(dissenting). In upholding the action of the commissioner in this case, the majority reaffirms the applicability of the rational basis test to administrative action.1 As a result, if under any conceivable set of circumstances, no matter how remote, the necessity for the regulation is fairly debatable, the “inescapable” conclusion is that the regulation must be upheld. The majority reasons that any other result would violate the principle of separation of powers, and would fail to give due deference to the expertise of administrative bodies. Ante at 553-554.
It is my view that the rational basis test as traditionally applied in this Commonwealth rests upon the twin infirmities of misapplied precedent and illogic. It therefore should no longer be used as the litmus for judicial scrutiny of administrative regulations. That the regulation in this case passes scrutiny under the rational basis test illustrates beyond dispute the ineffectiveness of that test as a limitation on the abuse of power by administrative bodies. Therefore, even assuming that the majority has correctly characterized the commissioner’s action *558as a regulation rather than a disguised adjudication,2 I would not apply the rational basis test in determining the legality of that action. In my view, the rational basis test, as applied to administrative action, results not in the proper separation of powers, but in the abdication of meaningful judicial review. See American Grain Prods. Processing Inst. v. Department of Pub. Health, 392 Mass. 309, 331 (1984) (Lynch, J., dissenting). Thus, I respectfully dissent from Part IIB of the majority opinion. Because of the view that I take of this case, I would not reach the questions whether the procedure required by G. L. c. 111, § 31, was followed and whether the action of the commissioner was in fact adjudication within the meaning of G. L. c. 30A, §§ 1 (1) & 14.
1. Flaws of the rational basis test. The rational basis test arose out of a recognition of the principle that “acts, passed by the legislature, according to the forms prescribed in the Constitution, are founded in authority given by the people to that department.” Portland Bank v. Apthorp, 12 Mass. 252, 253 (1815). Popularly elected representatives are accountable to the electorate, which provides a check on the arbitrary exercise of legislative power. For that reason, otherwise constitutional legislative action is not normally subject to intense judicial scrutiny. Indeed, it is our duty “to give effect to the will of the people as expressed in the statute by their representative body. It is in this way . . . that the doctrine of separation of powers is given meaning.” Shell Oil Co. v. Revere, 383 Mass. 682, 687 (1981), quoting Commonwealth v. Leis, 355 Mass. 189, 202 (1969) (Kirk, J., concurring). See Blue Hills Cemetery, Inc. v. Board of Registration in Embalming & Funeral Directing, 379 Mass. 368, 372 (1979).
*559However, the infirmities of the test flow from its application to judicial review of administrative action. I cannot agree with the principle that a “regulation stands on the same footing as would a statute, ordinance, or by-law.” Druzik v. Board of Health of Haverhill, 324 Mass. 129, 138 (1949). That logical leap, first taken by this court in Druzik, has been accepted and repeated as an article of faith, though its historical and philosophical justifications have never been probed. Indeed, the only support for the extension of this principle that was cited by the Druzik court was Pacific States Box & Basket Co. v. White, 296 U.S. 176, 186 (1935),3 a case that is clearly no longer good law. See, e.g., 1 K.C. Davis, Administrative Law § 6.13, at 508; § 6.16, at 524 (2d ed. 1978).4 While it may be easy to apply the same standard of review to legislative as to administrative action, to do so ignores material differences between the two.
Simply stated, “an agency is not a legislature.” State Farm Mut. Auto. Ins. Co. v. Department of Transp., 680 F.2d 206, 221 (D.C. Cir. 1982), vacated on other grounds, Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29 (1983). It is therefore “incorrect” to say that regulations have the “same legal force as do statutes.” 1 F. Cooper, State Administrative Law 264 (1965). See 1 K.C. Davis, supra § 6.13, at 509. Instead, we should recognize that administrative bodies “have become a veritable fourth branch of the Government, which has deranged our three-branch legal theories much as the concept of a fourth dimension unsettles our three-dimensional thinking. . . . [They] have been called quasi-legislative, quasi-executive or quasi-judicial, as the *560occasion required, in order to validate their functions within the separation-of-powers scheme of the Constitution. The mere retreat to the qualifying ‘quasi’ is implicit with confession that all recognized classifications have broken down.” FTC v. Ruberoid Co., 343 U.S. 470, 487-488 (1952) (Jackson, J.,
. dissenting). It is, however, not an impossible task to recognize the similarity of regulations to legislative enactments, while nevertheless taking due account of their differences. Blithely papering over these differences, as the rational basis test does, simply admits defeat in attempting to deal with a difficult problem.
