This appeal (under 28 U.S.C. § 1292(a)(1)) by the Justice Department from an order enjoining it from putting into effect in Chicago a provision of the Department’s new guidelines for FBI investigations requires us to consider the proper standards for interpreting an equity decree that restricts the executive branch of the federal government in the performance of its constitutional responsibilities.
The suit out of which the order under review arose was brought in 1973 and charged, so far as is pertinent here, that the FBI’s Chicago office had in the name of domestic security conducted investigations (some protracted) of the plaintiffs, among them the American Civil Liberties Union, the National Association for the Advancement of Colored People, the Independent Voters of Illinois, the Better Government Association, the American Friends Service Committee, and a Congressman. The complaint alleged that the plaintiffs posed no actual or potential threat to domestic security, that the motivation for the investigations was the FBI’s dislike of their political views, and that the goal and consequence of the investigations were to harass and intimidate the plaintiffs. These allegations repair the deficiencies that led the Supreme Court to hold in Laird v. Tatum, 408 U.S. 1, 92 S.Ct. 2318, 33 L.Ed.2d 154 (1972), that a challenge to the Army’s domestic surveillance activities was not justiciable because the surveillance was not alleged to have *1010affected the plaintiffs’ conduct in any way. See id. at 13 and n. 7, 92 S.Ct. at 2325 and n. 7; see also United Presbyterian Church v. Reagan, 738 F.2d 1375 at 1378-1381 (D.C.Cir.1984).
In 1976 the Department of Justice published guidelines for FBI investigations (the “Levi Guidelines,” after the then Attorney General). The guidelines led to a sharp reduction in the number of domestic security investigations and set the stage for the settlement of the suit. The parties signed a settlement agreement in the fall of 1980. The agreement runs to 31 double-spaced typewritten pages and includes among other extraordinary provisions a section that gives standing not only to the plaintiffs but to every resident of Chicago at the time of the settlement to enforce the decree by application to the district court, and a section that empowers the plaintiffs’ representatives to monitor compliance with the decree for five years with the help of periodic reports that the decree requires the Justice Department to furnish them. The provision of particular importance in this enforcement proceeding, however, is paragraph 3.4(a), which provides (we italicize the portion most pertinent to this appeal):
The parties agree that the following general principles apply to FBI activities relating to the domestic activities of United States persons:
(a) The FBI, in conducting domestic security investigations and inquiries, shall be concerned only with conduct and only such conduct as is forbidden by a criminal law of the United States, or by a state criminal law when authorized by federal'statute. The FBI shall not conduct an investigation solely on the basis of activities protected by the First Amendment of the Constitution of the United States, or on the lawful exercise of any right secured by the Constitution or laws of the United States.
After the usual comment period, Judge Getzendanner in August 1981 approved a consent decree containing the settlement agreement. See Alliance to End Repression v. City of Chicago, 91 F.R.D. 182 (N.D.Ill.1981).
Eighteen months later Attorney General Smith announced new guidelines for FBI investigations (“The Attorney General’s Guidelines on General Crimes, Racketeering Enterprise and Domestic Security/Terrorism Investigations,” March 7, 1983), superseding the Levi Guidelines of 1976. This too is a formidable document — 19 single-spaced typewritten pages. In Part I (“General Principles”) the following appears (again we italicize the key portion):
In its efforts to anticipate or prevent crime, the FBI must at times initiate investigations in advance of criminal conduct. It is important that such investigations not be based solely on activities protected by the First Amendment or on the lawful exercise of any other rights secured by the Constitution or laws of the United States. When, however, statements advocate criminal activity or indicate an apparent intent to engage in crime, particularly crimes of violence, an investigation under these Guidelines may be warranted unless it is apparent, from the circumstances or the context in which the statements are made, that there is no prospect of harm.
The plaintiffs applied to Judge Getzendanner for an injunction against the Smith Guidelines, charging that they were inconsistent with the consent decree in a variety of respects. She agreed to the extent of enjoining the last sentence quoted above (actually only the part of the sentence that we have italicized). 561 F.Supp. 575, 582-83 (N.D.Ill.1983). This appeal followed. The panel of this court that first heard the appeal modified the district court’s decree to change the injunction to a declaratory judgment and affirmed the decree as so modified (one judge, however, would have reversed). 733 F.2d 1187 (1984). The full court granted rehearing en banc because of the sensitive issue of separation of powers raised by a decision invalidating in a major city a part of the FBI’s nationwide investigatory guidelines.
