Opinion for the court filed by Circuit Judge J. SKELLY WRIGHT.
Concurring opinion filed by Senior Circuit Judge MacKINNON.
J. SKELLY WRIGHT, Circuit Judge:Almost 14 years ago, soon after passage of the National Environmental Policy Act (NEPA), 42 U.S.C. § 4321 et seq. (1982), this court faced the challenge of ensuring that the Act’s “important legislative purposes, heralded in the halls of Congress, [were] not lost or misdirected in the vast hallways of the federal bureaucracy.” Calvert Cliffs’ Coordinating Committee v. USAEC, 449 F.2d 1109, 1111 (D.C.Cir.1971). This case poses a no less formidable challenge: to ensure that the bold words and vigorous spirit of NEPA are not similarly lost or misdirected in the brisk frontiers of science.
For this appeal presents an important question at the dawn of the genetic engineering age: what is the appropriate level of environmental review required of the National Institutes of Health (NIH) before it approves the deliberate release of genetically engineered, recombinant-DNA-containing organisms into the open environment? More precisely, in the context of this case, the question is whether to affirm an injunction temporarily enjoining NIH from approving deliberate release experiments without a greater level of environ*125mental concern than the agency has shown thus far.
In September 1983 three environmental groups and two individuals filed suit against the federal officials responsible for genetic engineering decisions.1 Arguing that NIH had not complied with the requirements of NEPA, plaintiffs sought to enjoin a proposed NIH-approved experiment by University of California scientists that would represent the first deliberate release of genetically engineered organisms into the open environment. They also sought to enjoin NIH’s approval of any other deliberate release experiments. Plaintiffs later added Regents of the University of California as a defendant. On May 18, 1984 the District Court granted the requested relief and enjoined both the University of California experiment and NIH approval of all other deliberate release experiments.
We emphatically agree with the District Court’s conclusion that NIH has not yet displayed the rigorous attention to environmental concerns demanded by law. We therefore affirm the District Court’s injunction prohibiting the University of California deliberate release experiment until an appropriate environmental assessment is completed. We also share the District Court’s view that NIH should give greater consideration to the broad environmental issues attendant on deliberate release of organisms containing recombinant DNA, and to its own responsibility for approving these deliberate release experiments. We find, however, that the part of the injunction enjoining NIH from approving all other deliberate release experiments is, at this juncture, overly broad, and we therefore vacate the part of the injunction that prohibits NIH approval of those experiments. We wish to emphasize, however, that if NIH fails to give appropriate environmental consideration to any other experiment, as it has failed to do with the University of California experiment, injunctive relief would be clearly proper.
I. Background
This case arises against a backdrop of the National Environmental Policy Act, the emergence of genetic engineering, and federal attempts to regulate genetic engineering.
A. National Environmental Policy Act
On January 1,1970 the National Environmental Policy Act became law. Recognizing “the profound impact of man’s activity on the interrelation of all components of the natural environment,” 42 U.S.C. § 4331(a), Congress sought to “fulfill the responsibilities of each generation as trustee of the environment for succeeding generations,” id. § 4331(b)(1). The major “action-forcing” provision of NEPA is the requirement that “all agencies of the Federal-government” prepare a detailed environmental analysis for “major Federal actions significantly affecting the quality of the human environment.” Id. § 4332(C); S.Rep. No. 91-296, 91st Cong., 1st Sess. 19 (1969). Congress mandated that this detailed statement, long known as an Environmental Impact Statement (EIS), include such considerations as “the environmental impact of the proposed action,” “any adverse environmental effects which cannot be avoided should the proposal be implemented,” and “alternatives to the proposed action.” 42 U.S.C. § 4332(C).
Realizing that NEPA would be toothless if agencies could merely issue a conclusory statement that the action did not significantly affect the environment (and that therefore no EIS was required), the Council *126on Environmental Quality (CEQ), an entity created by NEPA, issued regulations establishing that, unless the major federal action falls within an agency-established “categorical exclusion,” 40 C.F.R. § 1508.4 (1983), the agency should support each finding of “no significant impact” with a “concise public document” called an “environmental assessment” (EA). Id. §§ 1501.4(a)-(b), 1508.9. The environmental assessment must “[bjriefly provide sufficient evidence and analysis for determining whether to prepare an environmental impact statement or a finding of no significant impact.” Id. § 1508.9(1). CEQ regulations apply to all federal agencies. Id. § 1501.2; Andrus v. Sierra Club, 442 U.S. 347, 351, 99 S.Ct. 2335, 2338, 60 L.Ed.2d 943 (1979).
Two fundamental principles underlie NEPA’s requirements: federal agencies have the responsibility to consider the environmental effects of major actions significantly affecting environment, and the public has the right to review that consideration. Baltimore Gas & Electric Co. v. Natural Resources Defense Council, Inc., 462 U.S. 87, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983). NEPA’s dual mission is thus to generate federal attention to environmental concerns and to reveal that federal consideration for public scrutiny.
In passing NEPA Congress emphasized its particular concern with the role of new technologies and their effect on the environment. The statute explicitly enumerates “new and expanding technological advances” as one of the activities with the potential to threaten the environment. 42 U.S.C. § 4331(a). The legislative history reveals an underlying concern with “[a] growing technological power * * * far outstripping man’s capacity to understand and ability to control its impact on the environment.” S.Rep. No. 91-296, supra, at 6. One of NEPA’s main functions was to bolster this capacity to understand and control the effects of new technology. See Scientists’ Institute for Public Information v. AEC, 481 F.2d 1079, 1089-1090 (D.C.Cir.1973).
NEPA thus stands as landmark legislation, requiring federal agencies to consider the environmental effects of major federal actions, empowering the public to scrutinize this consideration, and revealing a special concern about the environmental effects of new technology.
B. Genetic Engineering
Genetic engineering is an important development at the very cusp of scientific advances. More than a decade ago scientists discovered a method for transplanting deoxyribonucleic acid (DNA), the principal substance of genes. Although exchanges and mutations of DNA occur .in nature, genetic engineering provides the ability to control these fundamental processes of life and evolution. DNA segments can be recovered and cloned from one organism and inserted into another. The result is known as “recombinant DNA.” See generally 41 Fed.Reg. 27903-27904 (July 7, 1976), Joint Appendix (JA) 231-232; Office of Technology Assessment, U.S. Congress, Commercial Biotechnology: An International Analysis 33-38 (January 1984).
Recombinant DNA technology has been limited primarily to small organisms, usually bacteria. This production of new bacteria through altering genetic material has been confined to the laboratory; organisms with recombinant DNA have never been released into the general environment.
