In these consolidated appeals, we review (1) a ruling that an environmental impact statement (“EIS”) for an insecticide spraying program to combat infestations of gypsy moths satisfies the National Environmental Policy Act of 1969 (“NEPA”), 42 U.S.C. § 4332(C) (1982), and its implementing regulations, 40 C.F.R. §§ 1500-1508 (1986), and (2) the denial of attorneys’ fees under the Equal Access to Justice Act (“EAJA”), 28 U.S.C. § 2412(d)(1)(A) (Supp.III 1985), for certain parts of this litigation. We affirm on the adequacy of the EIS. On the attorneys’ fees issues, we affirm in part and vacate and remand in part.
I. PROCEDURAL BACKGROUND
This litigation commenced in 1982, when Oregon Environmental Council, Citizens for the Safe Control of the Gypsy Moth, and two Oregon residents (collectively “OEC”) brought suit to enjoin various state and federal officials, including the U.S. Secretary of Agriculture (“the Secretary”), from performing or authorizing aerial spraying of carbaryl, a chemical insecticide, in a federally assisted state program to eradicate gypsy moths in a. populated, residential area of South Salem, Oregon. OEC challenged the adequacy of a Programmatic Environmental Impact Statement (“the 1982 PEIS”) that had been developed for spraying in forested areas of the northeastern United States and a three-page environmental assessment (“EA”) describing the South Salem spraying program. The litigation progressed in four phases.
Phase I ended in 1983, when the Ninth Circuit held that the 1982 PEIS and EA did not satisfy NEPA’s requirements for the South Salem spraying program. Oregon Environmental Council v. Kunzman, 714 F.2d 901, 904-05 (9th Cir.1983) [hereinafter OEC I]. We determined that the differences between spraying in forested regions under the 1982 PEIS and spraying in cleared, suburban and urban areas were “too significant for compliance with NEPA.” Id. at 905. We noted that the 1982 PEIS did not discuss the effects of direct contact with carbaryl, of contact by children or sensitive individuals, and of new techniques for combatting gypsy moths. See id. at 904-05. On remand, the district court enjoined the defendants from spraying carbaryl in Oregon until the preparation of a legally adequate EIS. It retained jurisdiction to review the adequacy of such an EIS and awarded OEC attorneys’ fees under the EAJA for its work in Phase I.
*490Phase II began in March 1984 with the issuance of a new EIS (“the 1984 EIS”), which OEC challenged on numerous grounds. Two national environmental organizations, Friends of the Earth and the National Coalition Against the Misuse of Pesticides (collectively “FOE”), filed a complaint in intervention. Trial was scheduled for September 25, 1984, and a pretrial conference was set for August 31, 1984. On August 20, 1984, the federal defendants withdrew the contested 1984 EIS in order to supplement it. Because no final agency action then existed, the court dismissed OEC’s challenge to the 1984 EIS, but granted its request for injunctive relief based on the Secretary’s representation that no action would be taken until the EIS was supplemented. The court again retained jurisdiction to review the sufficiency of a final EIS.
Phase III commenced in March 1985 with the issuance of the final EIS, as supplemented (“the 1985 EIS”), which OEC and FOE again challenged. This EIS recommended “integrated pest management” to control gypsy moths, a program that permitted the use of chemical and/or biological insecticides. After trial, the district court found that the main text of the 1985 EIS was legally adequate, but that the worst case analysis in Appendix F violated 40 C.F.R. § 1502.8, which requires that an EIS be written in “plain language.” The court enjoined the use of carbaryl and three other chemical insecticides (trichlorfon, acephate, and diflubenzuron) in Oregon effective immediately and on a nationwide basis effective January 1, 1986. Oregon Environmental Council v. Kunzman, 614 F.Supp. 657, 665-66 (D.Or.1985) [hereinafter OEC II].
OEC and FOE sought attorneys’ fees under the EAJA for their work in Phases II and III. On October 4,1985, the district court denied fees for both phases, holding that in Phase II the plaintiffs were not “prevailing parties” under the EAJA and that in Phase III, although the plaintiffs were prevailing parties, the government's position was “substantially justified.” OEC and FOE both appealed (Nos. 85-4266, 85-4308).
Phase IV began in early 1986 with the issuance of an addendum to the 1985 EIS (“the 1986 Addendum”), which included a “plain language version” of the worst case analysis. In April 1986, the district court found that the worst case analysis met the readability and other requirements, and lifted the injunction. Oregon Environmental Council v. Kunzman, 636 F.Supp. 632, 640-41 (D.Or.1986) [hereinafter OEC III]. OEC and FOE appealed (No. 85-3779).