*1011The task of interpreting the decree in this case is complicated by the absence of a concrete factual context. Usually a question about the meaning of a consent decree arises because a party does, or proposes, an activity that may violate the decree or fails to undertake an activity required by the decree; and the task is then to apply the decree to the activity. There is no activity here. As far as we know, the FBI has yet to conduct or even propose a specific investigation that might run afoul of the decree; it only wants to be allowed to tell its agents in Chicago to follow the new guidelines that the Justice Department has issued to govern FBI investigations nationwide. Although the district judge had the power to interpret the decree even in this rather bare setting, cf. EEOC v. Safeway Stores, Inc., 611 F.2d 795, 798 (10th Cir.1979); Note, Implementation Problems in Institutional Reform Litigation, 91 Harv.L.Rev. 428, 440-48 (1977), the bareness of the setting requires us to resolve all reasonable doubts in favor of an interpretation of the guidelines that will avoid conflict with the decree. We do not lightly assume in advance of anything more ominous than a general statement of enforcement policy intended to be applicable nationwide that the Justice Department is trying to evade a judicial decree.
But that is what the plaintiffs are asking us to assume in pressing on us the following syllogism: Paragraph 3.4(a) of the decree forbids the FBI to “conduct an investigation solely on the basis of activities protected by the First Amendment”; the government may not suppress or punish statements advocating criminal activity unless they pose an immediate and substantial danger to the public safety (see, e.g., Brandenburg v. Ohio, 395 U.S. 444, 447, 89 S.Ct. 1827, 1829, 23 L.Ed.2d 430 (1969) (per curiam)); therefore the Smith Guidelines, in permitting investigations based on “statements [that] advocate criminal activity” without requiring that the statements be shown to pose an immediate and substantial danger, contravene paragraph 3.4(a). Reminding us that a consent decree “is to be construed for enforcement purposes basically as a contract,” United States v. ITT Continental Baking Co., 420 U.S. 223, 238, 95 S.Ct. 926, 935, 43 L.Ed.2d 148 (1975), and that “the scope of a consent decree must be discerned within its four corners,” United States v. Armour & Co., 402 U.S. 673, 682, 91 S.Ct. 1752, 1757, 29 L.Ed.2d 256 (1971); see also Firefighters Local Union No. 1784 v. Stotts, — U.S. —, 104 S.Ct. 2576, 2586, 81 L.Ed.2d 483 (1984), the plaintiffs argue that the quoted language from paragraph 3.4(a) is clear, and clearly inconsistent with the quoted language from the Smith Guidelines, and therefore that the injunction was proper.
We agree that our function is interpretation, and also that the words of paragraph 3.4(a) will bear the plaintiffs’ interpretation of them. But the question whether the interpretation is compelled is less clear. In searching for an answer we are of course not confined to the language of paragraph 3.4(a). The relevant “four corners” are those of the decree, not of one provision of the decree. White v. Roughton, 689 F.2d 118, 119 (7th Cir.1982). Like any document, a consent decree must be read as a whole. Cf. United States v. Morton, — U.S. —, —, 104 S.Ct. 2769, 81 L.Ed.2d 680 (1984) (“We do not, however, construe statutory phrases in isolation; we read statutes as a whole”); United States v. American Trucking Ass’ns., Inc., 310 U.S. 534, 543-44, 60 S.Ct. 1059, 1063-64, 84 L.Ed. 1345 (1940). The plaintiffs concede as much by citing in support of their interpretation of paragraph 3.4(a) paragraph 2.2(2), which states that “the [Levi] Guidelines are not intended to permit domestic security investigations of groups which advocate the necessity for violent revolution at some time in the indefinite future, but which do not now engage in serious crimes or violence or advocate imminent serious crime or violence.” Notice, however, that this does not say “the consent decree is not intended to permit____” It says the Levi Guidelines are not intended to permit, and they are not part of the decree. Paragraph 2.2(2) is not even in the consent agreement, which begins with Part *1012III. When Judge Getzendanner set forth what she called “the pertinent provisions of the FBI settlement agreement” in an appendix to her opinion approving the consent decree, she did not include paragraph 2. See 91 F.R.D. at 205-06.
Paragraph 2.2(2) appears in a statement of the reasons why the plaintiffs decided to settle the lawsuit: “the [Levi] Guidelines and other post-filing laws and procedures governing the FBI afford a reasonable basis upon which to settle this litigation. In arriving at this conclusion, plaintiffs take particular note of the following ____” H 2.2. In the same vein is the statement in the decree that the parties signed the agreement on the basis of “statutes, guidelines and other procedures referred to in” the decree, including the Levi Guidelines, and on the “understanding of them reflected in 112.2.” The Levi Guidelines and the other restrictions that had been placed on the FBI persuaded the plaintiffs that the defendants were acting in sufficiently good faith to warrant settling the litigation in advance of trial. The plaintiffs’ reasons for settling were relevant because the suit was a class action and the district judge therefore had to approve the settlement; and indeed some members of the class objected to the settlement that the plaintiffs’ lawyers had negotiated with the Justice Department. To read paragraph 2.2 as if it were a part of the agreement rather than a statement of the plaintiffs’ reasons for agreeing would thus neglect the balance of the passage quoted earlier from Armour: “the scope of a consent decree must be discerned within its four corners, and not by reference to what might satisfy the purposes of one of the parties to it.” 402 U.S. at 682, 91 S.Ct. at 1757. It would also neglect the fact that paragraph 3.6 of the decree envisages and accepts the possibility that the Levi Guidelines would be superseded; the plaintiffs assumed the risk that a subsequent set of guidelines might not be as much to their liking as the Levi Guidelines were. And it would neglect some pertinent remarks by Judge Getzendanner when she approved the decree: “The parties have candidly admitted that they would not necessarily agree on all possible applications of the language of the principles [in the decree] in the abstract or in hypothetical cases.” 91 F.R.D. at 200. The plaintiffs may not force their unilateral interpretation on the defendants.