Broad claims are made about both the potential benefits and the potential hazards of genetically engineered organisms. Use of recombinant DNA may lead to welcome advances in such areas as food production and disease control. At the same time, however, the environmental consequences of dispersion of genetically engineered organisms are far from clear. According to a recent report by a House of Representatives subcommittee, “The potential environ-' mental risks associated with the deliberate release of genetically engineered organisms or the translocation of any new organism into an ecosystem are best described as ‘low probability, high consequence risk’; that is, while there is only a small possibili*127ty that damage could occur, the damage that could occur is great.” The Environmental Implications of Genetic Engineering, Report by Subcommittee on Investigations & Oversight to House Committee on Science & Technology, 98th Cong., 2d Sess. 9 (1984) (hereinafter cited as Genetic Engineering Report), JA 167.
C. Federal Oversight of Genetic Engineering
Spurred by scientists involved in genetic research, NIH began efforts to oversee genetic engineering in the mid-1970’s. Federal oversight of deliberate release experiments falls into four periods: (1) NIH’s 1976 standards, which prohibited deliberate release of organisms containing recombinant DNA; (2) NIH’s 1978 revision, which gave the NIH Director power to approve deliberate release experiments; (3) the NIH Director’s approval of three such experiments in the early 1980’s; and (4) the District Court’s injunction prohibiting the University of California experiment and enjoining NIH approval of all other deliberate release experiments.2
1. NIH’s 1976 standards: prohibition on deliberate release. In 1976 the NIH Director issued “Guidelines for Research on Recombinant DNA Molecules.” 41 Fed. Reg. at 27902, JA 230. The Guidelines were an historic development, representing the first major federal effort to oversee genetic research and the culmination of intense scientific attention to the possible hazards of genetic research.
In 1974 scientists working in genetic research voluntarily called for a moratorium on certain kinds of experiments until an international meeting could be convened to consider the potential hazards of recombinant DNA molecules. Id. On October 7, 1974 NIH established the Recombinant DNA Advisory Committee (RAC) to consider genetic research issues. And in February 1975 NIH, the National Science Foundation, and the National Academy of Sciences sponsored an international conference at the Asilomar Conference Center in Pacific Grove, California to review the questions posed by the possibility of genetic engineering.
Finally, in the summer of 1976 the NIH Director announced the Guidelines that would govern NIH-supported genetic research experiments. In broad terms, the Guidelines permitted certain laboratory experiments to go forward under carefully specified conditions; certain other types of experiments were flatly prohibited. Deliberate release — “[djeliberate release into the environment of any organism containing a recombinant DNA molecule” — was one of five categories explicitly banned. 41 Fed. Reg. at 27915, JA 243. In announcing the Guidelines the Director noted that deliberate release of organisms with recombinant DNA was not yet feasible and that, if it became feasible, the ban could be reconsidered. Id. at 27907, JA 235. But he stressed that, if such reconsideration occurred, environmental concerns should be paramount: “It is most important that the potential environmental impact of the release be considered.” Id.
Significantly, NIH prepared an EIS to accompany its Guidelines, JA 244 — the only EIS NIH has ever completed on the subject of genetic engineering. The EIS did not specifically refer to deliberate release experiments; such experiments were banned. The EIS did, however, note that dispersion of organisms with recombinant DNA molecules loomed as a potential environmental hazard from the permitted experiments:
Should organisms containing recombined DNA be dispersed into the environment, they might, depending on their fitness relative to naturally occuring [sic] organisms, find a suitable ecological niche for their own reproduction. A potentially dangerous organism might then multiply and spread. Subsequent cessa*128tion of experiments would not stop the diffusion of the hazardous agent.
JA 357.
Thus in 1976 the NIH Guidelines prohibited deliberate release; the Director emphasized the importance of full environmental consideration of any possible future release; and the EIS identified dispersion of organisms with recombinant DNA as a possible environmental hazard.
2. The 1978 revision: permission to waive the prohibition against deliberate release. In 1978 the NIH Director undertook an effort to revise the Guidelines “in light of NIH’s experience operating under them and in light of [NIH’s] increasing knowledge about the potential risks and benefits of this research technique.” 43 Fed.Reg. 33042 (July 28, 1978), JA 433. Proposed in July and adopted in December, the revision changed the Guidelines in several respects. Most importantly for this appeal, the 1978 revision allowed the NIH Director authority to grant exceptions to the five absolute prohibitions in the Guidelines — including the prohibition on deliberate release of organisms containing recombinant DNA into the environment. 43 Fed. Reg. 60108 (December 22, 1978), JA 478.
NIH announced that the standard governing the use of this waiver authority would be the standard generally applicable to the Director’s exercise of his duties: “[T]he Director shall weigh each proposed action, through appropriate analysis and consultation, to determine that it complies with the Guidelines and presents no significant risk to health or the environment.” Id. at 60126, JA 496. NIH also declared that the Director would exercise his authority “with the advice of the Recombinant DNA Advisory Committee after appropriate notice and opportunity for public comment.” Id. at 60108, JA 478. The Director further stated that his “waiver decisions [would] include a careful consideration of the potential environmental impact, and certain decisions may be accompanied by a formal assessment or statement. This must be determined on a case-by-case basis.” 43 Fed.Reg. at 33051, JA 442.
On the subject of deliberate release experiments in particular, the Director suggested that clear standards might be necessary to guide his waiver discretion: “Recognizing the need expressed by * * * commentators for more definitive standards [to govern deliberate release waiver decisions], I will refer the matter to the Recombinant Advisory Committee (RAC) for its consideration. * * * [T]he RAC will be asked to address conditions under which exceptions to various prohibited categories of experiments may be granted.” 43 Fed. Reg. at 60083, JA 471. Thus the Director perceived a possible need for more definitive standards and suggested that such standards might be forthcoming.
NIH did not prepare an EIS to accompany its 1978 revision. It prepared two Environmental Assessments — one for the revision as proposed, 43 Fed.Reg. at 33111, JA 467, and one for the revision as adopted, 43 Fed.Reg. at 60101, JA 471. The assessments said little about the Director’s new waiver authority for deliberate release experiments. The first simply declared, “Waiver decisions will include a careful consideration of potential environmental impact,” 43 Fed.Reg. at 33111, JA 466; the second did not mention the waiver authority-
The 1978 revision also extended the coverage of the Guidelines to all experiments at institutions receiving NIH funds for recombinant DNA research, whether or not the particular experiment had received NIH funds. 43 Fed.Reg. at 60123, JA 493.