It will not do to say only that the Legislature, comprised of directly elected representatives, has delegated its power to administrative bodies and theoretically provides some check on administrative abuse. First, administrative bodies are still one step removed from the electorate, the source of legislative power and the most potent check on its possible abuse. See American Grain Prods. Processing Inst. v. Department of Pub. Health, supra at 333 (Lynch, J., dissenting) (“administrative officials are only tangentially responsive to the will of the people”). Moreover, “policymaking by a small group of political appointees may be objectionally nonrepresentative,” especially when no hearing and no formal opportunity to comment on proposed regulations takes place. L. Jaffe, Judicial Control of Administrative Action 566 (1965). Second, the plethora of State and local administrative bodies and the competing demands of other pressing legislation prevent effective legislative oversight. The Legislature does not have the opportunity “to examine more than an infinitesimal proportion of . . . regulations.” DeLong, Informal Rulemaking and the Integration of Law and Policy, 65 Va. L. Rev. 257, 281 (1979). See Bruff, Legislative Formality, Administrative Rationality, 63 Tex. L. Rev. 207, 231-232 (1984). For these reasons, other courts have recognized that “agencies do not have quite the prerogative of obscurantism reserved to legislatures.” UnitedStates v. Nova Scotia Food Prods. Corp., 568 F.2d 240, 252 (2d Cir. 1977). Indeed, the Supreme Court has recently stated that it does “not view as equivalent the presumption of constitutionality afforded *561legislation drafted by Congress and the presumption of regularity afforded an agency in fulfilling its statutory mandate.” Motor Vehicle Mfrs. Ass’n of the U.S., Inc. v. State Farm Mut. Auto. Ins. Co., supra at 43 n.9. See FCC v. RCA Communications, Inc., 346 U.S. 86, 90 (1953) (“Congress did not purport to transfer its legislative power to the unbounded discretion of the regulatory body”).
With the explosion of regulatory activity in the past twenty years, administrative bodies have become more independent, exercising more power at the same time that courts, at least in this Commonwealth, continued to adhere to only the most casual form of review. This situation has led to a perception that “agencies are out of control.” Breyer, Reforming Regulation, 59 Tul. L. Rev. 4, 6 (1984). It has added fuel to “the growing popular disenchantment with the pervasive bureaucratization of economic and social life.” Gellhom & Robinson, Rulemaking “Due Process”: An Inconclusive Dialogue, 48 U. Chi. L. Rev. 201, 260 (1981). These circumstances have called into doubt the “comfortable assumption that political remedies . . . may constitute an effective check” on administrative power, at the same time that it has become evident that administrative bodies are political, not neutral and disinterested, parties. DeLong, supra at 278, 281.