But in any event we are far from clear that there are substantive, as distinct from verbal, inconsistencies between the Levi Guidelines and the pertinent sentences in the Smith Guidelines. See Elliff, The Attorney General’s Guidelines for FBI Investigations, 69 Cornell L.Rev. 785, 798-99 (1984). The Levi Guidelines authorize a “full investigation” when there is reason to believe that a group “may be engaged in activities which involve the use of force or violence and which involve or will involve the violation of federal law ____” The reason to believe might be a statement that could not itself be made the basis of a prosecution without violating the First Amendment and that did not even indicate that any law was being violated. Moreover, the Levi Guidelines, we are authoritatively told, would authorize a “preliminary investigation” on the basis of so tepid a statement, delivered by a group with no known propensity for violence, as “The rulers have set the time for the party; let us bring the fireworks.” FBI Oversight, Hearings before the Subcomm. on Civil and Constitutional Rights of H. Comm, on Judiciary, 94th Cong., 1st and 2d Sess., ser. no. 2, pt. 3, at 268-69 (1976) (testimony of Attorney General Levi). And we must attend very carefully to the precise wording of paragraph 2.2(2). The domestic security investigations that the Levi Guidelines are said to forbid are investigations “of groups which advocate the necessity for violent revolution at some time in the indefinite future____” This is a veiled but unmistakable reference to those Marxist groups that are known to preach the inevitability of the proletarian revolution as a tenet of ideology rather than as a practical goal. It need not be read to forbid investigations of groups that may be planning to turn to *1013violence in pursuit of their aims, though not in the immediate future.
Although we do not think that paragraph 2.2 carries the day for the plaintiffs, we do agree with them, as we have said, that in interpreting paragraph 3.4(a) we are not limited to the particular words being interpreted. It is a great fallacy to think that by staring hard at an isolated sentence one can come up with a meaningful interpretation. The sentence may look clear and yet if one understood its background and context one might read it quite differently from its superficially clear meaning. To take an example far removed from law, almost everyone today thinks that “a custom more honored in the breach than the observance” means a custom that is not observed. That is what the expression viewed in isolation seems plainly to mean. But if you go back to the passage in Hamlet from which the expression comes (Act I, sc. iv, lines 8-20), you will see that the custom referred to is that of getting drunk on festive occasions, and that what “a custom more honored in the breach ...” actually means is a custom better disregarded than observed. The Complete Works of Shakespeare 1081 nn. 15-16 (Bevington ed., 3d ed. 1980).
The point is general: context, in the broadest sense, is the key to understanding language. That is why the Supreme Court in the Continental Baking case on which the plaintiffs rely refused “to ignore the flexibility of the English language, as well as the circumstances surrounding the [consent] order and the context in which the parties were operating.” 420 U.S. at 243, 95 S.Ct. at 938. It is why we held in White v. Roughton — like this a case requiring interpretation of a decree limiting a government agency’s freedom of action — that “it is permissible to construe unambiguous language where construction is necessary to determine the intended rather than the literal meaning of the decree.” 689 F.2d at 120. For “the overriding purpose in construing a contract is to give effect to the mutual intent of the parties at the time the contract was made; and while the language of the contract is normally the best evidence of that intent, a court can properly disregard even unambiguous language when it is convinced that the parties meant something different from what they said.” Id., citing 2 Restatement (Second) of Contracts, § 202, comment c; § 212, comment b (1981); see also Farnsworth, Contracts § 7.10, at pp. 492-93 (1982).