3. Approval of deliberate release experiments. The 1978 revision was the last significant revision of NIH’s guidelines regarding deliberate release experiments. A 1982 revision was largely semantic, 47 Fed. Reg. 17186-17187 (April 21, 1982), JA 503-504; a 1983 revision establishing slightly different procedures for deliberate release involving certain plants, 48 Fed.Reg. 24580 (June 1, 1983), is not part of this appeal. The “more definitive standards” suggested by the Director never emerged.
*129Although the guidelines have not changed, NIH’s role has begun to change dramatically. For, with the maturation of genetic engineering technology, NIH has been faced with applications for approval of deliberate release experiments.
The NIH Director, acting on the advice of RAC, has approved three deliberate release experiments at institutions receiving NIH funds for recombinant DNA research. On August 7, 1981 the Director approved a request by Dr. Ronald Davis of Stanford University to field-test corn plants containing recombinant DNA molecules. 46 Fed. Reg. 40331 (August 7, 1981). The goal was to increase the corn’s dietary value by improving its ability to store protein. However, the field tests were never conducted because feasibility problems developed. Genetic Engineering Report at 17, JA 171.
On April 15, 1983 the Director approved a request by Dr. John Sanford of Cornell University to field-test tomato and tobacco plants with recombinant DNA. 48 Fed. Reg. 16459 (April 15, 1983). The goal was to prove that pollen could serve as a “vector” for insertion of recombinant DNA. Again, however, due to feasibility problems, the experiment never went forward. Genetic Engineering Report at 17, JA 171.
On June 1, 1983 the Director gave final approval to the experiment at issue on appeal — the request by Drs. Nickolas Panopoulos and Steven Lindow of the University of California at Berkeley to apply genetically altered bacteria to plots of potatoes, tomatoes, and beans in northern California. 48 Fed.Reg. 24549 (June 1, 1983), JA 522. As discussed in greater detail below, the goal was to increase the crops’ frost resistance. Because .of the cancellation of the previous two experiments, the Panopoulos-Lindow experiment would be the first NIH-approved deliberate release experiment actually to be conducted.3
In February 1984 a congressional subcommittee report sharply criticized NIH’s method of reviewing deliberate release experiments. The report concluded that “the current regulatory framework does not guarantee that adequate consideration will be given to the potential environmental effects of a deliberate release.” Genetic Engineering Report at 10, JA 168. In particular, “the RAC’s ability to adequately evaluate the environmental hazards posed by deliberate releases is limited by both its expertise and its jurisdiction.” Id. The subcommittee report recommended a moratorium on deliberate release approvals until an interagency review panel was established to consider the potential environmental effects of each deliberate release experiment. “Each [deliberate release experiment] could result in major environmental damage or adverse public health effects.” Id. at 43-44, JA 176-177.4
4. The injunction. In September 1983 three public interest organizations and two individuals filed suit against the three federal officials ultimately responsible for NIH deliberate release decisions; they later added Regents of the University of California as a defendant. The University of California experiment was scheduled to begin on or about May 25, 1984. On May 18 the District Court issued an injunction enjoining the University of California experiment and NIH approval of other deliberate release experiments. The District Court found that plaintiffs were likely to succeed in showing that NIH should have completed at least a more complete environmental assessment, and perhaps an EIS, *130before approving the University of California experiment; it also found them likely to succeed in showing that NIH should have completed an Environmental Impact Statement in connection with both its 1978 policy change and its imminent “program” of deliberate release approvals.5
Both the federal defendants and the University of California Regents have filed appeals; the appeals were consolidated and the case heard by this court on an expedited basis.
II. Standard op Review
Two principles guide our analysis of this appeal: the standard for judicial review of an agency decision not to prepare an EIS and the standard for appellate review of a District Court decision to grant an injunction'.
A. Review of an Agency Decision Not to Prepare an EIS
That courts must play a cardinal role in the realization of NEPA’s mandate is beyond dispute. As the Supreme Court recently emphasized, the critical judicial task is “to ensure that the agency has adequately considered and disclosed the environmental impact of its actions and that its decision is not arbitrary or capricious.” Baltimore Gas & Electric, supra, 462 U.S. at 97-98, 103 S.Ct. at 2253. Since NEPA requires the agency to “take a ‘hard look’ at the environmental consequences before taking a major action,” id. (quoting Kleppe v. Sierra Club, 427 U.S. 390, 410 n. 21, 96 S.Ct. 2718, 2730 n. 21, 49 L.Ed.2d 576 (1976)), the judiciary must see that this legal duty is fulfilled. Although the “agency commencing federal action has the initial and primary responsibility for ascertaining whether an EIS is required,” Committee for Auto Responsibility v. Solomon, 603 F.2d 992, 1002 (D.C.Cir.1979), cert. denied, 445 U.S. 915, 100 S.Ct. 1274, 63 L.Ed.2d 599 (1980), the courts must determine that this decision accords with traditional norms of reasoned decisionmaking and that the agency has taken the “hard look” required by NEPA. Sierra Club v. Peterson, 717 F.2d 1409, 1413 (D.C.Cir.1983).
B. Review of an Injunction
We review the decision to grant an injunction as an exercise of the District Court’s discretion. Doran v. Salem Inn, 422 U.S. 922, 932, 95 S.Ct. 2561, 2568, 45 L.Ed.2d 648 (1975); Brown v. Chote, 411 U.S. 452, 457, 93 S.Ct. 1732, 1735, 36 L.Ed.2d 420 (1973). Under this circuit’s long-standing test, the District Court should consider (1) the'plaintiff’s likelihood of prevailing on the merits, (2) the threat of irreparable injury to the plaintiff in the absence of injunctive relief, (3) the possibility of substantial harm to other interested parties from the injunctive relief, and (4) the interests of the public. WMATC v. Holiday Tours, 559 F.2d 841, 842-843 (D.C.Cir.1977); Virginia Petroleum Jobbers Ass’n v. FPC, 259 F.2d 921, 925 (D.C.Cir.1958).