Despite these formidable practical and theoretical problems generated by assigning equivalence to statutes and regulations, this court has maintained its adherence to Druzik. Generally, the court has stated this theory of equivalence (upon which the rational basis test wholly rests) as an accepted fact, as though through simple repetition it could become so. See, e.g., American Grain Prods. Processing Inst. v. Department of Pub. Health, supra at 329; White Dove, Inc. v. Director of the Div. of Marine Fisheries, 380 Mass. 471, 477 (1980); Greenleaf Fin. Co. v. Small Loans Regulatory Bd., 377 Mass. 282, 293-294 (1979); Colella v. State Racing Comm’n, 360 Mass. 152, 155-156 (1971). Alternatively, the court has relied on justifications that may support some degree of judicial deference, but in no way command the application of the rational basis test. For example, the court has justified the rational basis *562test on the ground that the court “may not substitute its judgment for that of the Legislature.” Purity Supreme, Inc. v. Attorney Gen., 380 Mass. 762, 776 (1980). See Borden, Inc. v. Commissioner of Pub. Health, 388 Mass. 707, 723, cert. denied sub nom. Formaldehyde Inst., Inc. v. Frechette, 464 U.S. 936 (1983). Similarly, the court has based its use of the rational basis test on the avoidance of “an unhealthy tendency to substitute the court for the agency as policymaker.” Cambridge Elec. Light Co. v. Department of Pub. Utils., 363 Mass. 474, 491 (1973). See ante at 553; American Family Life Assurance Co. v. Commissioner of Ins., 388 Mass. 468, 477-478, cert. denied sub nom. American Family Life Assurance Co. v. Hiam, 464 U.S. 850 (1983); Grocery Mfrs. of Am., Inc. v. Department of Pub. Health, 379 Mass. 70, 79-80 (1979); Consolidated Cigar Corp. v. Department of Pub. Health, 372 Mass. 844, 853 (1977). As the previous discussion has illustrated, however, the idea that the Legislature has exercised any “judgment” is purely a legal fiction. Moreover, simply requiring that the administrative action be supported by “such evidence as a reasonable mind might accept as adequate to support a conclusion” (the “substantial evidence” test), does not require this court to engage in judicial legislation or to substitute its views for those of the administrative body. This is obvious from a review of our cases applying the substantial evidence test, where this court has held that it will not substitute its view of the facts for that of the administrative body, Arthurs v. Board of Registration in Medicine, 383 Mass. 299, 304 (1981), nor decide what weight should be placed on particular facts, General Elec. Co. v. Assessors of Lynn, 393 Mass. 591, 602 n.5 (1984), nor set aside a choice “between two fairly conflicting views, even though the court would justifiably have made a different choice.” Southern Worcester County Regional Vocational School Dist. v. Labor Relations Comm’n, 386 Mass. 414, 420 (1982). We have noted that our review, even under the substantial evidence test, is a “limited” one, calling for “deference” to the administrative body’s particular expertise. Workers’ Compensation Rating & Inspection Bureau v. Commissioner of Ins., 391 Mass. 238, 245-246 (1984). Indeed, *563under the substantial evidence test we have explicitly held that “we may not substitute our judgment for that of the [administrative body].” 1001 Plays, Inc. v. Mayor of Boston, 387 Mass. 879, 885 (1983). In light of these facts, it is evident that the only justification for the rational basis test is judicial precedent based on faulty premises.
2. A proposed standard. In developing a more appropriate standard of review of regulatory action, it is instructive to consider how the Federal courts have approached the same problem. Those courts, like the majority in this case, apply an “arbitrary or capricious" standard. However, Federal application of the arbitrary or capricious test bears little similarity to the test applied by the majority. From the lenient standard of Pacific States Box, which the majority still follows, Federal courts have evolved a much more substantial inquiry, subjecting regulatory action to “a thorough, probing, in-depth review.” Citizens to Preserve Overton Park, Inc. v. Volpe, supra at 415.5 See Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., supra at 43.
This so called “hard look” standard of review “compares the agency’s stated rationale for a decision with supporting or opposing data and policy views gathered by the agency as the ‘administrative record’ for judicial review. The court identifies the agency’s value choices and checks their consistency with the factual basis asserted for them, the agency’s other present or past policies, and the governing statute.” Bruff, supra at 238. Similarly, Federal courts will “not supply a reasoned basis for the agency’s action that the agency itself has not given” and actually relied on. Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto Ins. Co., supra at 43 & 50. This is in effect a rejection of the rational basis test as applied in this Commonwealth. That test, unlike the Federal test, validates any regulation so long as the imagination of some administrator, lawyer, or judge is fertile enough to envision a conceivable, *564though nonexistent, scenario, under which the regulation might possibly be labelled “rational.”
