The nation’s major railroads unite in this appeal to confront the union of rail mainte-naneé workers, the Brotherhood of Maintenance of Way Employees. The issue is the extent to'which the railroads must compensate their maintenance workers for travel expenses; the governing instrument is the collective bargaining agreement (the Agreement) of September 26, 1996. The union believes that the Agreement obliges the railroads to pay travel expenses for all “traveling employees.” The railroads counter that they need only pay travel expenses for employees in roving “regional and system gangs.”
For the most part, American labor law leaves the resolution of industrial- disputes to the parties’ own devices, subject to procedural safeguards of fairness.: But the legal milieu for railroads and airlines is different. A strike that halted traffic along these arteries of commerce could traumatize the nation’s economy. The specter of such labor unrest led Congress in 1926 to establish a two-track regime to govern the course of a dispute. Railway Labor Act (RLA or the Act), 44 Stat. 577, as amended, codified at 45 U.S.C. §§ 151-188; see Texas & N.O.R. Co. v. Brotherhood of Ry. & S.S. Clerks, 281 U.S. 548, 562-66, 50 S.Ct. 427, 431-32, 74 L.Ed. 1034 (1930). Congress charged the federal courts with a seminal but limited role under this regime — that of taxonomist. Courts are to sort labor disputes into two piles, major or minor. Minor disputes are about enforcing rights already agreed upon in a contract. The only way to resolve a minor dispute is to go to binding arbitration; the Act forbids strikes over a minor dispute. See Consolidated Rail Corp. v. Railway Labor Executives’ Ass’n, 491 U.S. 299, 302-04, 109 S.Ct. 2477, 2480-81, 105 L.Ed.2d 250 (1989). Major disputes are about creating new rights. The avenue for resolving this kind of dispute is ultimately open-ended. The disputants must first submit to a long course of bargaining and mediation, during which they are obliged to preserve the status quo. But, if at the end the railroad and union have still riot come to an agreement, the RLA steps out of the way, and the disputants may resort to raw “economic force.” See Conrail, 491 U.S. at 303, 109 S.Ct. at 2480-81.
This dispute arose soon after the Agreement was concluded. The railroads insisted that they had no obligation to pay the travel expenses of employees beyond those in regional and system gangs. Within a month (October 24, 1996) the union had filed suit in the Central District of Illinois, requesting a declaration that the railroads had tried to amend the Agreement unilaterally, which is to say that the dispute was major. Four of the railroads counter-claimed, urging that the dispute was minor. (The District Court for the Western District of Virginia transferred a parallel proceeding brought by the remaining two railroads, Norfolk Southern Railway Co. and Norfolk and Western Railway Co-., and the, suits were consolidated.) The railroads sought a prehminary injunction against any strikes, work stoppages or pick*639eting over the issue of travel expenses. The parties agreed to combine the hearing on the preliminary injunction with the trial on the merits. See Fed.R.Civ.P. 65(a)(2). The district court held that the dispute was major and ruled for the union. The railroads appeal. Our review is de novo. Railway Labor Executives Ass’n v. Norfolk & W. Ry. Co., 833 F.2d 700, 706 (7th Cir.1987).
The case turns on a question of interpretation: to which sources may a court look in interpreting a collective bargaining agreement under the RLA? We hold that a court may refer to the official, formal bargaining history when it construes an agreement under the RLA We reverse.
I. History of the dispute over travel expenses
The controversy between the parties turns on two views of the classes of employees that are entitled to have their travel expenses reimbursed. The union takes the broader view — the Agreement encompasses all “traveling employees.” A “traveling employee” is one whose work travels and who does not report to a fixed headquarters. An employee who commutes long-distance daily to a particular fixed site does not fit in this group. From what we can glean from the record, there are two kinds of traveling employees: those attached to district gangs, and those associated with regional and system gangs. The district gangs work within just one of the geographical units into which the railroads divide their workers, the seniority districts. Regional and system gangs, on the other hand, may traverse two or more seniority districts in the course.of their work. Because the seniority districts can stretch over hundreds of miles, an employee from either a district gang or a regional and system gang may have to work a considerable distance from his or her home.
The history of the instant dispute reflects the unique legal setting of labor strife in the rail industry — a setting on which the disposition of this ease depends. Railroad maintenance by its very nature demands that workers travel. The union, however, has bitterly opposed the use of regional and system gangs, which tends to impose an extraordinary travel burden on the gang members. The railroads refused to give up the regional and system gangs, and in 1991, the conflict impeded a negotiated agreement. With a strike looming, President Bush invoked the RLA 45 U.S.C. § 160, to appoint Presidential Emergency Board No. 219 (PEB 219). PEB 219 investigated the discord and rejected the union’s proposal to eliminate the regional and system gangs. The union opted to strike, but Congress intervened and imposed PEB 219’s recommendations wholesale on the railroads and the union. Pub.L. No. 102-29,105 Stat. 169 (1991).