II. THE 1985 EIS AND THE 1986 ADDENDUM
A few words are in order on the documents that are the focal point of the latest round of this strenuously and well litigated controversy. The 1985 EIS includes an 8-page summary, a 78-page main text, lists of preparers and 145 references, an index, and a glossary defining 90 terms. Eight appendices include EPA reports on N-nitrosocarbaryl and carbaryl; data on the history of gypsy moth eradication; comments on the draft documents and responses; a 134-page worst case analysis (Appendix F); a 43-page “plain language version” of Appendix F (Appendix H); and further information on Appendices F and H (Appendix I).
The EIS’s statement of purpose and need for action describes the history of the gypsy moth in the United States and the attempts to control it. The gypsy moth, Lymantria dispar L., native to Europe, Asia, and Africa, was accidentally released in Massachusetts in 1869. From 1889, when an outbreak of gypsy moth larvae first threatened severe tree defoliation, until the late 1950s, state and federal agencies used a variety of compounds, including dichlorodiphenyl-trichloroethane (DDT), to control infestations in New England. With the phasing out of DDT in the 1960s and 1970s, new compounds were developed — carbaryl, trichlorfon, acephate, and diflubenzuron. Almost two million pounds of carbaryl were used in gypsy moth programs in the northeastern United States between 1962 and 1977.
*491Despite these efforts, the gypsy moth spread to the mid-Atlantic states by the 1970s and has appeared as far south as South Carolina, in certain mid-western states, and in California and Oregon. The EIS reports that gypsy moths defoliated 11,955,486 acres between 1924 and 1969, 11,640,705 acres between 1970 and 1979, and 29,425,328 acres between 1980 and 1984. It estimates that, as of 1980, economic losses to home owners, forest industries, and recreation areas have totaled $272 million.
The EIS evaluates four alternatives for the gypsy moth problem: (1) no action, (2) the four chemical insecticides, (3) the biological insecticide Bacillus tkuringiensis, to which appellants do not object, and (4) the selected alternative, integrated pest management, which permits the use of both chemical and biological insecticides. The EIS identifies mitigating measures and the affected environment — tree species susceptible to gypsy moths, nontarget organisms that the insecticides would affect,1 and geographical areas. The analysis of the environmental consequences concludes that realistic dose levels (and certain worst case dose exposures) for the four chemical insecticides are below the acceptable daily intake level and estimates that certain worst case risks of human cancer and heritable mutations range from the order of 1 in 100,000,000 to the order of 1 in 1,000,-000. The highest worst case cancer risk (for acephate) is on the order of 1 in 100,-000.
The 1985 EIS and 1986 Addendum have the outward indicia of comprehensiveness. See 40 C.F.R. § 1502.10 (1986) (setting forth EIS format). At trial several witnesses testified to the thoroughness of the documents; one stated that he had “never seen a greater effort made to identify all possible avenues of exposure and all reasonably possible exposure contingencies.” If the EIS is inadequate, it is not for lack of substantial effort on the part of its preparers. Yet OEC and FOE assign numerous errors in this lengthy document. We address their principal arguments.
III. JURISDICTION IN THE DISTRICT COURT
As a preliminary matter the Secretary argues that, because the State of Oregon has indicated an intention to forgo the use of chemical insecticides, OEC lacks standing. We disagree. The plaintiffs alleged numerous violations of NEPA’s procedural requirements. See Trustees for Alaska v. Hodel, 806 F.2d 1378, 1380 (9th Cir.1986) (holding that an environmental group had standing to challenge an alleged agency violation of NEPA’s procedural rights). Procedural failures in EIS preparation create a risk that environmental impacts will be overlooked and provide sufficient “injury in fact” to support standing. City of Davis v. Coleman, 521 F.2d 661, 670-71 (9th Cir.1975).
The Secretary argues that to allow anyone to bring an action to enforce compliance with NEPA would allow even individuals in states without gypsy moths to bring this action. This case does not present such a situation. In Coleman, we held that a plaintiff had standing to challenge the failure to prepare an EIS if he has “a sufficient geographical nexus to the site of the challenged project that he may be expected to suffer whatever environmental consequences the project may have.” Id. at 671. In this case, OEC’s members reside in a state with an actual gypsy moth problem and thus may challenge a nationwide EIS that is applicable to them.