One thing to consider is the consequences of alternative interpretations. Suppose that the result of reading a contract in a particular way is that one of the parties assumed enormous risks and got nothing in return; this would argue against the reading. Cf. Fidelity & Deposit Co. v. City of Sheboygan Falls, 713 F.2d 1261, 1270 (7th Cir.1983). This principle of interpretation has particular relevance when a court is construing a consent decree regulating a public institution, rather than an ordinary contract (and we must keep in mind that such a consent decree “is no mere contract, even though reference to contract principles may be useful,” New York State Ass’n for Retarded Children, Inc. v. Carey, 596 F.2d 27, 37 (2d Cir. 1979)). The consequences of alternative interpretations may be much greater than in the case of an ordinary contract; and naturally a court will hesitate to assume that by signing a consent decree the government knowingly bartered away important public interests merely to avoid the expense of a trial. In White v. Roughton, for example, where the issue was whether a township had agreed to make a “commitment to a fixed level of benefits paid out of its tax revenues for the indefinite future [that] could seriously imperil its fiscal health,” we declined to read the decree literally, because it would have meant that “the township [had] made just such an improvident commitment,” id. at 121, and we were reluctant to assume that the township had discharged its duty to the public improvidently. Cf. Firefighters Local Union No. 1784 v. Stotts, supra, 104 S.Ct. at 2585.
*1014The panel majority in this case acknowledged that under its reading of the consent decree the Justice Department may have acted “improvidently” in signing the decree, 733 F.2d at 1191, and this acknowledgment gives us pause, especially as the degree of improvidence would be altogether greater than in the White case. This is because the First Amendment, as currently interpreted, places tight limits on the government’s authority to punish those who counsel or advocate violence. The Supreme Court’s decisions from Schenck v. United States, 249 U.S. 47, 39 S.Ct. 247, 63 L.Ed. 470 (1919), to Brandenburg v. Ohio, supra, describe a circle, ending where they began with a test of incitement. Compare 249 U.S. at 52, 39 S.Ct. at 249 (“The most stringent protection of free speech would not protect a man in falsely shouting fire in a theater, and causing a panic____ The question in every case is whether the words used are used in such circumstances and are of such a nature as to create a clear and present danger that they will bring about the substantive evils that Congress has a right to prevent”), with 395 U.S. at 447, 89 S.Ct. at 1829 (a state may forbid only “advocacy ... directed to inciting or producing imminent lawless action and ... likely to incite or produce such action”). If, therefore, a new sect of religious fanatics announced that unless Chicagoans renounce their sinful ways it may become necessary to poison the city’s water supply, or a newly organized group of white supremacists vowed to take revenge on Chicago for electing a black mayor, these statements, made by groups with no “track record” of violent acts, might well be privileged. See, e.g., NAACP v. Claiborne Hardware Co., 458 U.S. 886, 927, 102 S.Ct. 3409, 3433, 73 L.Ed.2d 1215 (1982) (holding that the First Amendment protected “references to the possibility that necks would be broken”); Noto v. United States, 367 U.S. 290, 296, 298, 81 S.Ct. 1517, 1520, 1521, 6 L.Ed.2d 836 (1961) (same, for the statements “ ‘he was the kind of a guy they hoped to shoot some day’ ” and “ ‘Sometime I will see the time we can stand a person like this S.O.B. against the wall and shoot him,’ ” and for other “off hand remarks that certain individuals hostile to the [Communist] Party would one day be shot”). Or suppose the leaders of a newly formed organization of Puerto Rican separatists went around Chicago making speeches to the effect that, if the United States does not grant Puerto Rico independence soon, it will be necessary to begin terrorist activities on the mainland United States. These speeches could not, in all probability, be made the basis of a prosecution. Therefore, under the interpretation of the consent decree urged by the plaintiffs, whereby the Justice Department would have no power to investigate anything that cannot be punished, the FBI probably could not even investigate any of these hypothetical groups. And since the sentence in the Smith Guidelines that the district court enjoined applies to all investigations, not just investigations of groups, many threats made by individuals against other individuals could not be investigated either, for such threats, too, enjoy broad protection under the First Amendment, as we know from cases like Watts v. United States, 394 U.S. 705, 706, 89 S.Ct. 1399, 1400, 22 L.Ed.2d 664 (1969) (per curiam) (“If they ever make me carry a rifle the first man I want to get in my sights is L.B.J.”).
We doubt that in agreeing to the consent decree the Justice Department tied its hands to such an extent; for if it did, it was trifling with the public safety of the people of Chicago and maybe even violating the President's constitutional obligation to “take Care that the Laws be faithfully executed.” Art. II, § 3, cl. 3. (A variety of federal laws forbid terrorist acts: e.g., 18 U.S.C. §§ 13, 231, 351, 372, 871, 875, 878, 1111, 1112, 1113, 1114, 1361, 1751, 2101, 2113, 2383, 2384; 49 U.S.C. § 1472.) The Levi Guidelines had not gone so far (and this is another reason why we find no substantive inconsistency between them and the Smith Guidelines, at least so far as the issue in this appeal is concerned). The FBI always has investigated people who advocate or threaten to commit serious vio*1015lations of federal law, even if the violations are not imminent; and it always will. It “has a right, indeed a duty, to keep itself informed with respect to the possible commission of crimes; it is not obliged to wear blinders until it may be too late for prevention.” Socialist Workers Party v. Attorney General, 510 F.2d 253, 256 (2d Cir. 1974) (per curiam).