The 1991 agreement allowed for re-opening negotiations in November 1994. The issue of regional and system gangs got the limelight again as the union once more sought to cut them back. The contentious issue stalled agreement a second time. President Clinton appointed Presidential Emergency Board 229 (PEB 229), which issued its report on June 23, 1996 (PEB 229 Rep.). Perhaps spurred by the memory of Congress’ intervention in 1991, this time the railroads and the union averted a strike. With regard to travel expenses (the issue before us), the union and the railroads could not agree on implementing language. They chose instead to copy verbatim the pertinent recommendation of PEB 229 as Article XIV (Travel Allowance) of the Agreement.
II. Article XIV viewed alone
We turn then to Article XIV to see whether the dispute is major or minor. The distinction hinges on whether old rights are being enforced or new rights created. Elgin, J. & E. Ry. Co. v. Burley, 325 U.S. 711, 723, 65 S.Ct. 1282, 1289-90, 89 L.Ed. 1886 (1945). Unsurprisingly, a railroad and a union often differ about classifying a particular dispute. Unions lean toward considering disputes major, and railroads the other way, because the chief consequence of pigeonholing a dispute as “major” — that a union may opt to strike or picket — grants the union considerable leverage over the railroads. In Conrail, the Supreme Court instructed how courts should sort disputes into the two piles. A suit is minor if the railroad’s interpretation “is arguably justified by the terms of the *640parties’ collective-bargaining agreement.” Conrail, 491 U.S. at 307, 109 S.Ct. at 2482-83 (emphasis added). “Where, in contrast, the employer’s claims are frivolous or obviously insubstantial, the dispute is major.” ' Id. (emphasis added). That, then, is how we frame the question: is the railroads’ reading of Article- XIV as applying only to regional and system gangs arguably justified, or is it frivolous or obviously insubstantial?
Were we to view Article XIV in a vacuum, we would conclude, as the district court did, that the railroads’ position is feeble. Article XIV has two substantive sections, each pulled from PEB 229’s report. The first section lays out a schedule of payments to workers based on distances traveled ($25 per 100 miles). The second provides the option of air transport for workers going more than 400 miles from their homes. Neither section defines which employees are to be covered. The first opens as follows:
At the beginning- of the work seasonf,] employees are required to travel from their homes to the initial reporting location, and at the end of the season they will return home. -This location could be hundreds of miles from their residences. During the work season the carriers’ service may place them hundreds of miles away from home at the end of each work week.
The second section is no more specific. The benefits for air travel, it states, accrue to “employees required to work over 400 miles from their residences.”
The union-says “employees” refers to all traveling employees (as defined above). The railroad says the term should be construed to cover only regional and system gangs. Either view is logically possible; neither is barred by the explicit terms of Article XIV. But while the term “employees” could refer solely to regional and system gangs, there is no hint in Article XIV that “employees” actually bears the narrower meaning. The railroads reply that Article XIV also gives a special allowance to employees “[a]t the start up and break up of a gang.” Again, the hitch for the railroads is that the word “gang” could refer just to regional and system gangs, or to both district gangs and regional and system gangs. The railroads propose a theoretically plausible distinction, but one that has no basis in the text. We would hold the railroads’ view “frivolous or obviously insubstantial” and affirm the district court— if-the act of interpretation were to stop at the four corners of the Agreement.