This analysis comports with the traditional requirements for standing. Injury sufficient to meet the standing requirement may be actual or threatened, Heckler v. Mathews, 465 U.S. 728, 738, 104 S.Ct. 1387, 1394, 79 L.Ed.2d 646 (1984), and may involve a contingent risk of injury. 13 C. Wright, A. Miller & E. Cooper, Federal *492Practice and Procedure § 3531.4, at 434-37 (2d ed. 1984). OEC alleged contingent risks of injury that would result if Oregon decided to use chemical insecticides pursuant to the 1985 EIS and the 1986 Addendum.2 See Rockford League of Women Voters v. United States Nuclear Regulatory Comm’n, 679 F.2d 1218, 1221-22 (7th Cir.1982) (finding that plaintiffs had standing to challenge the issuance of a construction license for a nuclear plant even though plant operation was contingent on the future issuance of an operating license). Moreover, Oregon’s indication that it will use only Bacillus thuringiensis does not render the case moot because, as we found in Phase I, the decision to use chemical insecticides is “capable of repetition yet could evade review.” See OEC I, 714 F.2d at 903 (citing, inter alia, Southern Pac. Terminal Co. v. ICC, 219 U.S. 498, 515, 31 S.Ct. 279, 283, 55 L.Ed. 310 (1911)). The controversy is ripe because the agency’s action in issuing the EIS is clear and final. See Trustees for Alaska, 806 F.2d at 1381; 40 C.F.R. § 1500.3 (1986) (indicating that judicial review is appropriate after the filing of a final EIS). We conclude that the district court had jurisdiction to hear OEC’s challenges.
IV. SUBSTANTIVE CHALLENGES
The purpose of NEPA is to ensure that federal agencies are fully aware of the impact of their decisions on the environment. Friends of Endangered Species, Inc. v. Jantzen, 760 F.2d 976, 985 (9th Cir.1985). To carry out this purpose, NEPA requires that all federal agencies “include in ... major Federal actions significantly affecting the quality of the human environment, a detailed statement ... on ... the environmental impact of the proposed action.” 42 U.S.C. § 4332(C) (1982). The requirement of an EIS serves two ends. A properly prepared EIS ensures that federal agencies have sufficiently detailed information to decide whether to proceed with an action in light of potential environmental consequences, and it provides the public with information on the environmental impact of a proposed action and encourages public participation in the development of that information. Citizens for a Better Henderson v. Hodel, 768 F.2d 1051, 1056 (9th Cir.1985). The Council on Environmental Quality’s (“CEQ”) regulations govern the form, content, and preparation of an EIS. See 40 C.F.R. §§ 1500-1508 (1986).
A. Standard of Review
NEPA is essentially a procedural statute. Trustees for Alaska, 806 F.2d at 1382. The district court’s review of an EIS is principally governed by the Administrative Procedure Act, 5 U.S.C. § 706(2)(D) (1982), which provides that agency action may be set aside if undertaken “without observance of procedure required by law.” See Enos v. Marsh, 769 F.2d 1363, 1372 (9th Cir.1985); see also Sierra Club v. Sigler, 695 F.2d 957, 964 (5th Cir.1983) (noting that review also extends to whether the EIS was “arbitrary, capricious, an abuse of discretion, or otherwise not according to law” under § 706(2)(A)). The reviewing court may not substitute its judgment for that of the agency concerning the wisdom or prudence of a proposed action. California v. Block, 690 F.2d 753, 761 (9th Cir.1982).
This circuit employs a . “rule of reason” that asks whether an EIS contains a “ ‘reasonably thorough discussion of the significant aspects of the probable environmental consequences.’ ” Id. (quoting Trout Unlimited v. Morton, 509 F.2d 1276, 1283 (9th Cir.1974)). The district court must make “a pragmatic judgment whether the EIS’s form, content and preparation foster both informed decision-making and informed public participation.” Id. The reviewing court may not “ ‘fly speck’ ” an EIS and hold it insufficient on the basis of inconsequential, technical deficiencies. Northwest Indian Cemetery Protective Ass’n v. Peterson, 795 F.2d 688, 695 (9th *493Cir.1986) (quoting Lathan v. Brinegar, 506 F.2d 677, 693 (9th Cir.1974) (en banc)). But an EIS may be found inadequate under NEPA if it does not “ ‘reasonably [set] forth sufficient information to enable the decisionmaker to consider the environmental factors and make a reasoned decision.’ ” Id. (quoting Adler v. Lewis, 675 F.2d 1085, 1096 (9th Cir.1982)).
Appellants argue that the district court applied an erroneous standard of review when it characterized its task as one of determining “whether the agency made a ‘good faith effort’ to take environmental concerns into account.” OEC III, 636 F.Supp. at 634. They correctly indicate that this court has rejected a test of subjective good faith in favor of the objective test described above. Coalition for Canyon Preservation v. Bowers, 632 F.2d 774, 782 (9th Cir.1980). We do not perceive reversible error here, however, because we are satisfied that in describing and applying its standard of review the district court adhered to the principles set forth above. See OEC III, 636 F.Supp. at 634; OEC II, 614 F.Supp. at 660.