And what did the Department get in return for abandoning that duty, if that is what it did? By the time the case reached the settlement stage all that the plaintiffs were seeking was an injunction, and they would not have gotten a broader injunction if they had pressed the case to trial and won than they got in the consent decree. Equity decrees may “fence in” the violator, see, e.g., FTC v. National Lead Co., 352 U.S. 419, 429-30, 77 S.Ct. 502, 509-10, 1 L.Ed.2d 438 (1957), but there are limits beyond which such a decree could not properly tie the hands of the FBI in investigating terrorist activities. The decree as it stands is a remarkable judicial intervention in vital executive functions; a proper decree formulated after a trial would not have been more Draconian. Cf. Judge Getzendanner’s remarks in 91 F.R.D. at 197-98. Thus, if the plaintiffs are right in their reading of the decree, the Justice Department bargained away some of its essential investigative powers and got nothing in return but a saving of some litigation expenses. It did not even save “face”; for though the consent decree contains a disclaimer of any admission of liability by the federal defendants, it admits 500 “black bag jobs” by them, and says they “resulted in the obtaining of information concerning lawful political activities of the persons or organizations targeted,” and in the disruption and harassment of “the lawful activities of some United States persons in the City of Chicago,” “much of that activity [having] no statutory basis or justification.”
We shall see in a moment that the plaintiffs, no longer willing to defend a position that would make consequences entirely irrelevant to interpretation of the consent decree, argue that the decree allows the government to make “preliminary inquiries” based on statements that could not be made the subject of a prosecution. And we shall also see that there is no basis in the decree for such an exception. The plaintiffs are willing to condone reparative surgery to save the decree from an absurd interpretation, but they reject the alternative: an interpretation of the decree with some basis in the language—a free but not a reckless interpretation—that avoids imputing to the Justice Department indifference to its responsibilities to the public. This interpretation is that the FBI may not base an investigation solely on the political views of a group or an individual; it must have a basis in a genuine concern for law enforcement. Thus it may not investigate a group solely because the group advocates Puerto Rican independence, a cause that most of the men and women of the FBI no doubt have little fondness for; but it may investigate any group that advocates the commission, even if not immediately, of terrorist acts in violation of federal law. It need not wait till the bombs begin to go off, or even till the bomb factory is found.
We are not speaking metaphorically. Between 1970 and 1980, domestic terrorist organizations committed more than 400 bombings in the United States. (Computed from Motley, US Strategy to Counter Domestic Political Terrorism 16 (1983) (table 1.1).) The FBI cannot hope to nip terrorist conspiracies in the bud if it may not investigate proto-terrorist organizations. That is why, as cases such as Zurcher v. Stanford Daily, 436 U.S. 547, 563-67, 98 S.Ct. 1970, 1980-82, 56 L.Ed.2d 525 (1978); Heller v. New York, 413 U.S. 483, 488-93, 93 S.Ct. 2789, 2792-95, 37 L.Ed.2d 745 (1973); Branzburg v. Hayes, 408 U.S. 665, 693-95, 92 S.Ct. 2646, 33 L.Ed.2d 626 (1972), and Reporters Comm, for Freedom of Press v. American Tel. & Tel. Co., 593 F.2d 1030, 1052 (D.C.Cir.1978), make clear, the FBI would not be violating the First Amendment itself (as distinct from the plaintiffs’ conception of the consent decree) if it decided to investigate a threat that was not so immediate as to permit punitive *1016measures against the utterer. Since the repressive effect of an investigation is less than that of a prosecution but the benefits in preventing violent crime may be as great, a less immediate danger will justify the government’s action. Cf. Socialist Workers Party v. Attorney General, supra, 510 F.2d at 256; Handschu v. Special Services Division, 849 F.Supp. 766, 771 (S.D.N.Y.1972) (Weinfeld, J.).
Admittedly the repressive effect will not be zero. No one wants his name in an FBI investigatory file; and the knowledge that the FBI investigates groups that advocate violent change could deter some people from joining such groups and deter the groups themselves from engaging in lawful though minatory forms of advocacy. There would therefore be a cost to the values protected by the First Amendment, if the groups never stepped over the boundary that separates privileged from indictable speech. But we think the cost would be outweighed by the benefits in preventing crimes of violence, provided that the FBI did not prolong its investigation after it became clear that the only menace of a group under investigation was rhetorical and ideological.
