” BCTD Brief at 64. OSHA itself plainly regards warnings and labels as a necessary supplement to its other controls. It seems obvious that a warning sign or label in English will only rarely warn or educate non-English-speaking workers. The number of such workers in the construction industry is significant. (E.g., there were 64,000 Hispanic workers in the construction industry in the second quarter of 1984, according to the Bureau of Labor Statistics. J.A. at 1422.) As we noted earlier, § 6(b)(5) directs the agency to provide for all workers.
OSHA offers two rationales for refusing to require bilingual warnings or universal symbols. First, it maintains that such a requirement would impose “an unduly stringent requirement on those construction employers whose work force is comprised solely of English-speaking persons.” 51 Fed.Reg. 22,724/3. OSHA’s refusal to impose a pointless burden on such employers is entirely legitimate. But the agency is surely capable of designing a rule covering only employers with a significant number of non-English speakers. Second, the agency points out that “information overload” will reduce a warning’s effectiveness. OSHA Brief at 94, citing 48 Fed.Reg. 53,-301 (1983). This is clearly true. Cf. part II.E. But neither common sense nor anything in the record suggests that a bilingual warning would confuse English speakers; the gain for others seems obvious.
We recognize that BCTD’s claim, suggesting use of the “languages that predominate in the workplace,” has a potential breadth that could readily lead to requirements that would be infeasible as a practical matter or would yield only de minimis benefits at the margin. But OSHA is free to prevent such excesses by limiting any requirement to instances where the number of employees affected is substantial. Accordingly, we think the issue deserves more consideration on remand.
9. Reporting requirements, (a) BCTD maintains that OSHA should require all construction industry employers to file reports with it prior to engaging in any asbestos projects. BCTD Brief at 67-68. It argues that the information generated by such reports would enable the agency to more efficiently enforce the regulations, BCTD Reply Brief at 32-33, and cites evidence in the record that requiring such reports would lead to significant improvements in employer compliance and consequential decreases in worker risk. Hattis & DiMauro, “Health Benefits and Costs of Supplementary Measures to Improve Compliance With Workplace Exposure Limits for Asbestos in the Construction Industry,” J.A. at 1093, 1098-1100. It also points out that in OSHA’s acrylonitrile and inorganic arsenic exposure standards, the agency does require employers to report the address of their workplace, the number of employees working within the regulated area, and a description of each operation creating employee exposures. 29 C.F.R. §§ 1910.1045(d), 1910.1018(d).
In response, OSHA argues that “the manner and method of selecting work-sites for inspection falls exclusively within *328[the agency’s] discretion,” and that the agency must to some degree rely upon self-policing by employers. OSHA Brief at 66-68. We do not see how these general principles are relevant to the issue at hand. Agency discretion in selecting targets for inspection is quite different from agency evaluation of a proposed requirement that would arguably generate better information for making these selections. And if the reports suggested by the petitioners would significantly improve employer compliance, the fact that the agency must largely rely on self-policing would not justify their rejection. As the record contains uncontradicted (and unanalyzed) evidence of non-de minimis benefits, and as OSHA has replied only with reasoning of marginal relevance, the proposal must be considered on remand.
(b) BCTD further argues that OSHA should require all employers contracting or subcontracting asbestos-related work to establish, maintain, and transfer to building owners written records of the presence and locations of any asbestos or asbestos products in any actual or prospective workplace. BCTD Brief at 68. It argues that such records would greatly facilitate identification and prevention of asbestos hazards.
OSHA did not even mention this proposal in its preamble, although it offered some purely descriptive observations to the effect that the regulations imposed no duties on building owners as such. 51 Fed.Reg. at 22,678/1. In its brief here it argues that building owners “are outside the domain of the OSH Act.” OSHA Brief at 96. This would not seem to be true insofar as future building owners are employers. Perhaps OSHA interprets the Act as conferring upon it jurisdiction over employers only as they directly control the workplaces of their employees. Some provisions of the Act may support that view. The statement of purposes, referring to Congress’s effort to assure safe working conditions “by encouraging employers and employees in their efforts to reduce the number of ... hazards at their places of employment ...,” 29 U.S.C. § 651(b)(1), may point that way. The definition of the standards that OSHA is to adopt seems somewhat broader: the standards are to “require[] conditions, or the adoption or use of one or more practices, means, methods, operations, or processes reasonably necessary or appropriate to provide safe or healthful employments and places of employment.” Id. § 652(8).
We think it inappropriate for us to rule on the point without OSHA having considered it explicitly. Cf. Phillips Petroleum Co. v. FERC, 792 F.2d 1165, 1170, 1172 (D.C.Cir.1986) (where agency erroneously believed Supreme Court decision required particular interpretation of ambiguous statute, court remands to agency for it to reach its own judgment on the issue). Accordingly, we remand to OSHA for that purpose.
10. Scope of “competent person” requirement. “Competent person” is a term of art, defined as “one who is capable of identifying existing asbestos, ... hazards in the workplace, and who has the authority to take prompt corrective measures to eliminate them....”29 C.F.R. § 1926.58(b). The regulations require employers to designate a competent person to supervise removal, demolition and renovation operations for which the employer is required to establish a negative-pressure enclosure. Id. at § 1926.58(e)(6)(ii). Along with their other duties, competent persons are directed to: (1) set up the negative-pressure enclosure and ensure its integrity, (2) supervise all employee exposure monitoring required by the standard, (3) ensure that employees are trained in the use of engineering controls, work practices, and personal protective equipment, and (4) ensure that engineering controls are functioning properly. Id. at §§ 1926.58(e)(6)(A), (B), (D), (F) and (H).
(a) BCTD argues that “employers engaged in any kind of asbestos-related construction work” should be required to designate competent persons to oversee asbestos-related safety measures. BCTD Brief at 54 (emphasis in original). Thus, it would expand the requirement beyond areas where the regulations require a negative-pressure enclosure.