In reviewing this exercise of discretion, we must be conscious of whether we are reviewing findings of fact, conclusions of law, or determinations about the balancing of the injunction factors. In all three we are deferential to the District Court’s determination about “what evidence can properly be adduced in the limited time that can be devoted to a preliminary injunction hearing.” Friends for All Children, Inc. v. Lockheed Aircraft Corp., 746 F.2d 816, 835 n. 32 (D.C.Cir.1984). We are most deferential to the District Court’s balancing of the *131injunction factors. Id. On questions of fact the usual “clearly erroneous” standard applies, id.; cf. Fed.R.Civ.P. 52(a). We are least deferential on questions of law. “[A] greater amplitude of judicial review is called for when the appeal presents a substantial issue that the action of the trial judge was based on a premise as to the pertinent rule of law that was erroneous. * * * When this can be identified, the appellate court furthers the interest of justice by providing a ruling on the merits to the extent that the matter is ripe, though technically the case is only at the stage of application for preliminary injunction.” Natural Resources Defense Council, Inc. v. Morton, 458 F.2d 827, 832 (D.C.Cir.1972); see also Ambach v. Bell, 686 F.2d 974, 979 (D.C.Cir.1982) (per curiam).
With these principles in mind, we turn to the District Court’s injunction as it applies, first, to NIH’s approval of the University of California experiment, and, second, to NIH’s possible approval of all other deliberate release experiments.
III. The University of California Experiment
The University Regents raise objections about the District Court’s view that the environmental assessment of the University of California experiment was inadequate; they also deploy a barrage of procedural objections. We will address the adequacy of the assessment and the procedural objections in turn. The federal appellants claim not to appeal the part of the injunction that applies to the U.C. experiment, but they continue to insist that the only flaw in NIH’s environmental assessment is that NIH did not publish the results of its environmental review in a document titled “Environmental Assessment.” Statement of counsel at oral argument; see also brief for federal appellants at 5 n. 3. We will address that contention as well.
A. The Adequacy of the Environmental Review
1. The proposed experiment. On September 17, 1982 Drs. Lindow and Panopoulos, scientists at Berkeley, submitted a request for NIH approval of an experiment that would involve deliberate release of genetically altered organisms in the open environment. NIH approval was required because the University of California receives NIH funds for recombinant DNA research. Lindow and Panopoulos proposed to apply the genetically altered bacteria to various crops, including potatoes, tomatoes, and beans. By changing the bacteria’s genetic composition, Lindow and Panopoulos hoped that the bacteria would change from frost-triggering bacteria to non-frost-triggering bacteria; they further hoped that the engineered non-frost-triggering bacteria would displace the natural frost-triggering bacteria. The ultimate goal was to protect the crops from frost and thus to extend their growing season. Such non-frost-triggering bacteria occur in nature as products of natural mutation, but Lindow and Panopoulos apparently hoped that the genetically engineered organisms would be more stable than the natural mutants. They sought to treat crops at six sites; the workers applying the recombinant-DNA-containing bacteria would wear respirators to reduce the risk of inhalation. Application of Drs. Lindow and Panopoulos, September 17, 1982, JA 728-736.
2. NIH review. NIH announced the Lindow-Panopoulos request for approval, the RAC meeting at which it would be considered, and the opportunity to comment. 47 Fed.Reg. 41925 (September 22, 1982). No comments were received. At the RAC meeting on October 25, 1982 RAC members raised questions about the number of sites, the lack of adequate information, and the possible effects on rainfall. RAC voted to recommend that the Director approve the project; the vote was seven in favor, five opposed, with two abstentions. RAC Minutes, JA 684-686. The Director decided to postpone approval and suggested further consideration.
Lindow and Panopoulos resubmitted their proposal with some modifications, including a reduction of experiment sites from six to one. On April 11, 1983, after *132some discussion, RAC voted to recommend approving the proposal by a vote of 19-0, with no abstentions. RAC Minutes, JA 713-716. The Director then approved the experiment. 48 Fed.Reg. 24549 (June 1, 1983), JA 523.
3. NEPA compliance. NIH’s consideration of the Lindow-Panopoulos experiment falls far short of the NEPA requirements. And, despite the government’s apparent belief, the deficiency is not a question of which document contains the environmental analysis. Rather, the deficiency rests in NIH’s complete failure to consider the possibility of various environmental effects.
Neither the government nor the University seriously disputes that an environmental assessment is necessary. The government has conceded that the approval is a “major action” and that it does not fall into a categorical exclusion to the EIS requirements, Federal Defendants’ Response to Plaintiffs’ Second Set of Interrogatories at 17, JA 50; see also Foundation on Economic Trends v. Heckler, 587 F.Supp. 753, 767 (D.D.C.1984). The University’s contention here — and the government’s contention below, as well as its apparent continuing belief — is that the environmental consideration given by NIH was equivalent to the necessary environmental assessment and that the injunction requires only a document labelled “Environmental Assessment.” We thus fear that the University and the government completely misapprehend the District Court’s hqlding and the requirements imposed by NEPA.
The most glaring deficiency in NIH’s review of the Lindow-Panopoulos experiment is its treatment of the possibility of dispersion of recombinant-DNA-containing organisms. As noted, NIH’s only EIS on genetic engineering specifically identified dispersion as one of the major environmental concerns associated with recombinant DNA research. The consequences of dispersion of genetically altered organisms are uncertain. Some observers believe that such dispersion would affect the environment and the climate in harmful ways. See, e.g., Genetic Engineering Report at 9, JA 167 (“the risk presented by the deliberate release of a genetically engineered organism is that it may cause environmental changes that perturb the ecosystem it encounters and/or that the organism itself may have negative effects if it establishes itself outside of the specific environment for which it was intended”).
Thus the problem of dispersion would seem to be one of the major concerns associated with the Lindow-Panopoulos experiment, the first experiment that would actually release genetically engineered organisms in the open environment. Yet in the minutes of the RAC meeting- — the only document on appeal that records any NIH consideration of the environmental impact of dispersion — the entirety of the consideration of dispersiop is the following statement: according to a RAC evaluator, “Although some movement of bacteria toward sites near treatment locations by insect or aerial transport is possible, the numbers of viable cells transported has been shown to be very small; and these cells are subject to biological and physical processes limiting survival.” RAC Minutes (April 11, 1983), JA 715 (emphasis added). In this sentence, which was taken almost verbatim from the Lindow-Panopoulos proposal, the RAC evaluator thus conceded the possibility of aerial or insect transport, but merely commented that the number of viable cells would be small, and that they were subject to processes limiting survival. Remarkably, therefore, RAC completely failed to consider the possible environmental impact from dispersion of genetically altered bacteria, however small the number and however subject to procedures limiting survival.6
*133In light of this complete failure to address a major environmental concern, NIH’s environmental assessment utterly fails to meet the standard of environmental review necessary before an agency decides not to prepare an EIS. The argument that this consideration would be adequate if contained in a document labelled “Environmental Assessment” simply misconceives the clear requirements of NEPA as articulated by the courts, Baltimore Gas & Electric, supra; Kleppe v. Sierra Club, supra, 427 U.S. at 410 n. 21, 96 S.Ct. at 2730 n. 21; Cabinet Mountains Wilderness v. Peterson, 685 F.2d 678, 682 (D.C.Cir.1982), and by the Council on Environmental Quality, 40 C.F.R. §§ 1501.4, 1508.9. An environmental assessment that fails to address a significant environmental concern can hardly be deemed adequate for a reasoned determination that an EIS is not appropriate. See Vermont Yankee Nuclear Power Corp. v. NRDC, 435 U.S. 519, 553, 98 S.Ct. 1197, 1216, 55 L.Ed.2d 460 (1978) (“NEPA places upon an agency the obligation to consider every significant aspect of the environmental impact of a proposed action”).