The plaintiffs argue, however, that the consent decree went further than the First Amendment — it must have, they say, or they got nothing in return for agreeing to settle the case rather than litigating it. Alternatively they argue that the sentence in paragraph 3.4(a) on which the district judge based the injunction would be redundant if it merely commanded the FBI to do what the Constitution commands it to do— indeed, what the preceding sentence commands. In either form the argument overlooks three points. The first is that, much lore to the contrary notwithstanding (see, e.g., Pierce v. MacNeal Memorial Hospital Ass’n, 46 Ill.App.3d 42, 54, 4 Ill.Dec. 615, 624-25, 360 N.E.2d 551, 559-60 (1977)), not every sentence in a negotiated document can be assumed to have independent operative significance. White v. Roughton, supra, 689 F.2d at 120. Paragraph 3.4(a) apparently emerged from a negotiating session. We do not know who proposed it, or how much forethought was given to it. The second sentence may have been included not as the plaintiffs conjecture to put the FBI under a different duty from the first (“The FBI, in conducting domestic security investigations and inquiries', shall be concerned only with conduct ... ”) but to ensure that the first would not be misunderstood — in other words, to emphasize by repetition.
Second, since if the case had gone to trial and the plaintiffs had won the district judge would almost certainly not have gone further than to enjoin the FBI from violating the Constitution, it is unlikely that the Justice Department agreed to go further in the consent settlement. A rational litigant does not give up in a settlement more than he can expect to lose at trial. True, by settling on any terms the Department avoided the expenses of trial and the slight probability of a decree that would hem it in more than the consent decree; but knowing that the plaintiffs too wanted to avoid the expense of litigation and the possibility that they would lose at trial or obtain an unsatisfactory decree, the Department could not have thought it must make all the concessions in the settlement negotiations. And yet on the plaintiffs’ reading of the decree they gave up nothing to settle the case and the Department gave up everything.
Third, even if the consent decree did no more than enjoin the FBI from violating the Constitution, the plaintiffs won a good deal — and we do not refer just to their symbolic victory. As a result of the decree, if the FBI now conducts an investigation in Chicago that violates the Constitution it will be violating a judicial decree, and will be subject to the summary and severe sanctions, criminal as well as civil, that are visited on government agencies and agents as on others who willfully violate federal court decrees. See 18 U.S.C. §§ 401, 402; Fed.R.Crim.P. 42; In re Attorney General, 596 F.2d 58, 64 (2d Cir.1979); Ex parte Craig, 282 Fed. 138, 153 (2d Cir.1922), aff’d under the name of *1017Craig v. Hecht, 263 U.S. 255, 44 S.Ct. 103, 68 L.Ed. 293 (1923). Nowhere else in the country does the FBI labor under such a constraint. Should it conduct an unconstitutional investigation elsewhere the victims would have to bring a suit like the present, which took eight years to be resolved; or they would have to bring damage actions against individual FBI agents, with all the immunity and solvency problems that such suits entail. And the consent decree did go further than just enjoining the FBI from violating the Constitution; only in Chicago must the FBI report on its investigatory activity periodically to civil rights lawyers.
The plaintiffs argued for the first time at the oral argument of this appeal that the district court’s injunction does not forbid the FBI to conduct a “preliminary inquiry,” as distinct from a full-fledged “investigation,” solely on the basis of statements that are not criminally punishable. The purpose of the argument is to show that the district court’s injunction does not leave the FBI helpless to investigate potential terrorist groups. Its basis is a comparison between the first sentence in paragraph 3.4(a) of the consent decree, which refers to “domestic security investigations and inquiries,” and the second sentence (the basis of the injunction), which refers only to “investigation.” But as neither “investigation” nor “inquiry” is defined anywhere in the consent decree and “inquiry” does not even appear in the Levi Guidelines, it seems more likely that the omission of “inquiries” from the second sentence of paragraph 3.4(a) was inadvertent, or that “investigation” in that sentence was understood to be a short-hand expression for “domestic security investigations and inquiries,” than that the decree meant to distinguish between more and less extensive investigations and give the FBI a completely free hand with regard to the latter. Maybe “inquiry” in paragraph 3.4(a) means “preliminary investigation” in the Levi Guidelines, but this is entirely speculative.
