POSNER, Circuit Judge.
The question presented by this appeal is whether the judge in a civil rights case may [513]*513order the losing party to reimburse the cost incurred by the winner to hire an expert witness. The Civil Rights Attorney’s Fees Awards Act of 1976, amending 42 U.S.C. § 1988, provides that in any suit under a variety of federal civil rights statutes, “the court, in its discretion, may allow the prevailing party, other than the United States, a reasonable attorney’s fee as part of the costs” paid by the loser. As interpreted by the courts, the Act entitles a prevailing plaintiff to such an award as a matter of course, but a prevailing defendant only if the suit was completely groundless. Hensley v. Eckerhart, 461 U.S. 424, 429 and n. 2, 103 S.Ct. 1933, 1937 and n. 2, 76 L.Ed.2d 40 (1983). The Supreme Court has also read the statute to allow as part of the “reasonable attorney’s fee” the award of paralegal fees, even though paralegals are not attorneys. Missouri v. Jenkins, — U.S. -, 109 S.Ct. 2463, 105 L.Ed.2d 229 (1989). And this court, in a decision the defendants do not question, has authorized the award — again as part of the reasonable attorney’s fee — of lawyers’ out-of-pocket expenses of travel, investigation, and other trial preparation. See Henry v. Webermeier, 738 F.2d 188, 192 (7th Cir.1984). Many other courts have done likewise. See cases cited in International Woodworkers v. Champion International Corp., 790 F.2d 1174, 1184-86 (5th Cir.1986) (en banc) (dissenting opinion), aff’d under the name of Crawford Fitting Co. v. J.T. Gibbons, Inc., 482 U.S. 437, 107 S.Ct. 2494, 96 L.Ed.2d 385 (1987); cf. Mennor v. Fort Hood National Bank, 829 F.2d 553, 557 and n. 22 (5th Cir.1987).
The specific question presented by this appeal, however, is new in this circuit and has caused a division among several others. See, e.g., West Virginia University Hospitals, Inc. v. Casey, 885 F.2d 11 (3d Cir.1989); SapaNajin v. Gunter, 857 F.2d 463, 465 (8th Cir.1988); Ramos v. Lamm, 713 F.2d 546, 559 (10th Cir.1983); Sevigny v. Dicksey, 846 F.2d 953, 959 (4th Cir.1988); cf. Denny v. Westfield State College, 880 F.2d 1465 (1st Cir.1989). It is whether the expenses reimbursable under the statute include fees that the prevailing party pays to an expert hired to advise or testify in the case. The district court held that experts’ fees are reimbursable under section 1988, and, in this successful suit under 42 U.S.C. § 1983 by a class of “breakdancers” and others to invalidate on First Amendment grounds a Chicago ordinance restricting street performances, see 619 F.Supp. 1129 (N.D.Ill.1985), awarded the plaintiffs almost $10,000 for the fees they had incurred in hiring two experts to advise and testify. The reasonableness of the expenditure is not in question. The total award of attorney’s fees, expenses (including the experts’ fees), and costs approached $42,000.
The Supreme Court and this court have rejected the strongest argument against the district court’s result, the argument from the statute’s “plain language.” The fee statute authorizes the award of a reasonable attorney’s fee, and an expert witness or consultant is not an attorney. But neither is a paralegal, yet paralegal fees may be awarded (<Jenkins). An attorney’s travel expense or long-distance telephone expense is not an attorney’s fee, yet is awardable too {Henry). And although the statute makes no distinction between a prevailing plaintiff and a prevailing defendant, the standard for an award of fees is different in the two eases {Hensley).
The defendants argue that a paralegal is more like an attorney than is an economist, a psychiatrist, a police commissioner (one of the experts here), or a sociologist (the other — William Whyte, of Organization Man fame). That may be, but it does not touch the question whether the fee statute is to be read literally. A sheep is more like a goat than it is like an ostrich; but if a statute regulating sheep had been applied to goats, an attempted application to ostriches could not be defeated simply by pointing out that an ostrich is not a sheep. If “attorney” in the fee statute can mean something different from attorney, and “fee” something different from fee, then maybe one of the other things “attorney’s fee” can mean is the fee paid an expert witness or consultant.