Appellants also contend that the adequacy of the environmental assessment can be divined from the NIH Director’s final approval — and his accompanying statement of “no significant risk,” 48 Fed.Reg. 24548 (June 1, 1983), JA 522, as required by the 1978 revision. This contention also reveals a fundamental misunderstanding about the adequacy of an environmental assessment. Simple, conclusory statements of “no impact” are not enough to fulfill an agency’s duty under NEPA. The only consideration of environmental consequences was apparently the RAC review; as shown, that consideration is inadequate to support the Director’s finding. To accept the Director’s conclusory statement would violate principles of reasoned decisionmaking, Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 245, 9 L.Ed.2d 207 (1962), NEPA’s policy of public scrutiny, Baltimore Gas & Electric, supra, and CEQ’s own regulations, 40 C.F.R. §§ 1501.4, 1508.9.
It should be stressed that this inquiry into the adequacy of an environmental assessment is ultimately relevant to the agency’s determination that its proposed federal action will not have a “significant impact” on the environment — and thus no EIS is required. In that connection, it is notable that NIH never directly addressed the question whether an EIS. should be prepared. Such an inquiry is, of course, the ultimate purpose of an environmental assessment. Maryland-Nat’l Capital Park & Planning Comm’n v. U.S. Postal Service, 487 F.2d 1029, 1039-1042 (D.C.Cir.1973).
To reiterate, NIH must first complete a far more adequate environmental assessment of the possible environmental impact of the deliberate release experiment than it has yet undertaken. That assessment must “provide sufficient evidence and analysis for determining whether to prepare an environmental impact statement or a finding of no significant impact,” 40 C.F.R. § 1508.9(a)(1). Ignoring possible environmental consequences will not suffice. Nor will a mere conclusory statement that the number of recombinant-DNA-containing organisms will be small and subject to processes limiting survival. Instead, NIH must attempt to evaluate seriously the risk that emigration of such organisms from the test site will create ecological disruption. Second, until NIH completes such an evaluation the question whether the experiment requires an EIS remains open. The University of California experiment clearly presents the possibility of a problem identified by NIH in its EIS as a potential environmental hazard. This fact weighs heavily in support of the view that an EIS should be completed, unless NIH can demonstrate either that the experiment does not pose the previously identified danger, or that its assessment of the previously *134identified danger has changed through a process of reasoned decisionmaking. Nor is it sufficient for the agency merely to state that the environmental effects are currently unknown. Indeed, one of the specific criteria for determining whether an EIS is necessary is “[t]he degree to which the possible effects on the human environment are highly uncertain or involve unique or unknown risks.” 40 C.F.R. § 1508.27(b)(5).
Thus we approve the District Court’s determination that, as a matter of law, plaintiffs are likely to prevail in showing that NIH’s environmental assessment of the University of California experiment — and its discharge of its statutory duty to consider the propriety of an EIS — was wholly inadequate. And, in light of the government’s contention, we emphasize as strongly as we can that the problem lay in the adequacy of the assessment itself, not in the document in which the assessment was contained.
B. Procedural Objections
The University appellant raises four procedural objections to the District Court’s injunction against the University of California experiment: (1) the District Court lacked power to issue an injunction against the University; (2) plaintiffs did not exhaust their administrative remedies; (3) the District Court’s review of the record was misconceived; and (4) the District Court’s balancing of the various factors regarding the injunction was flawed. None of these arguments carries the day.
1. Judicial power. The University claims that the District Court “had no power to enjoin the University, as a private party,” brief for appellant Regents at 49, because NEPA applies only to federal agencies. However, as the University recognizes, it is well established that judicial power to enforce NEPA extends to private parties where “non-federal action cannot lawfully begin or continue without the pri- or approval of a federal agency.” Biderman v. Morton, 497 F.2d 1141, 1147 (2d Cir.1974). “[Wjere such non-federal entities to act without the necessary federal approval, they obviously would be acting unlawfully and subject to injunction.” Id.
The University maintains, however, that the Guidelines require only NIH approval as set forth in the NIH Guidelines— and, since the experiment received that approval, the University discharged its responsibilities under the Guidelines, whether or not a court finds the approval invalid under NEPA. This argument is completely meritless. For an NIH approval to be valid under the Guidelines, the approval must comport with NEPA. This requirement is stated explicitly in the NIH Guidelines, 43 Fed.Reg. 33110 (July 28, 1978), JA 465; even if it were not, approvals by NIH — federal actions — would be valid only if the agency discharged its duties under NEPA. Without valid NIH approval under NEPA, the University cannot lawfully go forward with its experiment, and it can thus be enjoined by the court.7
We note also that the justification for NIH’s requirement of approval — a requirement that has not been challenged by the University — is that the University receives funds for recombinant DNA research. Federal funding has long been recognized as an appropriate basis to enforce NEPA’s requirements on non-federal parties. See Biderman v. Morton, supra, 497 F.2d at 1147; Bradford Township v. Illinois State Highway Authority, 463 F.2d 537, 540 (7th *135Cir.), cert. denied, 409 U.S. 1047, 93 S.Ct. 518, 34 L.Ed.2d 499 (1972); Named Individual Members of San Antonio Conservation Society v. Texas Highway Dep’t, 446 F.2d 1013, 1028 (5th Cir. 1971), cert. denied, 406 U.S. 933, 92 S.Ct. 1775, 32 L.Ed.2d 136 (1972). In view of the University’s commitment to abide by the binding conditions of the Guidelines, we need not decide whether the funding would itself provide a sufficient basis to support an injunction against the University.
2. Exhaustion. The University appellant argues that plaintiffs’ objections to the University of California experiment should be barred because plaintiffs failed to raise the objections during the period of NIH consideration. The University appellant points out that NIH announced the experiments and the RAC meetings in the Federal Register and invited public comment.