The elements of the Federal test “do not significantly differ for purposes of judicial review of rulemaking” from the “substantial evidence” test with which this court already has developed familiarity. 1 K.C. Davis, Administrative Law § 6.6, at 468 (2d ed. 1978). See DeLong, supra at 287-288. See also Associated Indus, of N.Y. State, Inc. v. United States Dep’t of Labor, 487 F.2d 342, 349-350 (2d Cir. 1973). Therefore, I would apply the substantial evidence test to review regulatory action.6 It is neither radical nor unreasonable to require merely that administrative bodies base their regulations on “such evidence as a reasonable mind might accept as adequate to support a conclusion” (G. L. c. 30A, § 1 [6]), “including any evidence that detracts” from it. Craven v. State Ethics Comm’ n, 390 Mass. 191, 201 (1983), quoting 1001 Plays, Inc. v. Mayor of Boston, supra at 885. Indeed, this court may have already begun to move in the direction of requiring something more from administrative bodies than “any conceivable basis.” See American Grain Prods. Processing Inst. v. Department of Pub. Health, supra at 325 (finding a “substantial basis” for regulation); Borden, Inc. v. Commissioner of Pub. Health, supra at 725 (not relying on a rational basis because, after review of the record, the court found actual evidence warranting the regulation).
3. The action in this case. It is unmistakably clear that the commissioner’s action is not supported by substantial evi*565dence.7 His action is purportedly based on scientific studies conducted for the commissioner. But these studies were undertaken only after the commissioner had originally issued the chemical ban. It is not surprising that it was reissued almost verbatim after the studies were completed, because when “the decision is announced prior to [its justification]. . . the process is no longer one of investigation. There is an overwhelming institutional bias in favor of justifying the result in any way possible.” Independent U.S. Tanker Owners Comm. v. Lewis, 690 F.2d 908, 920 (D.C.Cir. 1982).
Putting aside this consideration, the studies themselves also fail to provide substantial evidence for his action. First, these studies focused exclusively on ADL’s research to establish the threat to public health involved. There was no consideration of, or reliance on, any other realistic scenarios concerning the possible use of the banned chemicals in Cambridge by the small number of other foreseeable users. Second, the conclusions of those studies are flawed on their face. They do not consider ADL’s commitment to use significantly smaller amounts of the chemicals than the studies hypothesized. Thus, the studies are virtually useless in assessing even the “worst case” accident scenarios. When confronted with this fact, the commissioner has simply asserted that he did not have to consider ADL’s commitment, since there was no guaranty that other users would make the same commitment. But the studies did not examine the hazards that might be posed by other users; they addressed themselves solely to ADL. It is neither reasonable nor rational for the commissioner to rely on the studies to support his action, but to deflect any criticism of the results *566or methods chosen by shifting the focus to situations he has never considered.8
Nor was there sufficient information from which the commissioner could have made an informed judgment about the acceptability of the risk of harm from ADL’s research. In general, “[t]he concept of overall risk incorporates the significance of possible adverse consequences discounted by the improbability of their occurrence.” City of N.Y. v. United States Dep’t of Transp., 715 F.2d 732, 738 (2d Cir. 1983), cert. denied, 465 U.S. 1055 (1984). Quite apart from the commissioner’s faulty knowledge about possible adverse consequences, he had no reasonable idea of the possibility that they would occur. Neither study considered the probability of a harmful accident except in the most general terms. While I would not require mathematical precision in making this assessment, something more is required than unsupported statements that it is “very unlikely” or “unlikely but not impossible.” The quoted assessments offer the commissioner no support, since virtually anything could be considered by a reasonable person to be “not impossible. ”
The majority is correct in stating that we must give deference to the commissioner’s expertise. Ante at 553-554. But “[i]n judicial review, the court must evaluate the relevance and weight of expertness” (emphasis in original). L. Jaffe, supra at 579. Unless “the requirements for administrative action [are] strict and demanding, expertise, the strength of modem government, can become a monster which mies with no practical limits on its discretion” (emphasis in original). Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., supra at 48, quoting New York v. United States, 342 U.S. 882, 884 (1951) (Douglas, J., dissenting). It does no violence to the principles of deference to conclude that the commissioner lacked substantial evidence to bring his expertise into play in this situation.
4. Conclusion. The standard of review I have advanced would not unduly restrict role making, nor would it place this *567court in the position of determining the appropriateness of any particular rule. Instead, the court’s task would be simply to ensure that administrative bodies act in a manner that is both consistent with statutory policy and reasonably related to life in the real world, rather than some conjectured life in the fictional twilight zone of what is “conceivable.”