We are quite aware that appeals to literalism are common. The cases are thick with references to “plain meaning” and [514]*514with such tired saws as that interpretation must begin with the words of the statute— and stop there if they are clear. See, e.g., United States v. Ron Pair Enterprise, Inc., — U.S. -, 109 S.Ct. 1026, 1030, 103 L.Ed.2d 290 (1989), and other cases cited in In re Sinclair, 870 F.2d 1340, 1341-42 (7th Cir.1989). In fact, interpretation must begin with the linguistic and cultural competence presupposed by the author of the statute. “Language is a process of communication that works only when authors and readers share a set of rules and meanings.... To decode words one must frequently reconstruct the legal and political culture of the drafters.” Id. at 1342. (Goodness knows what the civil rights fees statute would mean to someone outside the legal culture!) As shown by Jenkins, a 7-1 decision on the meaning of the civil rights fee statute (the Court divided more closely on another issue), rendered less than a year ago, and by countless other cases including Hensley and Henry, judges realize in their heart of hearts that the superficial clarity to which they are referring when they call the meaning of a statute “plain” is treacherous footing for interpretation. They know that statutes are purposive utterances and that language is a slippery medium in which to encode a purpose. They know that legislatures, including the Congress of the United States, often legislate in haste, without considering fully the potential application of their words to novel settings. The presence of haste here is suggested by the fact that the civil rights fees statute was passed on the last day of the Ninety-Fourth Congress.
When a court can figure out what Congress probably was driving at and how its goal can be achieved, it is not usurpation — it is interpretation in a sense that has been orthodox since Aristotle — for the court to complete (not enlarge) the statute by reading it to bring about the end that the legislators would have specified had they thought about it more clearly or used a more perspicuous form of words. That is what the Supreme Court did in the Jenkins case. The civil rights fee statute had been passed the year after Alyeska Pipeline Service Co. v. Wilderness Society, 421 U.S. 240, 95 S.Ct. 1612, 44 L.Ed.2d 141 (1975), cut off a trend in the lower federal courts toward awarding prevailing parties in civil rights and other cases, conceived of as “private attorneys general,” a reasonable attorney’s fee as a matter of federal equity power. Noting that the fee statute had been enacted in order to overrule Alyeska with respect to civil rights cases, the Court in Jenkins asked whether an award of paralegal fees would serve the statutory purpose, and concluded that it would. See 109 S.Ct. at 2471. Paralegals provide a low-cost substitute for work that would otherwise be performed by the attorney himself. The substitute is part of the attorney’s work product. If fees for the services of paralegals could not be shifted to the losing party, as attorney’s fees can be, there would be an incentive to substitute attorney time for paralegal time. So even defendants would be made worse off in the long run; they would be paying for the same work but at a higher rate. Shipes v. Trinity Industries, Inc., 685 F.Supp. 612, 615 (E.D.Tex.1987). A nonli-teral interpretation was necessary to make sense of the statute.
Much the same argument can be made for the shifting of experts’ fees. Experts are not only hired to testify; sometimes they are hired, also or instead, to educate counsel in a technical matter germane to the suit. The time so spent by the expert is a substitute for lawyer time, just as paralegal time is, for if prohibited (or deterred by the cost) from hiring an expert the lawyer would attempt to educate himself about the expert’s area of expertise. To forbid the shifting of the expert’s fee would encourage underspecialization and inefficient trial preparation, just as to forbid shifting the cost of paralegals would encourage lawyers to do paralegals’ work. There is thus no basis for distinguishing Jenkins from the present case so far as time spent by these experts in educating the plaintiffs’ lawyer is concerned (we can find no merit in a distinction based on the difference between an employee and an independent contractor), and so we think the district judge must have been right at [515]*515least in part. If so, this is a reason for thinking he was right in whole, for it is difficult to separate time spent by an expert in educating the lawyer from time spent in preparing to testify. Sometimes it is impossible; the same time may serve both purposes.
Time actually spent on the stand, however, is not easily conceived of as a substitute for lawyer time or lawyer work product. And such time is easily separated out from the other time spent by an expert on a case, though usually it is just the tip of the iceberg; in this case, so far as we can determine, less than ten percent of the money paid the experts was for time actually testifying. But the Court in Jenkins did not insist on a relation of substitution for paralegal fees to be shifted, and of course paralegals do not just do at lower cost what lawyers do. In part because they are cheaper, they do things that lawyers would not consider worth their while to do. They expand the scope of trial preparation. And so it is with experts.