The general rule on exhaustion of administrative remedies is that “claims not presented to the agency may not be made for the first time to a reviewing court.” Washington Ass’n for Television & Children v. FCC, 712 F.2d 677, 680 (D.C.Cir.1983). However, this well-established exhaustion doctrine is ultimately an exercise of judicial discretion. Action for Children’s Television v. FCC, 564 F.2d 458, 469 (D.C.Cir.1977); Great Falls Community TV Cable Co. v. FCC, 416 F.2d 238, 239 (9th Cir.1969); Joseph v. FCC, 404 F.2d 207, 210 (D.C.Cir.1968). In “exceptional cases or particular circumstances * * * where injustice might otherwise result,” the reviewing court may “consider questions of law which were neither pressed nor passed upon by the * * * administrative agency below.” Hormel v. Helvering, 312 U.S. 552, 557, 61 S.Ct. 719, 721, 85 L.Ed. 1037 (1941). In this case the issue — the appropriate environmental review for the first deliberate release of genetically engineered organisms — is one of great public importance. Indeed, this imminent application of a new technology with unknown environmental consequences is precisely the kind of situation NEPA is intended to address. See Scientists’ Institute for Public Information v. AEC, supra, 481 F.2d at 1089. Given the public importance of this issue, we refuse to find that the decision to hear plaintiffs’ challenge is an abuse of judicial discretion, particularly in view of NEPA’s clear policy favoring public scrutiny of environmental considerations, Baltimore Gas & Electric, supra8
Additionally, we note that the exhaustion doctrine is premised on a view of fairness to the agency and to the litigants. United States v. L.A. Tucker Truck Lines, Inc., 344 U.S. 33, 37, 73 S.Ct. 67, 69, 97 L.Ed. 54 (1952). The University’s claims of unfairness to NIH and to itself are unconvincing. As noted, the possibility of dispersal was not a surprise: NIH explicitly recognized the possibility, but completely failed to consider its possible environmental impact. There was thus no question of unfairness or surprise; there was simply the lack of consideration to the environmental effects of the admitted possibility of dispersion. In these circumstances, a failure to object during the administrative process should not bar plaintiffs.
3. The District Court review. The University appellant also attacks the District Court’s review of the record on two grounds: it argues that the District Court improperly failed to review the entire administrative record, and that it improperly cast a burden of rebuttal upon the agency. Neither of these contentions is substantial. The District Court did note that “NEPA does not require this Court to reconstruct the environmental deliberations of an agency for their adequacy when the agency chooses not to follow the steps out*136lined in NEPA.” 587 F.Supp. at 767. However, in context this comment must be read to mean that the agency’s deliberation should be part of the administrative record; the court will not supply reasons that the agency has not revealed in its decisionmaking process. See Motor Vehicle Mfrs Ass’n v. State Farm Mutual Auto. Ins. Co., 463 U.S. 29, 41, 103 S.Ct. 2856, 2865, 77 L.Ed.2d 443 (1983) (hereinafter cited as State Farm to 103 S.Ct. only). Indeed, the District Court emphasized that it had reviewed both the “formal” and the “informal” record of the agency decisionmaking process. 587 F.Supp. at 767-768.9 Similarly, the University seeks to make much of the District Court’s use of the word “rebutted” in its conclusion that “plaintiffs have identified several areas of plausible environmental concerns which are not rebutted” in the agency record. Id. at 767. It is clear, however, that, far from putting a rebuttal burden on the agency, the District Court was merely referring to whether the agency had considered the problem. Thus both arguments can be dismissed as semantic.
4. Balancing of the injunction criteria. The University also argues that, whatever plaintiffs’ likelihood of success on the merits, the injunction is not justified under the other three criteria that the court must balance in weighing a request for injunctive relief — the threat of irreparable injury to the plaintiffs, the possibility of injury to other interested parties, and the public interest.
The District Court’s discretion in balancing these factors is broad. “The nature of the discretion peculiar to the preliminary injunction lies in the latitude given to the District Court to engage in a balancing of the traditional factors * * Friends for All Children, Inc. v. Lockheed Aircraft Corp., supra, 746 F.2d at 835 n. 32. The District Court’s determinations about the harm to plaintiffs, the harm to other parties, and the public interest were eminently sound, and far from the abuse of discretion that would justify reversal.
One point bears further elucidation. The University seeks to minimize the injury that would result if its experiment proceeds without adequate environmental consideration. In doing so the University fundamentally misconceives the nature and significance of NEPA’s requirements. The NEPA duty is more than a technicality; it is an extremely important statutory requirement to serve the public and the agency before major federal actions occur. Jones v. D.C. Redevelopment Land Agency, 499 F.2d 502, 513 (D.C.Cir.1974), cert. denied, 423 U.S. 937, 96 S.Ct. 299, 46 L.Ed.2d 271 (1975); Environmental Defense Fund v. TVA, 468 F.2d 1164, 1184 (6th Cir.1972); Scherr v. Volpe, 466 F.2d 1027, 1034 (7th Cir.1972); Calvert Cliffs’ Coordinating Committee v. USAEC, supra, 449 F.2d at 1115. If plaintiffs succeed on the merits, then the lack of an adequate environmental consideration looms as a serious, immediate, and irreparable injury. Although the balancing of this harm against other factors is necessarily particularized, State of Alaska v. Andrus, 580 F.2d 465, 485 (D.C.Cir.), vacated in part on other grounds sub nom. Western Oil & Gas Ass’n v. Alaska, 439 U.S. 922, 99 S.Ct. 303, 58 L.Ed.2d 315 (1978), the injury itself is clear.
*137In conclusion, we find that the University’s procedural objections are meritless. None of those objections prevents us from reaching our central holding: the District Court’s finding that plaintiffs are likely to succeed in showing that NIH failed to give adequate environmental consideration to the University of California experiment was entirely correct, and the resulting injunction completely proper.
IV. The Injunction Against NIH Approval of All Other Deliberate Release Experiments
The District Court also found that plaintiffs were likely to succeed in showing that NIH should have completed an EIS (1) in 1978, when it revised the Guidelines to make deliberate release possible, and (2) in 1982 or 1983, when it became clear that NIH would be receiving proposals for deliberate release of genetically engineered organisms. The District Court thus enjoined NIH from approving all other deliberate release experiments.