III. Bargaining history: the Presidential Emergency Boards
So the railroads’ position would be frivolous, if we could not incorporate sources of meaning beyond the Agreement itself. But we can; in fact, we must. The Supreme Court long ago ruled that courts should interpret collective bargaining agreements with a keen eye for the nuances of context. “A collective bargaining agreement is not an ordinary contract for the purchase of goods and services, nor is it governed by the same old common-law concepts which control such private- contracts.” Transportation-Communication Employees Union v. Union Pac. R.R. Co., 385 U.S. 157, 160-61, 87 S.Ct. 369, 371, 17 L.Ed.2d 264 (1966); see also Southeastern Pa. Transp. Auth. v. Brotherhood of R.R. Signalmen, 882 F.2d 778, 784-85 (3d Cir.1989) (SEPTA). The assumptions courts make in reading a contract do not translate readily into the sphere of labor agreements. Contracts at common law are presumed to grow out of free choice, “in the sense that there is no real compulsion [for the parties] to deal with one another, as opposed to dealing with other parties.” United Steelworkers of Am. v. Warrior & Gulf Navigation Co., 363 U.S. 574, 580, 80 S.Ct. 1347, 1352, 4 L.Ed.2d 1409 (1960). Courts hold the parties to the letter of their contracts, because the parties' affirmatively chose those words to structure a legal mini-universe out of the void. Collective bargaining agreements, in contrast, typically govern pre-existing relationships not easily broken and remade. For each side, the real options áre governance “by an agreed-upon rule of law or leaving each and every matter subject to a temporary resolution dependent solely upon the relative strength, at any given moment, of the contending forces.” Id. Under these conditions, a labor agreement cannot be presumed to represent a true “meeting of the minds.” Conrail, 491 U.S. at 317, 109 S.Ct. *641at 2488. A collective bargaining agreement thus falls into a different order of legal instrument from an ordinary contract. Rather, “‘it is a generalized code to govern a myriad of cases which the draftsmen cannot wholly anticipate.’” Conrail, 491 U.S. at 311-12, 109 S.Ct. at 2484-85 (quoting Warrior & Gulf, 363 U.S. at 578-79, 80 S.Ct. at 1350-51 (internal citation omitted)).
The notion of a “meeting of the minds” breaks down still further under the Railway Labor Act. The RLA takes as its “primary goal” “to settle strikes and avoid interruptions to commerce.” Burlington Northern R.R. Co. v. Brotherhood of Maintenance of Way Employes, 481 U.S. 429, 451, 107 S.Ct. 1841, 1854, 95 L.Ed.2d 381 (1987). This goal imposes a heavy duty upon railroads and unions alike “to exert every reasonable effort ... to settle all disputes.” 45 U.S.C. § 152, First. The RLA compels a railroad and a union to bargain, and to do so under the shadow of the emergency powers of the President and Congress. See, e.g., Burlington Northern, 481 U.S. at 452, 107 S.Ct. at 1854-55 (noting 1986 congressional intervention). Should the parties fail to agree to some text — any text — a real possibility exists of Congress’ forcing its own solution. Under these circumstances, we cannot presume that the parties act at their peril if their wording is misleading, just as we would not read a statute or treaty as if the lawgiver or contracting powers drafted their text under strict liability.
The railroads and the union did not negotiate Article XIV in a vacuum, and we will not act as if they had. For collective bargaining agreements under the RLA, we must look beyond the document itself. We must took to the parties’ “practice, usage and custom.” Union Pac. R.R. Co., 385 U.S. at 161, 87 S.Ct. at 371. We must took to parallel labor agreements, even those involving other parties. Id. And in this case wé must took to a portion of the bargaining history. But such extrinsic evidence cannot be used to “contradict the express provisions of a written contract.” Brotherhood of Ry., Airline & S.S. Clerks v. Atchison, Topeka & Santa Fe Ryi, 847 F.2d 403, 406 n. 2 (7th Cir.1988). The RLA may direct us to took beyond the agreement, but it does not free either party from the consequences of assenting to a text. Our purpose instead is to ascertain whether the ostensible meaning of Article XIV — lucid and definitive as it might seem — is in fact the right one: This is a court’s responsibility to some extent even under contract law, where a court may unveil a latent ambiguity through extrinsic evidence. Bank of the United States v. Dunn, 31 U.S. 51, 54, 6 Pet. 51, 8 L.Ed. 316 (1832) (noting this exception to parol evidence rule); FDIC v. W.R. Grace & Co., 877 F.2d 614, 620-21 (7th Cir.1989); DeKalb Bank v. Purdy, 166 Ill.App.3d 709, 117 Ill.Dec. 606, 520 N.E.2d 957, 963 (1988). A federal court must enjoy at least the same authority under the RLA. See, e.g., SEPTA, 882 F.2d at 783 (a court may took to “side letter agreements, policy statements and any other evidence of the intent of the parties which may properly be considered”).1
Here the railroads point to a single slice of the bargaining history, the hearings and report of Presidential Emergency Board *642229. It is a particularly apposite slice for our purposes, because the negotiators took a chunk of PEB 229’s report and inserted it word-for-word in the Agreement as Article XIV. The bargaining history hence is of a peculiar sort. There is a danger in opening the door too wide to extrinsic evidence, so that the next judge will have to confront illegible notes scribbled on a bar napkin, or testimony about the musings of coffee-sated negotiators at 2 a.m. But that danger is not present here. We have a manageable and definable set of documents, collected from an official hearing, from which the parties drew the language at issue in this case. If an ambiguity is asserted to exist beneath the surface of that text, we cannot believe that Congress would have us blind ourselves to what its original drafters meant to say.
