POSNER, Circuit Judge.

These are consolidated appeals from an order by Judge Grady cutting by roughly one-half the attorney's fees requested by the plaintiffs' counsel in a class action. 750 F. Supp. 868 (N.D.Ill. 1990). The principal defendant, the Continental Bank, had purchased more than a billion dollars in oil and gas loans from the Penn Square Bank, an Oklahoma bank whose collapse in 1982 made the loans that Continental Bank had bought from Penn Square largely uncollectable. Caught thus in the undertow of the Penn Square disaster, Continental itself became insolvent and was taken over by the Federal Deposit Insurance Corporation. Several class actions were brought against Continental and its officers in 1982 under the federal securities laws on behalf of investors who had bought stock in the bank before the crash and who claimed that Continental had made misleading statements about its financial condition. These suits were consolidated before Judge Grady. Extensive discovery and collateral litigation ensued. Between 1986 and 1988 the suits were settled for a total (including interest to the end of 1989) of $45 million.

Having employed their professional skills to create a cornucopia for the class, the lawyers for the class were entitled under the principles of restitution to suitable compensation for their efforts. Boeing Co. v. Van Gemert, 444 U.S. 472, 100 S.Ct. 745, 62 L.Ed.2d 676 (1980). They had agreed at the outset that they would not seek an award of attorneys' fees in excess of 20 percent of the amount of recovery. That capped their request at $9 million. But proceeding under the so-called "lodestar" method, by which the fee award is built up from the number of hours spent by the lawyers on the case, Blanchard v. Bergerson, 489 U.S. 87, 94, 109 S.Ct. 939, 944, 103 L.Ed.2d 67 (1989), they managed to justify (to their own satisfaction anyway) an award that, but for the ceiling, would have been even higher. So they requested the entire $9 million, which Judge Grady, in a meticulous 81-page order, reduced as we have said by roughly a half. We respect the care with which he reviewed the lawyers' submissions, despite the absence of an adversary presentation. (The class was notified of the fee request, but no member of the class objected. There is no appellee.) And we review his decision under a deferential standard. Hensley v. Eckerhart, 461 U.S. 424, 437, 103 S.Ct. 1933, 1941, 76 L.Ed.2d 40 (1983); Rivera v. Benefit Trust Life Ins. Co., 921 F.2d 692, 698 (7th Cir. 1991); Ustrak v. Fairman, 851 F.2d 983, 987 (7th Cir. 1988). Even so, we are unable to accept most of his rulings.

1. The judge placed a ceiling of $175 on the hourly rates of all lawyers for the class, including lawyers whose regular billing rates were almost twice as high. He did this on the theory that the most demanding work on the case had been done by a rather junior lawyer whose billing rate is only $175. The more experienced, higher-paid lawyers simply were not, in the judge's view, needed for this case. This is highly implausible, when one considers that the defendants hired a crowd of pricey lawyers to defend the case and that the FDIC, in a parallel suit, hired one of the class counsel and paid him the same market rate that the district judge refused to authorize — even though the class action was contingent, and the FDIC suit was not. Nor were the lawyers who defended Continental and its officers against the class, at rates similar to the normal billing rates of the lawyers for the class, at risk of not being paid.

It is apparent what the district judge's mistake was. He thought he knew the value of the class lawyers' legal services better than the market did. What the market valued at $350 he thought worth only half as much. He may have been right in some ethical or philosophical sense of "value" but it is not the function of judges in fee litigation to determine the equivalent of the medieval just price. It is to determine what the lawyer would receive if he were selling his services in the market rather than being paid by court order. Missouri v. Jenkins, 491 U.S. 274, 285, 109 S.Ct. 2463, 2470, 105 L.Ed.2d 229 (1989); Harsch v. Eisenberg, 956 F.2d 651, 663 (7th Cir. 1992); Bandura v. Orkin Exterminating Co., 865 F.2d 816, 822 (7th Cir. 1988); Tomazzoli v. Sheedy, 804 F.2d 93, 98 (7th Cir. 1986) (per curiam); Kirchoff v. Flynn, 786 F.2d 320, 323-26 (7th Cir. 1986); Henry v. Webermeier, 738 F.2d 188, 195 (1984).