We review findings of fact underlying the district court’s decision for clear error. Northwest Indian Cemetery, 795 F.2d at 696. Interpretations of CEQ regulations present questions of law, which are reviewable de novo. See Sierra Club v. Union Oil Co. of Cal., 813 F.2d 1480, 1489 (9th Cir.1987); Felkner v. Dean Witter Reynolds, Inc., 800 F.2d 1466, 1468 (9th Cir.1986). The ultimate conclusion that an EIS is legally adequate under NEPA and the CEQ regulations presents an essentially legal question, which we review de novo. See United States v. McConney, 728 F.2d 1195, 1204-05 (9th Cir.) (en banc), cert. denied, 469 U.S. 824, 105 S.Ct. 101, 83 L.Ed.2d 46 (1984). Put another way, whether a particular deficiency, or combination of deficiencies, in an EIS is sufficient to warrant holding it legally inadequate, or constitutes merely a “fly speck,” is essentially a legal question, reviewable de novo.
B. The Readability or Understandability Requirement
In order to achieve the purpose of informing decisionmakers and the public of potential environmental consequences of a proposed agency action, the CEQ regulations require:
Environmental impact statements shall be written in plain language and may use appropriate graphics so that decision-makers and the public can readily understand them. Agencies should employ writers of clear prose or editors to write, review, or edit statements, which will be based upon the analysis and supporting data from the natural and social sciences and the environmental design arts.
40 C.F.R. § 1502.8 (1986); see also id. § 1500.2(b) (stating that an EIS “shall be concise, clear, and to the point”); id. § 1502.1 (same); id. § 1502.2(a) (stating that an EIS “shall be analytic rather than encyclopedic”); id. § 1500.4(e) (“clear format”); id. § 1502.10 (“clear presentation”). Appellants argue that neither the main text of the 1985 EIS nor the worst case analysis in the 1986 Addendum meets this requirement.
This circuit has not defined the level of readability that § 1502.8 requires. The district court stated that an EIS “must translate technical data into terms that render it an effective disclosure of the environmental impacts of a proposed project to all of its intended readership.” OEC II, 614 F.Supp. at 665. The scant case law indicates that an EIS must translate technical data into terms that effectively disclose environmental impacts to its “intended readership,” including “interested members of the public,” National Resources Defense Council, Inc. v. United States Nuclear Regulatory Comm’n, 685 F.2d 459, 487 n. 149 (D.C.Cir.1982), rev’d on other grounds sub nom. Baltimore Gas & Elec. Co. v. NRDC, 462 U.S. 87, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983), that an EIS should be written “in clear, concise, easily readable form so as to provide a reasonably intelligent non-professional an understanding of the environmental impact,” Warm Springs Dam Task Force v. Grib*494ble, 378 F.Supp. 240, 252 (N.D.Cal.1974), aff'd, 621 F.2d 1017 (9th Cir.1980); see also Upper West Fork River Watershed Ass’n v. Corps of Eng’rs, United States Army, 414 F.Supp. 908, 930 (N.D.W.Va.1976), aff'd mem., 556 F.2d 576 (4th Cir.1977), cert. denied, 434 U.S. 1010, 98 S.Ct. 720, 54 L.Ed.2d 752 (1978), and that an EIS must be “organized and written in language understandable to the general public and at the same time contain sufficient technical and scientific data to alert specialists to particular problems within their expertise,” Alabama ex rel. Baxley v. Corps of Eng’rs of the United States Army, 411 F.Supp. 1261, 1267 (N.D.Ala.1976); see also Sierra Club v. Froehlke, 359 F.Supp. 1289, 1342-43 (S.D.Tex.1973) (“‘understandable to non-technical minds’”), rev’d on other grounds sub nom. Sierra Club v. Callaway, 499 F.2d 982 (5th Cir.1974) (quoting Environmental Defense Fund, Inc. v. Corps of Eng’rs of the United States Army, 348 F.Supp. 916, 933 (N.D.Miss.1972), aff'd, 492 F.2d 1123 (5th Cir.1974)).
We hold that § 1502.8 imposes a requirement that an EIS must be organized and written so as to be readily understandable by governmental decisionmakers and by interested non-professional laypersons likely to be affected by actions taken under the EIS. The main text of an EIS will routinely include some scientific data and reasoning necessary to apprise decisionmakers and the public of potential environmental consequences. The more complicated the science underlying those consequences is, the more challenging the preparer’s task will be to convey the information clearly. Overly technical material and supporting data, however, should ordinarily appear in appendices. See 40 C.F.R. § 1502.18 (1986).