But when applied to the time the expert spends on the stand, this analysis reckons without the general federal cost and witness-fee statutes, and their interpretation by the Supreme Court in Crawford Fitting Co. v. J.T. Gibbons, Inc., supra, anticipated in Chicago College of Osteopathic Medicine v. George A. Fuller Co., 801 F.2d 908, 912 (7th Cir.1986). Section 1920 of the Judicial Code lists items that a federal judge may tax as costs. “Compensation of court appointed experts” is one of these, see § 1920(6), but the experts in this case were not appointed by the court and the statute makes no explicit reference to experts retained by a party. Another item in the list is witness fees, § 1920(3). But section 1821(a)(1) of the Judicial Code provides that “except as otherwise provided by law, a witness in attendance at any court of the United States ... shall be paid the fees and allowances provided by this section,” which goes on to provide in other subsections that “a witness shall be paid an attendance fee of $30 per day for each day’s attendance,” § 1821(b), plus his travel time to and from the court, and lodging if an overnight stay is required, §§ 1821(c), (d). Sections 1821 and 1920 long predate the civil rights fees act. In Crawford the Court held that they bar the award of higher witness fees by federal judges under Rule 54(d) of the Federal Rules of Civil Procedure, the rule for the taxing of costs.
Crawford was not however a case in which the judge had rested his award of witness fees on another statute. And section 1821(a)(1) is, by its own terms, inapplicable where a statute makes other provision for costs or witness fees. The question left open by Fuller (see 801 F.2d at 912) and by Crawford (see 482 U.S. at 445, 107 S.Ct. at 2499, 2500 n. 1 (concurring opinion), 446 n. 1, 107 S.Ct. 2500 n. 1 (dissenting opinion)), but squarely posed by the present appeal, is whether the civil rights fees statute makes other provision for witness fees within the meaning of section 1821(a)(1). Notice that however this question is answered, the plaintiffs in this case would still be entitled to an award of fees incurred for advice, as distinct from testimony and preparation for testimony, by experts. Sections 1821 and 1920 do not bear on fees for nontestimonial services.
The defendants argue that Crawford requires reversal because the Court said, “We will not lightly infer that Congress has repealed §§ 1920 and 1821, either through Rule 54(d) or any other provision not referring explicitly to witness fees.... We hold that absent explicit statutory or contractual authorization for the taxation of the expenses of a litigant’s witness as costs, federal courts are bound by the limitations set out in 28 U.S.C. § 1821 and § 1920.” 482 U.S. at 445, 107 S.Ct. at 2499. Read literally this statement provides strong support for the defendants’ argument, but the literal interpretation of judicial language is no more defensible than the literal interpretation of statutory language. Read literally, the statement we have just quoted is in tension with the Court’s recognition in the Alyeska opinion that the general cost-shifting statute (§ 1920) does not bar all forms of fee shifting that are grounded in equitable rather than statutory powers — does not for example bar an award of fees under the com[516]*516mon-fund doctrine or as a sanction for a frivolous claim or defense. See 421 U.S. at 255-60, 95 S.Ct. at 1620-23. In any event, all Crawford says is that Congress’s desire to bring a case within the exception to section 1821(b) must be plain; this is not the same as saying that the statute appealed to as making other provision for witness fees must, as the defendants argue, contain the words “witness fees” or recite that it is repealing section 1920 or section 1821(b). And unlike Crawford, this is not a case where a general grant of authority, such as that contained in Rule 54(d), is appealed to against a specific provision, such as the witness fee statute. That statute is no more specific than the civil rights fees statute. If the former is specific with respect to shifting witness fees, the latter is specific with respect to shifting fees in civil rights cases. They are equally specific, or equally general.
The defendants press the following argument on us. When Congress passed the civil rights fees act it must have known, or at least it should be presumed to have known, that 28 U.S.C. §§ 1821 and 1920 were on the books and read together limited witness fees to $30 a day unless Congress provided otherwise. Therefore if Congress had wanted to lift the limit for expert witnesses in civil rights cases it would have said so. Repeals by implication are disfavored. See, e.g., Morton v. Mancari, 417 U.S. 535, 550, 94 S.Ct. 2474, 2482, 41 L.Ed.2d 290 (1974).