We share the District Court’s view that the administrative record is almost completely bereft of evidence suggesting that NIH has sought to address the question of deliberate release experiments systematically. We find, however, that as a matter of law we must reverse the District Court’s holding that plaintiffs are likely to succeed in showing that NIH is now required to undertake an EIS in connection with the 1978 policy change, or a programmatic EIS in connection with the impending requests for approval of deliberate release experiments.
A. The 1978 Policy Change
The District Court found that a review of the 1978 revision “suggests that plaintiffs will prevail on their claims of a violation of either NEPA, 42 U.S.C. § 4332(2)(C), or the Administrative Procedure[ ] Act, 5 U.S.C. § 706(2)(D).” 587 F.Supp. at 762. The District Court reasoned that the revision — permitting the possibility of deliberate release — was a “major federal action.” NIH itself decreed the 1978 change a “major revision” of the EIS-accompanied 1976 Guidelines. The District Court further found that the administrative record revealed no evidence that the NIH Director had considered the possibility of a significant environmental effect. The first EA merely stated that any approval must “include a careful consideration of potential environmental impact,” 43 Fed.Reg. 33111 (July 28, 1978), JA 466; the second EA was silent on the impact of the change; and the Director, in his decision to publish the final Guidelines, deferred elaboration of standards until later. The District Court thus concluded that plaintiffs were likely to succeed in showing that the NIH decision not to complete an EIS accompanying its 1978 revision violated NEPA.
The federal appellants and the University vigorously dispute this finding, and they raise several substantive and procedural objections to it. We need not consider most of these contentions. The emphasis on the 1978 revision is misguided for two reasons. First, the policy revision did not irrevocably commit NIH to any decision. Direct release was not then an imminent possibility, and NIH retained authority to grant or deny approval on a case-by-case basis. “NEPA requires an agency to evaluate the environmental effects of its action at the point of commitment,” Sierra Club v. Peterson, supra, 717 F.2d at 1414, and the 1978 policy change did not necessarily represent the “point of commitment” that triggers NEPA. Indeed, the District Court itself recognized that the 1978 decision could be seen as a “decision deferred,” 587 F.Supp. at 769, rather than a commitment. Second, significant changes have occurred since 1978, including the maturation of genetic engineering to a point where deliberate release is feasible and imminent. Agency determinations about EIS requirements are supposed to be forward-looking, Nat’l Wildlife Federation v. Appalachian Regional Comm’n, 677 F.2d 883, 889 (D.C.Cir.1981), and it makes little sense to rest a judgment on prospective relief on decisions *138made several years ago under very different circumstances.
Given these two circumstances, the focus on the 1978 revision as an independent basis for an injunction and an EIS requirement is inappropriate.
B. The Programmatic EIS
The District Court found that plaintiffs are likely to prevail in showing that NIH should complete a programmatic EIS. As this court has explained, under NEPA two types of EIS are possible: programmatic and specific. “A programmatic EIS reflects the broad environmental consequences attendant upon a wide-ranging federal program. The thesis underlying programmatic EISs is that a systematic program is likely to generate disparate yet related impacts. * * * Whereas the programmatic EIS looks ahead and assimilates ‘broad issues’ relevant to [the program], the site-specific EIS addresses more particularized considerations * * Nat’l Wildlife Federation v. Appalachian Regional Comm’n, supra, 677 F.2d at 888.
Various standards guide judicial review of whether an agency should complete a programmatic EIS. The Supreme Court has emphasized that the question of preparing a programmatic EIS is initially committed to the agency. Kleppe v. Sierra Club, supra, 427 U.S. 390, 96 S.Ct. 2718, 49 L.Ed.2d 576. Under CEQ regulations a programmatic EIS should be prepared if actions are “connected,” “cumulative,” or sufficiently “similar” that a programmatic EIS is “the best way” to identify the environmental effects. 40 C.F.R. § 1508.25. As this court has noted, “Two considerations are especially helpful in reviewing the responsible officials’ decision not to prepare a programmatic EIS: (a) Could the programmatic EIS be sufficiently forward looking to contribute to the decisionmakers’ basic planning of the overall program? and (b) Does the decisionmaker purport to ‘segment’ the overall program, thereby unreasonably constricting the scope of * * * environmental evaluation?” Nat’l Wildlife Federation v. Appalachian Regional Comm’n, supra, 677 F.2d at 889. Thus a programmatic EIS should be prepared if it can be forward-looking and if its absence will obstruct environmental review.
The District Court found that “the federal defendants have passed the point at which a programmatic EIS is required by NEPA and the Council on Environmental Quality’s regulations.” 587 F.Supp. at 764. Though we share the District Court’s view that a programmatic EIS would be helpful, we are not prepared to agree, on the record before us, that, as a matter of law, plaintiffs are likely to succeed in showing that the absence of a programmatic EIS unreasonably constricts environmental evaluation. If NIH gives adequate environmental consideration to each deliberate release experiment — and, as noted, that consideration must be far more complete and rigorous than the consideration NIH has thus far undertaken — then we cannot, at this point, say that such consideration will obstruct environmental review. We thus reverse the District Court’s finding that plaintiffs are likely to succeed in showing, at this point, that the governing law requires NIH to complete a programmatic EIS on deliberate release experiments.
This reversal, however, should not be construed to mean that NIH need give no consideration to the environmental effects of deliberate release experiments beyond the individual experiments. Since NIH has given no serious consideration to whether a programmatic EIS is justified, we cannot evaluate its claims that deliberate release experiments are neither so “cumulative” or “connected” that a programmatic EIS is required under the CEQ regulations, nor so “similar” that a programmatic EIS may not be “the best way to assess adequately” their environmental effects. 40 C.F.R. § 1508.25. Additionally, we note that federal development of a new technology with unknown environmental consequences is the type of action in which programmatic considerations are particularly important. Scientists’ Institute for Public Information v. AEC, supra, 481 F.2d at 1089-1090. NIH is about to begin a process of review*139ing what will be a stream of applications for approval of a new technology with unknown environmental consequences. Under these circumstances, NIH should at least consider whether a programmatic EIS is required to fulfill the agency’s legal obligations under NEPA and the CEQ regulations. We thus conclude that, if NIH does not at least consider the advisability of a programmatic EIS, its approval of individual deliberate release experiments is likely to violate established principles of reasoned decisionmaking. See State Farm, supra, 104 S.Ct. at 2867 (agency may not “entirely fail[] to consider an important aspect of the problem”). And, unlike NIH’s completely conclusory statement that it would not prepare a programmatic EIS, 49 Fed. Reg. 697 (January 5, 1984), reasoned consideration of these important aspects of the problem should reflect an articulated, “rational connection between the facts found and the choice made.” Burlington Truck Lines, Inc. v. United States, supra, 371 U.S. at 168, 83 S.Ct. at 246.10
In short, we vacate the part of the injunction enjoining NIH approval of all deliberate release experiments, but we stress that NIH’s attention to important issues regarding deliberate release experiments will continue to influence whether its individual decisions may be found to reflect the “rule of reason,” Committee for Auto Responsibility v. Solomon, supra, 603 F.2d at 1003, that measures the adequacy of environmental review.