In this inquiry, we adhere to the general principle that the reviewing court must make a pragmatic judgment as to the adequacy of an EIS. If so inclined, the parties may introduce evidence concerning the reading level of the affected public and expert testimony on any indicia of the inherent readability of an EIS, as was done in this case. See OEC III, 636 F.Supp. at 634-36 (discussing expert testimony on the theoretical constructs of readability and on computer analyses of number of words per sentence and length of words). But the district court must also read the document. Its conclusion on whether an EIS is understandable under § 1502.8 is essentially a factual finding, which we review for clear error.
We have reviewed the evidence and read the 1985 EIS. The district court’s reasons for finding the main text of the 1985 EIS readable are given in OEC II, 614 F.Supp. at 665. We too find the main text clearly organized and sufficiently well written. The discussion of certain risks is at times dry and dense, but accessible to the interested layperson. We do not have a “‘definite and firm conviction’” that the district court committed a mistake. See United States v. Skokomish Indian Tribe, 764 F.2d 670, 673 (9th Cir.1985) (quoting Anderson v. Bessemer City, 470 U.S. 564, 573, 105 S.Ct. 1504, 1511, 84 L.Ed.2d 518 (1985)).
The worst case analysis in the 1986 Addendum presents an additional question. We agree with the district court that the readability requirement of § 1502.8 applies to the worst case analysis. When required, a worst case analysis is central to a full understanding of potential environmental impacts of proposed agency action. Clearly conveying that information to decisionmakers and to the public is as important as the identification of known adverse environmental impacts. As noted above, an EIS may include appendices to provide technical material that supports or amplifies a worst case analysis for the benefit of specialists, see 40 C.F.R. §§ 1502.18, 1502.24 (1986), and such supplemental technical material may be exempt from the readability requirement. But an agency may not circumvent its obligation to provide a clear assessment of environmental impacts simply by placing a worst case analysis in an appendix.
As with the main text of the 1985 EIS, the district court’s finding on the readability of the worst case analysis in Appendix H is not clearly erroneous. The district *495court’s reasoning amply supports its conclusion. See OEC III, 636 F.Supp. at 636-38. As one might expect, the material in Appendix H is less accessible than that in the main text, for the subject matter is more complex. As appellants argue, the materials might have been more accessible had it been presented in a single part of the EIS instead of in the main text and three appendices. But we are unpersuaded that the mode of presenting the worst case analysis rendered it unreadable in this case.3
C. Use of a Worst Case Analysis
The CEQ regulations provide that an EIS must include a worst case analysis when an agency is faced with incomplete or unavailable information. 40 C.F.R. § 1502.22 (1985).4 The regulation thus “requires disclosure and analysis of the ‘cost[s] of uncertainty — i.e., the costs of proceeding without more and better information.’ ” Southern Oregon Citizens Against Toxic Sprays, Inc. v. Clark, 720 F.2d 1475, 1478 (9th Cir.1983) [hereinafter SOCATS] (quoting Sigler, 695 F.2d at 970), cert. denied, 469 U.S. 1028, 105 S.Ct. 446, 83 L.Ed.2d 372 (1984); see Friends of Endangered Species, 760 F.2d at 988. In general, NEPA imposes a duty on federal agencies to gather information and do independent research when missing information is “important,” “significant,” or “essential” to a reasoned choice among alternatives. Save Our Ecosystems v. Clark, 747 F.2d 1240, 1244 n. 5, 1248-49 (9th Cir.1984). “Only if the costs are exorbitant or the means of obtaining the information is beyond the state of the art is the agency excused from compliance and allowed to perform a worst case analysis.” Id. at 1249. Reliance on a worst case analysis is justified if the costs of research are exorbitant “in light of the size of the project and/or the possible harm to the environment.” Id. at 1244 n. 5. Appellants argue that the Secretary violated § 1502.22 because the EIS does not establish either that the costs of obtaining the missing information are exorbitant or that the means are beyond state-of-the-art scientific research techniques.
The 1985 EIS justified its decision to forgo additional independent research and to prepare a worst case analysis principally on the ground that the costs would be exorbitant.5 It cites estimates ranging from tens of thousands to hundreds of thousands of dollars per study for each of several dozen studies that might be carried out. The EIS also indicates that completion of such studies would require several years, thus implicitly recognizing the harm to the environment that might be caused by infestations of gypsy moths during that period without chemical insecticides. The EIS, however, explains only that “[bjecause of the cost and time involved, a worst case analysis was done.”