In a moment we will consider what this extrinsic evidence reveals; but first we must explain why we cannot adopt the railroads’ rationale for consulting it. The issue before us is whether the tussle .over travel allowances is a major or minor dispute. The railroads say that the way to resolve this question is to ask what an arbitrator might do. If an arbitrator might rule for the railroads, then that’s it — the dispute is minor and it must go to arbitration. That a court should put on an arbitrator’s hat has some intuitive appeal. The content of . an agreement depends as a practical matter on who is going to interpret it and how she will do so. If an agreement’s drafters know that arbitrators will likely interpret things in way X (say, by looking at bargaining history) and not way Y, then that knowledge will inform what the drafters set down on paper and how they expect it to be seen.
Yet mimicking arbitrators is not what Congress commanded us to do. • Congress committed the classification' of RLA disputes to federal judges, not to arbitrators; and we have no warrant to cede that commission de facto. Nowhere do we find a suggestion by Congress or the Supreme Court (or the inferior courts, for that matter) endorsing the railroads’ proposal. This absence is telling, for the initial charm of the railroads’ position dissolves on further inspection. The rulings of arbitrators are subject to judicial review of an especially narrow kind. Brotherhood of Locomotive Engineers v. Atchison, Topeka & Santa Fe Ry. Co., 768 F.2d 914, 921 (7th Cir.1985) (“Perhaps ‘review’ is a misnomer.”). This brand of judicial review guards against only the arbitrator run amok. Arbitral awards that fall short of that berserk level, even if rash, foolish and ill-advised, must stand. They thus would make a poor source of law for courts to tap in distinguishing major and minor disputes. (Though facially neutral, the railroads’ attempt to elide the major/minor test with the judicial review test would work to their favor. The railroads would define a minor dispute as anything fitting into the gamut of arbitral awards that can survive judicial review. Given the deference paid even to slapdash awards, the class of major disputes would dwindle to meager proportions indeed.) To pin judicial rulings to the vagaries of nearly unconstrained awards would also suffer from a subtler but perhaps graver defect. For the classification of RLA disputes, the railroads would have us put the cart before the horse. It is for the judges to say what the law is, not for the arbitrators. The tether that binds arbitrators to judicial rulings may be a loose one, but the pull is from the judges to the arbitrators, not the other way around.
But even if for the wrong reasons, the railroads are right: the district court should have examined the bargaining history of PEB 229. That history comprises the Presidential Emergency Board’s report and the parties’ written and oral submissions. There the railroads claim to find proof positive that when the PEB drafted the future Article XIV, the Board thought that the union’s overarching concern was regional and system gangs. With this we readily agree. When PEB 229 pronounced its understanding of the union’s general contentions, it explicitly linked the issue of travel allowances to regional and system gangs, not all traveling employees. See JA-0362. The PEB also explained the union’s specific proposal on travel allowances in terms of regional and system gangs.2 JA-0369. In, several in*643stances the union itself straightforwardly framed the issue of travel allowances around the regional and system gangs. The starkest example was the presentation to the PEB by the union’s director of arbitration. He began by “set[ting] forth three fundamental proposals.” “Here is what we are after, here is what the [union] wants out of this proposal,” he said. The first two requests had to do with meals and lodging; the third with travel allowances. “Third, we believe maintenance-of-way employees should not be required to subsidize the carriers’ most productive gangs, by traveling long hours without pay or travel expense to work on the gangs.” JA-0646. Uncompensated travel time for regional and system gang members in particular came up again when the union’s general counsel addressed the PEB. JA-0628, 0634, 0635. There are further instances we might repeat, but the point is made. JA-0442, 0568, 0572.
From this the railroads would have us infer that the PEB’s recommendation .about travel allowances referred strictly to regional and system gangs. This one is harder to swallow. The PEB’s recommendation is, of course, silent on this point. The railroads’ submissions to the PEB themselves do not support their case: they sometimes speak in the broad language the unions urge today, as if allowances for all traveling employees were at stake. JA-0661-63; 0673-78; 0714; 0726-27. The union’s brief to the PEB on “Expenses Away from Home” seems to encompass all traveling employees on the issue of travel allowances, albeit with some ambiguity. JA-0411, 0414. But see JA-0444.