We expressed our doubts about this well-, known “canon of construction” in Edwards v. United States, 814 F.2d 486, 488-89 (7th Cir.1987), feeling free to do so because, while the Supreme Court frequently cites the canon with favor, it frequently refuses to apply it, too. The canon is indeed a mixed bag. It protects some old statutes from, as it were, inadvertent destruction, but it threatens to impale new statutes on the concealed stakes planted by old ones. Congressmen do not carry the statutes of the United States around in their heads any more than judges do. The canon may therefore cause a forgotten old statute inadvertently to wreck a considered new one. That is a danger in this ease. There is no indication that anyone drew Congress’s attention to the witness fee statute during the deliberations preceding enactment of the civil rights fee statute. Although an appendix to the House committee report cites a number of fee-shifting statutes some of which expressly allow the award of expert-witness fees, this feature of those statutes — which are merely cited, and not described — is not mentioned in the appendix or elsewhere. See H.R.Rep. No. 1558, 94th Cong., 2d Sess. 13-14 (1976).
If Congress’s knowledge of the entire contents of the United States Code cannot realistically be “presumed” — and it cannot — then the best that can be said for the canon against repeals by implication is, first, that it provides a mechanical rule for deciding difficult cases and thus makes law simpler and curbs judicial discretion, and, second, that it encourages legislators and their staffs to do a thorough search of previous statutes before enacting a new one. The first point identifies real, although limited, social goods (legal simplification and the reduction of judicial discretion), but the second strikes us as unrealistic about the conditions under which legislatures work. The canon seems in any event particularly inapt in a case such as this, because there is nothing to indicate that the general federal cost-shifting statutes express hostility to awarding the fees of party-retained expert witnesses in cases governed by a statute designed to enlarge the judicial power to shift costs. The issue had not arisen during the deliberations that preceded the enactment of the general statutes. If there is no conflict at the level of policy or purpose, then to give effect to the policy behind the later statute is not a “repeal” of the earlier one in a practical sense.
Perhaps the main problem with the canon is semantic — the connotations of “canon” and “canonical,” which makes it sound like a rigid rule, whereas it may just be a reminder not to assume lightly that Congress repealed a statute without saying so. But we need not continue this exploration of the “canon” against repeals by im[517]*517plication, for it is not in fact applicable to this case. The witness fee statute by its own terms does not apply where Congress has made other provision for witness fees, so if the civil rights fees statute is such other provision there is no repeal or impairment of the earlier statutes. The plaintiffs say it is, but acknowledge that this appears not from the text of the statute but from its history and purpose. They emphasize a statement on the House floor by one of the bill’s sponsors, Congressman Drinan, “that the phrase ‘attorney’s fee’ would include the values of the legal services provided by counsel, including all incidental and necessary expenses incurred in furnishing effective and competent representation.” 122 Cong.Rec. H35123 (Oct. 1, 1976). This statement is entitled to little if any weight in determining the meaning of the statute, and not only because it does not mention expert-witness fees. Floor debates are not a reliable guide to the meaning of statutes. They are poorly attended and members of Congress can make additions to the Congressional Record after the debate without notice to their colleagues. See Katzmann, Summary of Proceedings, in Judges and Legislators: Toward Institutional Comity 162, 175 (Katzmann ed. 1988).
Of slightly greater significance is the fact that the Senate committee report cites, as examples of cases that had been overruled by Alyeska and would be reinstated as it were by the new statute, several cases in which courts had awarded expert-witness fees along with attorney’s fees in the literal sense. The citation is indirect. The report notes that in the overruled cases the courts had been “following Congressional recognition in newer statutes of the ‘private attorney general’ concept [and] exercising their traditional equity powers to award attorneys’ fees.” S.Rep. No. 1011, 94th Cong., 2d Sess. 4 (1976), U.S.Code Cong. & Admin.News 1976, pp. 5908, 5911. In support of this proposition the report cites to a list of cases in the committee’s hearings, see id. at 4 n. 3, U.S.Code Cong. & Admin.News 1976, p. 5911 — and if one goes to this list one will find several case summaries that state that expert-witness fees had been awarded in the case. See Legal Fees — Hearings before the Sub-comm. on Representation of Citizen Interests of the S. Comm, on the Judiciary, 93d Cong., 1st Sess. 911, 917, 1060 (1973).