It is worth bearing in mind that “the knowledge of an administrative agency is rarely complete, an administrative agency is not ordinarily a representative body, its deliberations are not usually conducted in public, and its members are not subject to direct political controls in the same way as are legislators.” Bonfield, The Iowa Administrative Procedure Act: Background, Construction, Applicability, Public Access to Agency Law, the Rulemaking Process, 60 Iowa L. Rev. 731, 845 (1975). Oras one noted writer has said of independent agencies: “They have been founded on a basically undemocratic concept of the political process and have helped perpetuate naive notions about regulation, ... the virtues of group decision and the use of expertness.” M. Bernstein, Regulating Business By Independent Commission 294 (1955). Because of these considerations, we should abandon the principle that the court “should cast about to discover” any ground on which to uphold administrative action. Cotter v. Chelsea, 329 Mass. 314, 318 (1952). For “[w]hen a court declines to scrutinize (or even assumes the existence of) supporting facts, ignores the presence of alternatives, and requires nothing more than minimally rational explanation, almost any outcome can pass muster.” Garland, Deregulation and Judicial Review, 98 Harv. L. Rev. 507, 556 (1985). The result here is, therefore, “one more step toward agency nonaccountability and carte blanche.” American Grain Prods. Processing Inst. v. Department of Pub. Health, supra at 332 (Lynch, J., dissenting). At some point this process must stop.
The majority speaks of a “conceivable ground” necessary to validate the regulation, a test which is indistinguishable from the traditional rational basis test applied to legislative enactments. See, e.g., American Family Life Assurance Co. v. Commissioner of Ins., 388 Mass. 468, 477-478, cert. denied sub nom. American Family Life Assurance Co. v. Hiam, 464 U.S. 850 (1983).
Although the regulation is couched in general terms, ADL is the only present or likely user of the banned chemicals, and the scientific studies allegedly supporting the regulation focus exclusively on the risks associated with ADL’s research, and not on other problems that might arise if the chemicals were used in the future by some other business or individual. I also note that the regulation essentially renders wasted approximately $1 million spent by ADL to construct, with the knowledge and approval of the city of Cambridge, a special laboratory to conduct this research.
The Druzik court also cited Fieldcrest Dairies, Inc. v. Chicago, 122 F.2d 132, 135 (7th Cir. 1941), rev’d on other grounds, 316 U.S. 168 (1942), which itself relied on Pacific States Box, but only for the proposition that a municipal ordinance, not merely a regulation, stands on the same footing as a statute.
The Supreme Court has since interpreted the language in Pacific States Box relied on by the Druzik court to mean only that administrative actions are entitled to a presumption of regularity. See Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 415 (1971).
Although Overton Park involved administrative action that was neither strictly regulatory nor adjudicatory, the principles of that case have been interpreted broadly so as to “revamp the law of informal rulemaking.” DeLong, supra at 266.
Of course, the test would operate somewhat differently in regulatory cases than in review of adjudicatory action. For example, the “evidence” necessary to support the rule would be provided in court, rather than established before the administrative body and would include whatever information the regulation’s challenger and the administrative body choose to present in support of, or in opposition to, the rule. As in adjudicatory review, the initial burden of producing sufficient evidence to call into question the basis for a regulation would rest on the challenger. The administrative body would then have the usual choice of testing the challenger’s evidence by a motion to dismiss (and appealing an adverse result), or by presenting its own case, and holding the challenger to its ultimate burden of persuasion that the regulation is unsupported by substantial evidence. Cf. General Elec. Co. v. Assessors of Lynn, 393 Mass. 591, 599-600 (1984).
I note that G.L. c. 111, § 31, requires the commissioner’s regulations to be “reasonable” In Workers’ Compensation Rating & Inspection Bureau v. Commissioner of Ins., 391 Mass. 238, 245 (1984), we held that there is “no principled basis” for distinguishing between “substantial evidence,” defined as evidence “a reasonable mind might accept as adequate to support a conclusion,” and “reasonable support in the evidence.” Thus, even without adopting the new general standard I have put forth, the commissioner’s regulation should be reviewed under the substantial evidence test. To determine its validity under the rational basis test ignores the Legislature’s choice of a stricter standard.
It was also open to the commissioner to include in the regulation a restriction similar to the commitment made by ADL.