Appellants correctly point out that the EIS does not explicitly evaluate the cost of the studies “in light of the size of the project and/or the possible harm to the environment.” A more thorough EIS might well have done so. We believe, however, that the EIS implicitly embodies such a calculus. Despite some evidence introduced by appellants that indicated that the costs of research might be somewhat lower, we believe that the district court did not clearly err in finding that the costs were in fact exorbitant. The omission of an explic*496it exposition of the agency’s determination of exorbitance does not warrant deeming this EIS legally inadequate. To do so in this case would impermissibly “fly speck” the EIS.6
D. Other NEPA Claims
Appellants charge that the EIS and worst case analysis are insufficient under NEPA on numerous other grounds. These challenges principally fall in two categories: (1) the EIS does not adequately disclose the cumulative effects associated with the program, and (2) the EIS manipulates and ignores data to conceal health risks.
We have considered the numerous contentions. Many concern the use of particular scientific methodologies, decisions to rely on certain scientific studies instead of others, and the use and application of Acceptable Daily Intake (ADI) and No Observable Effect Levels (NOELs) in various risk assessments. We are convinced that the district court’s treatment of these challenges is sufficient and does not warrant restating here. See OEC III, 636 F.Supp. at 636-40; OEC II, 614 F.Supp. at 661-64. We do observe, however, that “NEPA does not require that we decide whether an [EIS] is based on the best scientific methodology available, nor does NEPA require us to resolve disagreements among various scientists as to methodology.” Friends of Endangered Species, 760 F.2d at 986; see also 40 C.F.R. § 1502.24 (1986). “Our task is simply to ensure that the procedure followed by the [agency] resulted in a reasoned analysis of the evidence before it, and that the [agency] made the evidence available to all concerned.” Id. The use of methodologies, studies, and data in this EIS was not arbitrary or capricious.
E. Conclusion on the Merits
Although we have noted a few areas in which the 1985 EIS and 1986 Addendum might have been improved, we hold that under the “rule of reason” these documents contain a “reasonably thorough discussion of the significant aspects of the probable environmental consequences.” Trout Unlimited, 509 F.2d at 1283. Appellants may be dissatisfied with the EIS’s substantive recommendation to permit the use of the four chemical insecticides, in view of their assessment of the level and acceptability of the environmental risks. But if the EIS has met NEPA’s procedural requirements and is not arbitrary or capricious, “an informed and deliberate agency decisionmaker is entitled to be arguably ‘wrong.’” North Slope Borough v. Andrus, 642 F.2d 589, 599 (D.C.Cir.1980). We therefore affirm the district court’s decision on the substantive challenges.
V. ATTORNEYS’ FEES
The EAJA provides that a court shall award attorneys’ fees to a “prevailing party” in a civil action brought against the United States unless the government’s position was “substantially justified” or “special circumstances make an award unjust.” 28 U.S.C. § 2412(d)(1)(A) (Supp. III 1985). We review a decision denying attorneys’ fees under the EAJA for abuse of discretion. Merrell v. Block, 809 F.2d 639, 640 (9th Cir.1987). An abuse of discretion occurs if the district court based its decision on an erroneous legal conclusion or a clearly erroneous finding of fact. International Woodworkers of America, Local 3-98 v. Donovan, 792 F.2d 762, 765 (9th Cir.1985); In re Hill, 775 F.2d 1037, 1040 (9th Cir.1985). Interpretation of the EAJA is a question of law subject to de novo review. Merrell, 809 F.2d at 640. A finding that a party is not a prevailing party is one of fact that will be set aside if clearly erroneous or if based on an incorrect legal standard. See Lummi Indian Tribe v. Oltman, 720 F.2d 1124, 1125 (9th Cir.1983); see also *497McQuiston v. Marsh, 707 F.2d 1082, 1085 (9th Cir.1983) [hereinafter McQuiston I] (noting that the term “prevailing party” is interpreted consistently with the law under 42 U.S.C. § 1988).
A. Phase II: “Prevailing Party”
Appellants contend that the district court erred in holding that they were not prevailing parties in Phase II, which witnessed the Secretary’s withdrawal of the 1984 EIS several weeks prior to the scheduled trial. In its request for fees, appellants argued that “but for” its challenge, the 1984 EIS would not have been withdrawn and supplemented. The government countered that it withdrew the 1984 EIS because of “the availability of new information from various sources.” The district court concluded:
I find that although this litigation certainly had some effect on the government’s decision to supplement the 1984 EIS, it was not the cause for the supplementation. The government drafted a document it thought complied with NEPA standards. Thereafter, it received information from various sources outlining the deficiencies of the document and it then withdrew that EIS in order to correct those shortcomings. Thus, I find that plaintiffs are not prevailing parties for this portion of the litigation, and decline to award them attorney’s fees for phase II.