That the PEB’s recommendations, as swept into Article XIV of the Agreement, referred to all traveling employees (and not just regional and system gangs) seems to us plausible, maybe even probable. For the union to prevail under the RLA, however, the evidence must do more than tilt; it must slide into a heap on one side. Railway Labor Executives Ass’n v. Norfolk and W. Ry. Co., 833 F.2d at 705 (“Because a major dispute can escalate into a strike, if there is any doubt as to whether a dispute is major or minor a court will construe the dispute to be minor.”); see also General Comm, of Adjustment, United Transp. Union, W. Md. Ry. Co. v. CSX R.R. Corp., 893 F.2d 584, 591 (3d Cir.1990) (“close cases” to be treated as minor). The evidence is not so lopsided. Based on what the PEB said and how the parties had east their views before it, we think that a latent ambiguity may exist in the Agreement, namely that the PEB meant only to refer to regional and system gangs. The railroads’ interpretation therefore is “arguably justifiable” and not “frivolous.” See Conrail, 491 U.S. at 320, 109 S.Ct. at 2489-90 (noting employer bears only “light burden” in establishing arbitral jurisdiction, the meeting of which in “no way” suggests employer “is entitled to prevail before the Board on the merits of the dispute”). The dispute is minor and the judgment of the district court is Reversed. The ease is Remanded with directions for the district court (1) to dismiss the union’s suit (Dist. Ct. No. 96 C 1515) for lack of subject matter jurisdiction, and (2) to enter judgment for Norfolk Southern Rail*644way Co. and Norfolk and Western Railway Co. on their suit (Dist. Ct. No. 96 C 1524) for declaratory and injunctive relief.
. The dissent presents thoughtful policy arguments against considering any extrinsic evidence when the plain language of an agreement seems clear. But the Rubicon has been crossed by the Third Circuit in SEPTA. See generally SEPTA, 882 F.2d at 783-86. We would find it very difficult to maintain that the suggestion to consult extrinsic evidence, supported by Third Circuit precedent, is "frivolous or obviously insubstantial,” Conrail, 491 U.S. at 307, 109 S.Ct. at 2482-83. The dissent also cites the views of a majority of the judges in Bidlack v. Wheelabrator Corp., 993 F.2d 603 (7th Cir.1993) (en banc), as persuasive authority against considering extrinsic evidence in this context. See Dissent. Op. at 645. Even Judge Posner’s lead opinion in Bid-lack, however, recognized that "extrinsic evidence is admissible to show that a written contract which looks clear is actually ambiguous” when there exists "contractual language on which to hang the label of ambiguous,” 993 F.2d at 608 — such as the term "employee" in Article XIV. And, of course, Bidlack did sanction consideration of extrinsic evidence in the interpretation of the agreement in that case. Further, Bidlack dealt with the merits of a dispute under the National Labor Relations Act, and for that reason alone could not establish that, in an RLA case, a suggestion to consult extrinsic evidence to uncover a latent contractual ambiguity is a frivolous legal argument. This does not mean, and we do not read SEPTA to suggest, that all forms *642of extrinsic evidence are entitled to equal, or even significant, weight.
. The dissent observes that."[t]he recommendations regarding travel allowances and those re*643garding regional and system gangs are in wholly separate sections (sections 7 and 10, respectively)” of part VI of PEB 229’s report. Dissent. Op. at 646. Part V of the report describes "The Parties’ Contentions,” and part VI contains PEB 229’s "Recommendations.” Section 7 of part V deals with "expenses away from home,” and includes the following language about travel expenses: “BMWE [the union] contends that .... maintenance of way employees subsidize the Carriers' regional and system gangs with staggering amounts of unpaid travel time and unreim-bursed automobile expense for travel between their homes and distant work locations.” PEB 229 Rep. at 10, JA-0369. The union’s position on meal and lodging allowances is also discussed in this section. Section 10 of part V, covering "regional and system-wide production gangs,” discusses the union’s position on the minimum size of regional and system-wide gangs and various other rules relating to such gangs, but does not include any discussion of meal, lodging or travel expenses. PEB 229 Rep. at 12, JA-0371. The possibility that, in part VI, the drafters would choose to discuss a recommendation for travel reimbursements (exclusively) for regional and system gangs under the heading "expenses for meals and lodging and travel allowance” (section 7) — along with other topics such as meal and lodging reimbursements — rather than under the heading "regional and systemwide production gangs” (section 10) — along with various rules pertaining to such gangs — is not so remote as to significantly undermine the railroads' proposed interpretation of the disputed provision.