We are surprised that the plaintiffs are willing to interpret paragraph 3.4(a) as permitting the FBI to conduct, on the basis of constitutionally protected advocacy, “preliminary inquiries” as defined in the Smith Guidelines. A preliminary inquiry is a limited investigation, but not so limited that it could not be used to harass inoffensive organizations; it could include the planting of undercover agents in the organization. See Part II.B(4)-(6) of the Smith Guidelines. So in conceding that preliminary inquiries are not regulated by paragraph 3.4(a) the plaintiffs have conceded too much. See Elliff, supra, at 804-05. But they have also conceded too little, because preliminary inquiries are not an adequate investigative tool for coping with groups or individuals that may be engaged in the long-range planning of terrorist acts. The term “preliminary inquiry” appears in the part of the Smith Guidelines that deals with “general crimes investigations” rather than the part that deals with “domestic security/terrorism investigations.” The latter are defined as “investigations of enterprises, other than those involved in international terrorism, whose goals are to achieve political or social change through activities that involve force or violence.” (Neither the decree nor the Guidelines deal with terrorism within or without the United States by foreign terrorist organizations.) The Guidelines describe “preliminary inquiries” as intended for situations where the FBI has ambiguous or incomplete information regarding a specific crime and needs to make further, generally unobtrusive, inquiries before deciding whether to launch a full-scale investigation. In contrast, a domestic security/terrorism investigation “is concerned with the investigation of entire enterprises, rather than individual participants and specific criminal acts, and authorizes investigations to determine the structure and scope of the enterprise as well as the relationship of the members.” Normally, moreover, preliminary inquiries must be completed within 90 days. In short, they are not designed for keeping watch on an organization not suspected of engaging in criminal activity at present or of planning criminal activity in the immediate future.
*1018We understand the plaintiffs’ concern with the repressive effect of prolonged and intrusive investigations of inoffensive groups and individuals. But their concern does not require that paragraph 3.4(a) be interpreted to protect organizations with violent aims from such investigations. The organizations that the decree sought to protect, organizations such as the ACLU and the NAACP, do not go around making threats to commit violent acts, and are not the acorns from which grow such trees as the FALN, the Posse Comitatus, the White Knights of the Ku Klux Klan, the Black Liberation Army, the New World Liberation Front, the Republic of New Africa, the Weather Underground, the Jewish Armed Resistance, Omega 9, and other menacing groups discussed in Motley, supra. Nor should the plaintiffs worry that unless Judge Getzendanner’s injunction stands, overzealous government officials will interpret “statements advocating] criminal activity” in the sentence she enjoined in the Smith Guidelines so broadly as to embrace the harmless statements of people whose politics the officials happen to dislike. That would be a reckless interpretation, especially since the Guidelines do not authorize the FBI to base a domestic security/terrorist investigation on the advocacy of non -violent crimes such as draft resistance — advocacy that might conceivably if improbably be imputed to an organization like the ACLU by hostile law-enforcement officers. It still would not be an organization that advocated political or social change through violence; passive resistance and civil disobedience are nonviolent.
But if despite all this some law enforcers do misread the Smith Guidelines as authorizing them to resume the abuses that led to this lawsuit in the first place, they will, of course, be violating the consent decree. Whether intelligently or stupidly interpreted the Smith Guidelines confer no immunities and do not supersede the decree (except for the dispensation in 113.6, not in issue here, that permits new guidelines to supersede parts of 11 3.5, which relate to the techniques rather than purposes of investigations). Those who follow them are not thereby raised above the law; and the plaintiffs’ lawyers are busy monitoring compliance with the decree to make sure the new guidelines do not result in investigations that violate the decree.
Moreover, the sentence the district court enjoined is not the only sentence in the Smith Guidelines that defines the scope of FBI domestic security investigations. It says that an investigation “may be warranted,” but does not explain when it would be warranted. For that you must go elsewhere in the Guidelines — to the provision that states:
A domestic security/terrorism investigation may be initiated when the facts or circumstances reasonably indicate that two or more persons are engaged in an enterprise for the purpose of furthering political or social goals wholly or in part through activities that involve force or violence and a violation of the criminal laws of the United States____ In determining whether an investigation should be conducted, the FBI shall consider all of the circumstances including (1) the magnitude of the threatened harm; (2) the likelihood it will occur; (3) the immediacy of the threat; and (4) the danger to privacy and free expression posed by an investigation.
Thus, before the FBI can be authorized to investigate, it is required to assess the (magnitude, likelihood, and imminence of the threatened harm from the activities to be investigated (and as is evident from the first sentence in the quoted passage, the harm must be of a type that the government is entitled to prevent though its criminal laws). Having been directed by the Guidelines to make such an assessment, the FBI is not free to ignore the results.
The Smith Guidelines, like the Levi Guidelines before them, are nationwide in their intended application. We doubt that any neutral observer would think it appropriate that the FBI should be governed by other than a uniform national set of investigatory standards — that it should operate under one set of constraints everywhere but Chicago, and under another and tighter *1019set in Chicago, so that this city can become a sanctuary for nascent terrorist organizations. To some extent the FBI is unavoidably under a tighter rein in Chicago than elsewhere, because of the consent decree, which applies only to Chicago. But there is no reason to magnify the disparity by looking for conflict between two sets of general language — the general principles of the decree and the general principles of the Smith Guidelines. Should they conflict in practice there will be time to force revision of the Guidelines. It would be different if the Guidelines and the decree were completely irreconcilable on their face. They are not. See Elliff, supra, at 812.