A party need not obtain formal relief on the merits to be deemed a prevailing party. McQuiston I, 707 F.2d at 1085. When the case ends before a final judgment is reached, a party may be deemed a prevailing party if it establishes ‘“some sort of clear, causal relationship between the litigation brought and the practical outcome realized.’” Greater Los Angeles Council on Deafness v. Community Television, 813 F.2d 217, 220 (9th Cir.1987) (quoting American Constitutional Party v. Munro, 650 F.2d 184, 188 (9th Cir.1981)). This standard was grounded on the pronouncements of other circuits:
The efforts of prevailing parties have been labelled variously as “at least a ‘contributing ... factor’ in the bringing about of [the desired] procedural changes,” although not the “sole cause”, as a “material factor in bringing about the defendant’s action”, and as contributing “in a significant way.”
American Constitutional Party, 650 F.2d at 187 (footnotes omitted). Hence, we have applied a test that inquires whether the suit was “at least a ‘material factor’ ” or played a “catalytic role” in bringing about the desired result. Id. at 188. “At a minimum, the lawsuit must have been a catalyst that prompted the opposing party to take action.” McQuiston I, 707 F.2d at 1085; accord McQuiston v. Marsh, 790 F.2d 798, 800 (9th Cir.1986) [hereinafter McQuiston II]; Harris v. McCarthy, 790 F.2d 753, 759 (9th Cir.1986); Rutherford v. Pitchess, 713 F.2d 1416, 1422 (9th Cir.1983).
The district court seems to have applied a legal standard that required a higher degree of causation, perhaps because of the way in which the appellants initially framed their position. Its decision does not discuss the appropriate legal standard of whether the litigation was a “material factor” or played a “catalytic role” in the Secretary’s action. Cf. McQuiston II, 790 F.2d at 801 (upholding a finding that the plaintiff was not a prevailing party when the district court “specifically found that no causal nexus existed between the ... lawsuit and the [government’s] action, and that his lawsuit did not prompt [the action]”). The court’s finding that “this litigation certainly had some effect on the government’s decision” suggests that, had the correct legal standard been applied, a finding that the appellants were prevailing parties would have been warranted.7 See Pitchess, 713 F.2d at 1419-22 (reversing a denial of attorneys’ fees when the district court’s findings acknowledged that the pendency of the proceedings resulted in improvements in the defendant’s conduct *498and the litigation had “an impact on encouraging the defendant’s good faith efforts”).
Although it appears that the district court most likely did not apply the material factor/catalytic role test, we pretermit our discussion of this issue. Even if we were to conclude that the appellants were prevailing parties, we would have to remand to determine whether the government’s position on the adequacy of the 1984 EIS was substantially justified. See Lummi Indian Tribe, 720 F.2d at 1126. The prudent course to take with this ambiguously resolved attorneys’ fees issue is to vacate this portion of the district court’s decision. Cf. id. at 1125 (reversing a decision denying attorneys’ fees because the district court plainly applied an incorrect standard for determining prevailing parties). On remand, the district court should clarify whether it in fact employed the material factor/catalytic role standard. If not, it should make the required factfinding applying the correct standard and, if necessary, decide whether the government’s position was substantially justified and determine the amount of any award by assessing the extent of OEC’s and FOE’s success. Id.
B. Phase III: “Substantially Justified’’
The district court held that the appellants were prevailing parties in Phase III because the worst case analysis in Appendix F did not satisfy the readability requirement. It denied attorneys’ fees, however, on the ground that the government’s position was substantially justified. The district court must consider the totality of the circumstances present prior to and during litigation. Hill, 775 F.2d at 1042. That the government lost does not raise a presumption that its position was not substantially justified. Id. The government bears the burden of showing that its position was substantially justified in law and in fact. See id.
This court has long applied a test of reasonableness in determining whether the government’s position was substantially justified. League of Women Voters v. FCC, 798 F.2d 1255, 1257 (9th Cir.1986); Foster v. Tourtellotte, 704 F.2d 1109, 1112 (9th Cir.1983) (per curiam). Appellants ask us to determine the precise level of reasonableness required in light of a House of Representatives committee report accompanying the 1985 EAJA amendments:
Several courts have held correctly that “substantial justification” means more than merely reasonable. Because in 1980 Congress rejected a standard of “reasonably justified” in favor of “substantially justified,” the test must be more than mere reasonableness.
H.R.Rep. No. 120, 99th Cong., 1st Sess. 9, reprinted in 1985 U.S. Code Cong. & Admin. News 132, 138 (footnote omitted); see also Spencer v. NLRB, 712 F.2d 539, 558 (D.C.Cir.1983) (stating that this rejection suggests a “slightly more stringent” standard than one of reasonableness), cert. denied, 466 U.S. 936, 104 S.Ct. 1908, 80 L.Ed.2d 457 (1984). The Secretary counters that the legislative history of the 1985 amendments is not so clear-cut. See Russell v. National Mediation Bd., 775 F.2d 1284, 1289 (5th Cir.1985) (finding the 1985 EAJA legislative history “puzzling,” “conflicting and inconclusive,” and resolving to adhere to its previous standard of reasonableness). Although this court has recently reiterated its adherence to the general test of reasonableness, it did not address the effect of the 1985 amendments. See Minor v. United States, 797 F.2d 738, 739 (9th Cir.1986) (relying on the 1980 legislative history). This case, however, does not require us to determine whether more than mere reasonableness is required, because the government clearly meets a higher standard.