*329BCTD points to testimony in the record supporting an expansion of the competent person requirement. For instance, Dr. Morton Corn, a professor in the Department of Environmental Health Science at Johns Hopkins University’s School of Hygiene and Public Health, testified that OSHA should require the presence of an industrial hygienist with knowledge of the operation and authority to shut down the work site. J.A. at 1707-08. John A. Pritchard testified that the regulations should require the presence of an expert in respiratory protection on site to ensure effective respirator fit and use. J.A. at 1718-19. While we recognize that the testimony relied on by the petitioners is somewhat conclusory, their recommendations accord with common sense. OSHA has pointed to nothing in the record that contradicts this testimony, but merely points out its belief that “[t]he operations covered by the ‘competent person’ requirement ... comprise the activities where exposure levels are the highest ...,” and argues that “[implementing this requirement for the most dangerous operations is neither arbitrary nor irrational.” OSHA Brief at 87. While we may take this as asserting the sound point that OSHA must draw the line somewhere, it hardly helps explain why it drew the line where it did. We recognize that it will rarely be possible to offer a compelling reason for the exact placement of a line of this sort, and we do not ask for such a reason. But we think something less conclusory is due. We note that the proposal aims directly at enhancing the human capacity for compliance, for the difference between a “competent person” and others is essentially one of training. 29 C.F.R. § 1926.58(e)(6)(iii). Thus, the proposal is not as vulnerable as many others are to the observation in II.E that at some point, given the frailty of human beings, the multiplication of compliance duties becomes counterproductive. The Secretary must therefore either expand the applicability of the competent person requirement or provide a more persuasive explanation of his refusal to do so.6
(b) BCTD also contends that the standard should explicitly provide that the “competent person” has the authority to suspend work and remove affected employees to a safe area when hazardous conditions exist. BCTD Brief at 54. OSHA notes that the competent person is defined as one with “the authority to take prompt corrective measures to eliminate” workplace hazards. See 29 C.F.R. § 1926.58(b). BCTD has pointed to nothing suggesting that its proposed change in the definition would have more than a de minimis effect, and the choice of language is therefore within the Secretary’s discretion.
11. Exemption from removal, demolition and renovation requirements. The regulations exempt “small-scale, short-duration operations” from the competent person and negative-pressure enclosure requirements applicable to employers engaged in removal, demolition and renovation operations. 29 C.F.R. § 1926.58(e)(6)(iv). The regulations do not define “small-scale, short-duration operations” explicitly, but rather attempt to do so by example, noting that the term is meant to include operations “such as pipe repair, valve replacement, installing electrical conduits, installing or removing drywall, roofing, and other general building maintenance or renovation.” Id. BCTD asserts that OSHA has not drawn the parameters of the exemption with enough specificity. Indeed, the exception as now worded seems to erase the rule. OSHA responds that explanatory language in the preamble adequately narrows the exception. There OSHA stated that the exemption was meant to apply only to “those work operations where it is impractical to construct a negative-pressure enclosure because of the configuration of the work environment.” 51 Fed.Reg. at 22,711/2. In so far as the exemption applies to the negative-pressure enclosure requirement, this language seems marginally clear enough and makes apparent sense, but on remand the Secretary should consider in-*330eluding this clarification in the regulations themselves. (In so far as the exemption applies to the competent person requirement, we can discern no logic whatever in OSHA’s choice of language, with or without the preamble’s gloss. See note 6 supra.)
IV. Conclusion
Accordingly, we remand the case to the Secretary for further proceedings consistent with this opinion.
So ordered.
. 51 Fed.Reg. 22,653/3. OSHA vigorously disputes, however, industry's assertions as to the magnitude of the random fluctuations problem, contending that most of the observed fluctuations in asbestos levels are in fact controllable through proper maintenance and employee training and supervision. See OSHA Brief at 63-65. But see AIA Reply Brief at 13 n. 5.
. I.e., value multiplied by likelihood of occurrence.
. See, e.g., RTI, Asbestos Dust: Technological Feasibility Assessment and Economic Impact Analysis of the Proposed Federal Occupational Standard Part I at 11-15 (September 1978), J.A. at 187 (1978 feasibility study finding that the use of encapsulated asbestos products produced fiber counts below detectable limits); Versar, Inc., Draft Final Report: Exposure Assessment for Asbestos at 253 (January 9, 1984), J.A. at 431 (report for EPA commenting that "[s]ince the banning of water-based products, exposures due to application of coatings are unlikely; petroleum-based compounds do not release fibers on application.”).
. As the Secretary set the action level at one half the PEL, on remand we instruct the Secretary to consider an appropriate modification of the action levels where, in response to this opinion, he changes the PELs.
. AIA appears to argue that Steelworkers’s concept of feasibility — that OSHA may infer feasibility from the "reasonable possibility" that a typical firm will be able to comply — is applicable only where the agency is engaged in “technology-forcing,” i.e., where the technology of compliance is not currently available. AIA Brief at 13-14. This seems manifestly backwards. The inference is surely more supportable where the technology of compliance is available.
No party here argues that there is a difference between the standard of feasibility for what the Secretary must prescribe (at issue here) as opposed to the standard for what he may prescribe (at issue in Steelworkers). Accordingly, we do not address the issue. We note that as a practical matter the deference we owe the agency is likely to create a practical gap between the minimum and maximum levels of stringency.
. We observe that neither the record nor common sense provides clear support for the "small-scale short term operations" exception from the competent person requirement for removal. demolition and renovation operations, 29 C.F.R. § 1926.58(e)(6)(iv), as OSHA currently appears to define that exception. See our general discussion of this exception in part III.C.ll infra.