Consistent with our skepticism about legislative omniscience, we are not convinced that these documents show that Congress actually adverted to the question whether expert-witness fees could be shifted along with attorney’s fees. See Blanchard v. Bergeron, — U.S. -, 109 S.Ct. 939, 946-47, 103 L.Ed.2d 67 (1989) (concurring opinion). That would require that members of Congress had read with some care the hearings cited in the Senate committee report and appreciated the significance of the references to awards of expert-witness fees. There is no indication that they did, and it is unlikely. But this conclusion does not end the analysis. If Congress is not omniscient, this weakens one of the arguments of the plaintiffs but also shows that — the defendants to the contrary notwithstanding — little weight can be attached to Congress’s failure to imitate the five statutes listed in the appendix to the House committee report that specify expert-witness fees as a type of expense that the district court can shift to the prevailing party. All five were environmental statutes, where the need for expensive technical experts was too obvious to escape conscious legislative consideration; the last four were copies of the first.
The question what a statute means is only in part a function of what the legislators thought it meant. The scope of legislation is not circumscribed by the conscious thoughts of the legislators. Nor does the answer to an interpretive question turn on how the statute might be read by someone utterly ignorant of its background or by the same someone reading with the same literalism statements in the legislative history. “Clarity depends on context, which legislative history may illuminate.” In re Sinclair, supra, 870 F.2d at 1342. We ask what meaning would be obvious to one familiar with the circumstances of enactment, including, here, the judicial history that provides the backdrop of the statute. We ask whether we can be [518]*518confident that if someone had told Congress in the deliberations leading up to enactment that it had neglected to say anything about the shifting of expert-witness fees, Congress would have added language making clear to the most literal-minded that such fees could be shifted. We think it would have, and one reason is simply that we are given and can think of no reason against such shifting — especially given our earlier point that expert fees for advice and consultation can be shifted along with paralegal and other incidental expenses normally incurred in litigation. There would be a reason if the civil rights fees statute had been a hard-fought compromise between those who wanted judges to have the broad equitable authority they had exercised before Alyeska and those who wanted no inroads made on Alyeska, for then the court’s duty would be to give effect to the compromise, not to give proponents a victory that had eluded them in the legislative arena. See Rodriguez v. United States, 480 U.S. 522, 525-26, 107 S.Ct. 1391, 1393-94, 94 L.Ed.2d 533 (1987) (per curiam). But there is no indication of compromise on any issue relevant to this case, and the defendants do not argue it. All the evidence is to the contrary. Recognizing that most civil rights suits were not lucrative for the plaintiffs’ lawyers, Congress wanted to make losing defendants bear the expenses of suit beyond the usual items taxable as costs. Why exclude expert-witness fees? Why in particular exclude those fees when they are incurred for testimony and its preparation but not when they are incurred for advice and consultation?
The most telling clue to the statute’s meaning is the passage quoted above from the Senate committee report, discussing the equitable power exercised by courts prior to Alyeska, and a companion statement that “this bill creates no startling new remedy — it only meets the technical requirements that the Supreme Court has laid down if the Federal courts are to continue the practice of awarding attorneys’ fees which had been going on for years prior to the Court’s May decision [Alyeska].” S.Rep. No. 1011, supra, at 6, U.S. Code Cong. & Admin.News 1976, p. 5913. The courts that until Alyeska had shifted attorney’s fees and other expenses, including expert-witness fees, had relied on the existence of a federal equity power. “There is ample authority, outside the civil rights area, to support the proposition that the allowance of attorneys’ fees and expenses of preparation for trial is in the discretion of the district court sitting in equity where exceptional circumstances call for their allowance in order to do justice between the parties.” Lee v. Southern Home Sites Corp., 429 F.2d 290, 295 (5th Cir.1970) (emphasis in original). This equitable power was thought to extend to expert-witness fees, see, e.g., Wallace v. House, 377 F.Supp. 1192, 1206-07 (W.D.La.1974); Welsch v. Likins, 68 F.R.D. 589, 596-97 (D.Minn.1975); Sabala v. Western Gillette, Inc., 371 F.Supp. 385, 394 (S.D.Tex.1974) — indeed to any expense that the district court might reasonably believe should be shifted in order to do equity. See Wallace v. House, supra, 377 F.Supp. at 1206-07, and cases cited there. Alyeska held that with immaterial exceptions this equitable discretion did not exist. The Civil Rights Attorney’s Fees Awards Act of 1976 restored it, using “reasonable attorney’s fee” as a shorthand expression for such fees and other expenses as the district court might in the exercise of its equitable discretion reasonably believe should be reimbursed to the prevailing party. The term is the label for a power first asserted, then withheld, and finally restored; it is not an exhaustive description of the power.