The district court stated that the defendant’s position on the adequacy of the worst case analysis was substantially justified because “no court had previously invalidated an EIS on the[] grounds” that it did not satisfy § 1502.8. We do not affirm on this broad ground. Although the absence of adverse precedent on an issue is relevant to the determination of the “substantially justified” question, it is not dis-positive. Hill, 775 F.2d at 1042 (stating *499that the government may sustain its burden by showing “its position is ‘a novel but credible exension or interpretation of the law’ ”) (quoting Hoang Ha v. Schweiker, 707 F.2d 1104, 1106 (9th Cir.1983)); see also League of Women Voters, 798 F.2d at 1260 (involving an issue on which “reasonable minds could differ”); Minor, 797 F.2d at 739 (involving an “important and doubtful question of tax law”). The CEQ regulations unambiguously require an EIS to be “written in plain language” “readily understandable]” to decisionmakers and the public. 40 C.F.R. 1502.8 (1986). The EIS must also be “concise, clear, and to the point.” Id. §§ 1500.2(b), 1502.1. Several cases in other jurisdictions have demonstrated adherence to the understandability requirement. See supra Part IV.B. The position that an EIS need not comply with a clear CEQ regulation would be novel, but it is not a credible extension or interpretation of the law. See Hill, 775 F.2d at 1042.
We find merit in a narrower position. We do not decide whether the Secretary’s position that § 1502.8 did not apply to a worst case analysis was substantially justified. Cf. SOCATS, 720 F.2d at 1481 (holding that the worst case requirement of § 1500.22 “is straightforward and means what it says” but affirming the denial of fees when there was no precedent holding that the regulation was applicable to an EA). We have no hesitation in holding that, in view of the previous uncertainty in the level of readability that § 1502.8 requires and some expert testimony that the document was understandable, the Secretary’s position that the worst case analysis satisfied § 1502.8 was substantially justified as a matter of law and fact. The Secretary’s position in Phase III was well argued and justified, though erroneous. See Trustees for Alaska, 806 F.2d at 1384. We therefore affirm the denial of attorneys’ fees for Phase III.
CONCLUSION
The appeal on the merits in No. 86-3779 is affirmed. The appeal on the attorneys’ fees in Nos. 85-4266 and 85-4308 is affirmed in part and vacated and remanded in part. The parties shall bear their own costs on this appeal.
AFFIRMED IN PART, VACATED AND REMANDED IN PART.
. The EIS states that carbaryl is toxic to honeybees, some aquatic insects, and shellfish; that trichlorfon is toxic to flies; that acephate is toxic to honeybees; and that diflubenzuron is toxic to some aquatic organisms. It further states that the expected doses of these pesticides should not adversely affect fish, wildlife, livestock, or domestic animals.
. Members of the nationally based Friends of the Earth and the National Coalition Against the Misuse of Pesticides reside in states in which chemical spraying has been proposed or conducted.
. The text of the EIS did not, as appellants suggest, exceed the length limitations. See 40 C.F.R. § 1502.7 (1986).
. The applicable regulation provides in pertinent part:
When an agency is evaluating significant adverse effects on the human environment in an environmental impact statement and there are gaps in relevant information or scientific uncertainty, the agency shall always make clear that such information is lacking or that uncertainty exists.
(a) If the information relevant to adverse impacts is essential to a reasoned choice among alternatives and is not known and the overall costs of obtaining it are not exorbitant, the agency shall include the information in the environmental impact statement.
40 C.F.R. § 1502.22(a) (1985) (amended 1986).
. At tried, the Secretary justified forgoing research on one issue, the risks from heritable human mutations, as beyond the state of the art. Appellants do not contest this assertion.
. Appellants also argue that the failure to list the source of the information on the costs of such studies (a personal communication with an employee of Dow Chemical) violated 40 C.F.R. § 1502.21 (1986), which provides that an EIS may not incorporate material "unless it is reasonably available for inspection by potentially interested persons within the time allowed for comment.” In the face of the comprehensiveness of the EIS, however, we believe that the omission is a “fly speck.”
. Counsel for the government conceded at oral argument that the government was "not oblivious” to the litigation and that, in view of several recent adverse decisions on herbicide spraying in this court, the government was being “cautious.”