There is no longer any novelty to federal district judges’ presiding over prison systems, school systems, and mental institutions, often pursuant to consent decrees. See Horowitz, Decreeing Organizational Change: Judicial Supervision of Public Institutions, in The Courts: Separation of Powers 39 (Goulet ed. 1983). But to put a district judge in charge of the FBI in Chicago would write a new chapter in the annals of federal judicial enterprise, and though the consent decree does not go that far it does give the judge a great deal of power over the FBI in Chicago just by virtue of the inherent ambiguity of language— enough power to make her role in enforcing the decree an extraordinary one that if it is to be played well must be underplayed. A due regard for the separation of powers, the flexibility of equity, the ambiguity of the decree, the generality of the new Justice Department guidelines, the sensitivity and importance of the subject matter, and the limitations of judicial competence argues against precipitating a premature confrontation between the judicial and executive branches in a setting where inevitably some people will say, with pardonable exaggeration, that the federal judiciary is playing fast and loose with the public safety. Fine-tuning investigative guidelines is the responsibility of the executive branch, not us. Our duty is to enforce the decree. Any actual as distinct from purely verbal threat to compliance with the decree lies in the future; we decline to anticipate that the Justice Department will so interpret the new FBI guidelines as to violate the district court’s decree.
We do not depreciate the abuses of police power that led to the decree — abuses that were part of a national pattern that included such grotesque incidents as the following: after Martin Luther King was “named ‘Man of the Year’ by Time magazine, the FBI decided to ‘take him off his pedestal,’ [‘jreduce him completely in influence,’ and select and promote its own candidate to ‘assume the role of the leadership of the Negro people.’ ” Intelligence Activities and the Rights of Americans, S.Rep. No. 755, 94th Cong., 2d Sess. 11 (1976). See generally Donner, The Age of Surveillance: The Aims and Methods of America’s Political Intelligence System (1980); Theoharis, Spying on Americans: Political Surveillance from Hoover to the Huston Plan (1978). These abuses ought not be forgotten. But neither do we want to emulate the Bourbon kings, of whom it was said that they learned nothing and forgot nothing. The abuses that gave rise to this suit ended a decade ago, and much has changed since then, including the management of the Bureau, whose present director is a former federal circuit judge, and the amount of supervision of the FBI by the Attorney General, which, as the issuance of published guidelines suggests, is greater than it used to be. Also, although the number of bombings and other violent acts by domestic terrorist organizations has not risen over this period, public concern with domestic terrorism has risen — which is understandable considering the continued rise in international terrorism and the increasing lethality of the weapons available to terrorists. See generally S.Rep. No. 755, supra, at 112 n. 20; Motley, supra, ch. 1; Monroe, Addressing Terrorism in the United States, 463 Annals of the AAPSS 141, 143 (1982); Snitch, Terrorism and Political Assassinations: A Transnational Assessment, 1968-80, 463 Annals of the AAPSS 54, 58 (1982) (figure 1); Special Report/Terrorism, U.S. News & World Rep., Jan. 9, 1984, at 24. By reading the *1020critical sentence in the decree as prohibiting improperly motivated investigations (other parts of the decree, not in issue in this appeal, prohibit improper investigative techniques), we can reconcile the decree and the new guidelines, maintain a proper separation of powers, yet protect the First Amendment rights of the plaintiffs and the other class members.
If the Justice Department had really consented to a foolish decree, the only recourse for the government would be to seek modification of it. District judges who preside over “institutional reform litigation” in this circuit should be mindful of Judge Friendly’s recent observation that, “As experience with this type of litigation increases, a consensus is emerging among commentators in favor of modification with a rather free hand.” New York State Ass’n for Retarded Children, Inc. v. Carey, 706 F.2d 956, 970 (2d Cir.1983). But even if the more demanding requirements of United States v. Swift & Co., 286 U.S. 106, 119, 52 S.Ct. 460, 464, 76 L.Ed. 999 (1932), for the modification of equitable decrees had to be satisfied, there would be no justification for misreading the decree. Not even the government will benefit in the long run from being excused from having to honor its agreement; for who will make a binding agreement with a party that is free to walk away from an agreement whenever it begins to pinch? But it is one thing to misread a decree and another to draw back from concluding that a coequal branch of government “improvidently” surrendered its constitutional obligations and exposed the people of the nation’s third-largest city — exposed everyone in this country, for there is no way to keep a Chicago-nurtured terrorist organization in Chicago, any more than the Nazi party could be confined to Munich — to the dangers posed by embryonic terrorist groups that are immunized from FBI investigation. If the consent decree will bear an alternative interpretation it ought to be given one.
The order of the district court is reversed with directions to dissolve the injunction.