IV. Conclusion
NIH, no less than any other federal agency, must ensure that its decisions meet the standards of environmental concern and reasoned decisionmaking required by law. The complexity and uncertainty of the issues before NIH do not diminish these responsibilities one iota. Indeed, “one of the functions of a NEPA statement is to indicate the extent to which environmental effects are essentially unknown. * * * [W]e must reject any attempt by agencies to shirk their responsibilities under NEPA by labeling any and all discussion of future environmental effects as ‘crystal ball inquiry.’” Scientists’ Institute for Public Information v. AEC, supra, 481 F.2d at 1092.
We affirm the injunction as it applies to the University of California experiment. We vacate the injunction as it applies to NIH’s approval of all other deliberate release experiments, but we share and support the District Court’s concern that NIH has not yet given adequate consideration to broad and important issues relating to its role in approving deliberate release experiments.
Affirmed in part and vacated in part.
. The organizational plaintiffs are Foundation on Economic Trends, Environmental Action, Inc., and Environmental Task Force. The original individual plaintiffs are Jeremy Rifkin and Michael W. Fox. (Individual plaintiffs David Brower and William Turnage were added subsequently.) The federal defendants are Margaret M. Heckler, Secretary of the Department of Health and Human Services; James F. Wyngaarden, Director, National Institutes of Health; and Richard M. Krause, Director, National Institute of Allergy and Infectious Diseases, National Institutes of Health.
. In addition to overseeing genetic research, NIH has provided significant financial support to stimulate development of genetic engineering technology. Office of Technology Assessment, U.S. Congress, Commercial Biotechnology; An International Analysis 309-310 (January 1984).
. "NIH approval” in this opinion refers to approval that is mandatory under NIH Guidelines — experiments at institutions receiving NIH funds for recombinant DNA research. The Guidelines also provide a mechanism for NIH review of voluntarily submitted private experiments. 49 Fed.Reg. 24563-24564 (June 1, 1983). Although NIH has considered some deliberate release experiments through this mechanism, see Genetic Engineering Report at 18 n. 23, JA 173, the record contains no evidence that any deliberate release experiment, with or without NIH approval, has ever been undertaken.
. See also Commercial Biotechnology, supra note 2, at 39 (NIH "is in a position of potentially conflicting interests. It serves both as a quasi-regulator of genetic research through the NIH Guidelines and as a promoter of genetic research through its sizable funding of genetic research.").
. The District Court refused to enjoin NIH’s approval of voluntarily submitted private experiments because of jurisdictional concerns. Foundation on Economic Trends v. Heckler, 587 F.Supp. 753, 766-767 (D.D.C.1984). However, the significance of findings about NIH’s approvals extends beyond those experiments that require approval for several reasons. First, private companies voluntarily comply with the Guidelines. Commercial Biotechnology, supra note 2, at 357. Second, some states and localities require compliance with the Guidelines from all businesses engaged in genetic research. Genetic Engineering Report at 28, JA 175. And third, the NIH Guidelines and practice are likely to be the yardstick for any common law liability-
. The University’s arguments defending the NIH review are completely unconvincing. For instance, the University emphasizes that the use of chemically induced non-frost-triggering bacteria has produced “no untoward environmental consequences.” Brief for appellant Regents at 42. But the Lindow-Panopoulos proposal itself stressed that the genetically engineered bacteria would have greater "genetic stability” and "competitive fitness” than the chemically-induced *133bacteria. JA 829. Similarly, the University points to the presence of non-frost-triggering bacteria in nature. Brief for appellant Regents at 42. At oral argument, however, the University reported that these natural populations, like the chemical mutants, are probably less stable and competitive than the genetically engineered bacteria.
. The University argues that this appeal differs from the cases cited in Biderman because the University’s obligation to receive NIH approval arises as a condition of the University's decision to receive federal funds for recombinant DNA research. However, NIH approval of deliberate release experiments is an explicit, regulatorilycreated condition to receipt of federal funds for recombinant DNA research, as the University well knew in accepting the federal funds. Whether the requirement of federal approval arises as a statutorily-created condition, as in the Biderman cases, or as a regulatorily-created condition, as in this case, is of no particular moment to the underlying question of judicial power. Most fundamentally, we reject the notion that a federal court is unable to enjoin a party who proposes to defy an explicit regulatory condition governing receipt and expenditure of federal funds.
. For all of its salutary effect on rational government decisionmaking, see S. Breyer & R. Stewart, Administrative Law and Regulatory Policy 63 (1979); 5 U.S.C. § 552 (1982), the Federal Register was never intended to be a means of limiting public discourse and judicial access to those who scrutinize its daily notices. Such a strong-armed invocation of Federal Register publication as a bar to subsequent objection raised in a reasonably timely fashion is particularly inappropriate in the context of a statute like NEPA that seeks to further public attention and debate.
. Whatever the District Court’s review, this court has reviewed the entire administrative record — the researchers’ proposals, the RAC minutes, and the Federal Register notices — and we fully support the District Court’s findings. We have had no indication that there is anything more to the administrative record. The RAC process, moreover, is relevant only insofar as it reveals the NIH Director's decisionmaking process. As the NIH Director has made clear, "It is not the function of the RAC to determine what NEPA and the CEQ regulations require. The RAC is not constituted to interpret points of law and the requirements of NEPA. Specifically, it is not a function of RAC to determine when an environmental impact statement or an environmental assessment is required by NEPA.” 49 Fed.Reg. 697 (Jan. 5, 1984). Thus the relevant administrative record is the record that was before the Director at the time of his decision. Walter O. Boswell Memorial Hospital v. Heckler, 749 F.2d 788, 789, 792-794 (D.C.Cir.1984).
. The choice of the document in which to consider these issues is, of course, the agency’s, but the first environmental assessment accompanying a deliberate release experiment may well be an appropriate and convenient forum for NIH to address these concerns.
We note also that, despite the Director’s comments in 1978 about the helpfulness of general standards in guiding individual experiment decisions, no such standards have yet been implemented. Through post-argument submissions, however, the parties advise us that RAC is now taking steps to develop standards.