2. The judge committed the same error when he refused to allow paralegal services to be compensated at market rates but instead attempted to compute an hourly expense of each paralegal, consisting of the paralegal's weekly salary divided by 40 plus overhead directly attributable to that paralegal, such as health benefits, but excluding general office overhead, such as rent. Paralegals take up space, and they're paid even when they're not working; so the judge plainly underestimated their hourly expense. But his mistake went deeper. He was again trying to determine the value of a service that the market has set its own value on. The Supreme Court has disapproved the approach of basing reimbursement for paralegal expenses on the "cost" of the paralegal, as distinct from the market value of his services, when it is customary to bill separately for the value of those services. Missouri v. Jenkins, supra, 491 U.S. at 287-89, 109 S.Ct. at 2470-72. The approach is not only unsound; it is futile from the standpoint of economizing on the expense of litigation. It will lead lawyers to substitute their own time, for which they are entitled to be compensated at market rates rather than at some constructed hourly cost, for that of cheaper paralegals.

3. The judge refused to award a risk multiplier — that is, to give the lawyers more than their ordinary billing rates in order to reflect the risky character of their undertaking. This was error in a case in which the lawyers had no sure source of compensation for their services. Suppose a lawyer can get all the work he wants at $200 an hour regardless of the outcome of the case, and he is asked to handle on a contingent basis a case that he estimates he has only a 50 percent chance of winning. Then if (as under the lodestar method) he is still to be paid on an hourly basis, he will charge (if risk neutral) $400 an hour for his work on the case in order that his expected fee will be $200, his normal billing rate. If the fee award is to simulate market compensation, therefore, the lawyer in this example is entitled to a risk multiplier of 2 (2 X $200 = $400). The need for such an adjustment is particularly acute in class action suits. The lawyers for the class receive no fee if the suit fails, so their entitlement to fees is inescapably contingent.

The example is oversimplified, because the risk of loss varies over the life of a case. It may be very high at first, but if the plaintiff surmounts the hurdles of the defendant's pretrial motions to dismiss or for summary judgment and proceeds to trial before a jury, it may fall dramatically. That is a refinement a district judge may or may not want to take into account. We are not trying to put him into a straitjacket but are merely emphasizing that the failure to make any provision for risk of loss may result in systematic undercompensation of plaintiffs' counsel in a class action case, where as we have said the only fee that counsel can obtain is, in the nature of the case, a contingent one.

The judge understood all this but said there was no need for any contingency adjustment here because there was no contingency — the risk of the suit's failing was zero. He based this surmise on the fact that most suits, including most class actions, are settled rather than abandoned or decided adversely to the plaintiff — that most plaintiffs therefore get something for having sued — and that Continental's shenanigans with respect to its Penn Square loan portfolio had been thoroughly aired in congressional hearings held and widely publicized before these suits were filed. These are not adequate reasons. Skelton v. General Motors Corp., 860 F.2d 250, 258 (7th Cir. 1988). Most suits settle for something — have, at the very least, some nuisance value — but the question is, settle for how much? The high-priced lawyers for the class put in thousands of hours before they achieved a settlement, and although the judge found some overstaffing he did not suggest that any of it was due to the lawyers' desire to postpone settling the case. That after years of stubborn negotiation Continental and the other defendants doubtless would have been willing to pay something to get the case off their backs does not imply that class counsel were certain to obtain a settlement entitling them to a fee commensurate with their efforts. They could have lost everything, because there was considerable doubt whether Continental had actually made any material misrepresentations. A codefendant, Ernst Whinney, refused to settle and was exonerated of all charges after a long trial.