It had long been understood that when expenses of litigation were shifted pursuant to the equity powers of the district court, the statutes providing for the taxing of “costs,” such as section 1920 (which in turn incorporates by reference the limitations on witness fees in section 1821), did not apply. See, e.g., Henkel v. Chicago, St. Paul, Minneapolis & Omaha Ry., 284 U.S. 444, 447-48, 52 S.Ct. 223, 225, 76 L.Ed. 386 (1932). So when Congress restored the equitable power to shift litigation expenses in civil rights cases, it took the question of expert-witness fees in such cases out of 28 U.S.C. §§ 1821 and 1920. No doubt it [519]*519would be unrealistic to suppose that Congress actually considered the bearing of Henkel and like cases. But it is only on an assumption of omniscience that the failure to specify expert-witness fees becomes significant, for while an omniscient Congress would remember sections 1821 and 1920, equally it would remember Henkel. To assume omniscience not only is unrealistic, but leads nowhere.
We anticipate a floodgates argument of the following form. If the civil rights fees statute is interpreted to shift the entire expenses of the prevailing plaintiff to the defendant, then plaintiffs will seek awards for the opportunity costs of their time spent in testifying or preparing to testify, or for medical expenses for treating the anxiety that litigation can trigger, or for the other consequential damages of participating even in successful litigation. But there is a reasonably bright line, which we do not think the framers of the fee statute meant to cross in seeking to resurrect a line of cases that had respected the line, between expenses incurred in hiring the inputs (of lawyer and other time, copying, and the like) necessary to prosecute a suit, and expenses incurred as a consequence of being a party even if you are indemnified for the full costs of suit. The latter category of expense raises problems of measurement and causality that the former does not and that even the most generous fee-shifting regimes exclude. See, e.g., United States v. McPherson, 840 F.2d 244 (4th Cir.1988). We think the civil rights fees statute excludes these expenses too. See Duncan v. Poythress, 750 F.2d 1540, 1543 n. 12 (11th Cir.1985); but see Laffey v. Northwest Airlines, Inc., 746 F.2d 4, 30 (D.C.Cir.1984). Our decision has, therefore, a limited scope.
It could of course be argued that one-way fee-shifting, which is essentially the regime created for civil rights litigation by the cases cited in Hensley, induces too much litigation. If so, then allowing expert-witness fees to be shifted is a move in the wrong direction. Maybe a big move, since those fees can be substantial — in this ease they are almost one-fourth of the total award of fees, which in turn was the only form of monetary relief that the plaintiff and his class obtained. But the propriety of one-way fee shifting in civil rights cases was resolved when the Supreme Court held on the basis of the legislative history of the civil rights fees statute that prevailing plaintiffs are to be treated more generously than prevailing defendants. It would be both arbitrary and inconsistent with our best guess as to the meaning of the statute to hold that, of all expenses ordinarily shifted under fee-shifting statutes, only expert-witness fees are excluded.
We do not want to exaggerate the significance of our holding, however. There is little evidence that the shifting of fees has ever made much difference in the propensity to bring civil rights suits. See Eisenberg & Schwab, The Reality of Constitutional Tort Litigation, 72 Cornell L.Rev. 641, 688-89, 693 (1987); Schwab & Eisenberg, Explaining Constitutional Tort Litigation: The Influence of the Attorney Fees Statute and the Government as Defendant, 73 Cornell L.Rev. 719, 755, 766 (1988); cf. Percival & Miller, The Role of Attorney Fee Shifting in Public Interest Litigation, 47 Law & Contemp.Prob., winter 1984, at 233, 246-47 (1984). If it has not, then it is unlikely that our decision today will have a large practical impact, since expert-witness fees will ordinarily be only a fraction of attorney’s fees.
AFFIRMED.