4. The judge made large across-the-board cuts in two categories of lawyer time — research and conferring. He cut legal research time by 40 percent on the ground that experienced securities counsel don't need to do much research. That clearly is incorrect. No matter how experienced a lawyer is, he has to conduct (or have conducted for him) research to deal with changes in the law, to address new issues, and to refresh his recollection. No one carries the whole of federal securities law — not only the many detailed statutes and regulations but the thousands of decided cases — around in his head, and a lawyer who tries to respond to a motion or brief without conducting fresh research is courting sanctions or a malpractice suit. The judge gave no examples of excessive time spent on legal research — he just had a gestalt reaction that there was too much. That isn't good enough. Likewise with the slashing of hours spent conferring.

Of course we must give weight to the judge's greater familiarity than our own with this suit. But this weight has to be tempered by the fact that most of the lawyers' activity was conducted outside the courtroom (and so the judge's observation), that the case was settled rather than litigated, that it was nonroutine, and that rather than disallowing specific items of unreasonable activity the judge slashed broad categories of activity by arbitrary percentages. That won't do, especially when millions of dollars are at stake. In re Fine Paper Antitrust Litigation, 751 F.2d 562, 594-95 (3d Cir. 1984). The qualification is important: in small cases, the amount at issue in the request for lawyers' fees may be too slight to justify cutting it with laser precision. The meat-axe approach (we called it "trimming fat from a fee application" in Tomazzoli v. Sheedy, supra, 804 F.2d at 98) may be acceptable in such a case. (The judge awarded only $6,000 in attorney's fees in Tomazzoli.) See also Ustrak v. Fairman, supra, 851 F.2d at 989 ($30,000). It is true that in a large case, where the stakes in the attorney's fee determination are greater, so is the burden on the judge because the fee application will be longer and more complex. But in Evans v. City of Evanston, 941 F.2d 473, 476-77 (7th Cir. 1991), we upheld the district judge's use of a sampling procedure to determine the reasonableness of a fee request so that he wouldn't have to scrutinize every entry on the lawyers' billing sheets. A judge is not permitted to destroy substantial entitlements to attorney's fees on the basis of his inarticulable and unsubstantiated dissatisfaction with the lawyers' efforts to economize on their time and expenses.

5. The judge refused to allow the lawyers to bill any of their out-of-pocket expenses of using a computerized legal research service (LEXIS). He thought those expenses should be part of the lawyers' overhead. This was another clear error. Harman v. Lyphomed, Inc., 945 F.2d 969, 976 (7th Cir. 1991). If computerized research expense were customarily treated in this fashion, lawyers' hourly rates would be even higher than they are, requiring an adjustment in the lodestar. Missouri v. Jenkins, supra, 491 U.S. at 286-87, 109 S.Ct. at 2470-71. But the more important point is that the market — the paying, arms' length market — reimburses lawyers' LEXIS and WESTLAW expenses, just as it reimburses their paralegal expenses, rather than requiring that these items be folded into overhead. Markets know market values better than judges do. And as with paralegals, so with computerized research: if reimbursement at market rates is disallowed, the effect will be to induce lawyers to substitute their own, more expensive time for that of the paralegal or the computer.

6. The judge refused to award the lawyers prejudgment interest. His reason was that he was using as the lodestar their 1988 billing rates and thus automatically compensating them for the loss of the use of their money. The alternative (described in Fleming v. County of Kane, 898 F.2d 553, 562-65 (7th Cir. 1990)) would have been to base the award on the rates the lawyers charged when they rendered the services to the class and to add interest on that amount to the present. Both approaches have been deemed acceptable. Missouri v. Jenkins, supra, 491 U.S. at 282-83, 109 S.Ct. at 2468-69; Skelton v. General Motors Corp., supra, 860 F.2d at 255 n. 5. In many, perhaps most, cases the second is more straightforward. The cost of delay in receiving money to which one is entitled is the loss of the time value of money, and interest is the standard form of compensation for that loss. Changes in billing rates from year to year reflect at most only one component of an interest rate — the rate of inflation — and that component may be swamped by factors having nothing to do with the time value of money, such as the demand for and supply of the services of particular lawyers, firms, or specialties, or of the profession as a whole. Suppose that, because of the troubled state of the legal-services market today, the billing rates of the class counsel have actually fallen since the time when they rendered their services to the class. In that event, to award them a fee on the basis of the current rates would be in effect to award them negative interest. Yet — to recur to the essential touchstone of analysis in these cases — if the lawyers had rendered the services to paying clients, their fees would have been computed and paid at the lawyers' then current billing rates, and the lawyers would have earned interest on those fees to the present. A further consideration is that there may be better information about past than about current billing rates. Past rates are rates that were actually charged to paying clients; current rates may be mere estimates of rates to be charged to those clients.

Again, we don't want to put the judge into a straitjacket. But if he uses current rates, he must be sure to make some provision for the interval between the "current" period and the date the lawyers actually receive their money from the fund for the class. Here the district judge awarded attorneys' fees, on the basis of 1988 billing rates, not in 1988 but late in 1990, leaving a gap of two years.

7. The judge refused to award any money to the named plaintiff for his admittedly modest services; the class lawyers had requested $10,000 for him. The threshold question is whether a named plaintiff is ever entitled to a fee. The basis for an award of fees in a common-fund case is, as we said, restitutionary, and the law of restitution (excepting salvage in admiralty) generally confines the right to restitution to professionals, such as doctors and lawyers. 2 George E. Palmer, The Law of Restitution, ch. 10 (1978). If you dive into a lake and save a drowning person, you are entitled to no fee. The named plaintiff is not a professional; he is, at most, a public-spirited member of the class. Yet the usual formulations of the common-fund doctrine describe the plaintiff rather than his lawyer as the person entitled to be compensated for the expenses he has incurred in conferring a benefit on the (other) beneficiaries of the common fund. See, e.g., Trustees of the Internal Improvement Fund v. Greenough, 105 U.S. 527, 26 L.Ed. 1157 (1882); Sprague v. Ticonic National Bank, 307 U.S. 161, 59 S.Ct. 777, 83 L.Ed. 1184 (1939); Boeing Co. v. Van Gemert, supra, 444 U.S. at 478, 100 S.Ct. at 749. The principal expense is the attorney's fee, but there can be others, provided they are not personal. Trustees of the Internal Improvement Fund v. Greenough, supra, 105 U.S. at 537-38; Granada Investments, Inc. v. DWG Corp., No. 91-3297, slip op. at 9 (6th Cir. April 30, 1992). Since without a named plaintiff there can be no class action, such compensation as may be necessary to induce him to participate in the suit could be thought the equivalent of the lawyers' nonlegal but essential case-specific expenses, such as long-distance phone calls, which are reimbursable.

The named plaintiff in this case was deposed, which took a few hours, and bore a slight risk of being made liable for sanctions, costs, or other fees should the suit go dangerously awry. Rand v. Monsanto Co., 926 F.2d 596 (7th Cir. 1991). The judge's ground for refusing him a fee was that the risk was negligible because the case was a clear winner and that if the named plaintiff had dropped out because he couldn't hope to be compensated for his modest efforts there were plenty of others to take his place without demanding compensation. The implicit reasoning is that the market would have produced a named plaintiff willing to charge a price of zero, and we think there is enough basis for this conjecture, given the state of the record, to let it stand. The "clear winner" point makes more sense here than in regard to the issue of a risk multiplier, because there the essential question was how large the settlement would be and here the question is whether the named plaintiff was at risk of being made to pay costs or some other sanction if the case went to trial and flopped. That risk was small. Not zero, and not only because the defendants might have won a complete victory at trial as Ernst Whinney did; an additional point is that if the case had gone to trial and the defendants had made an offer of settlement under Fed.R.Civ.P. 68 and the class had gone on to win a judgment smaller than the offer, the defendants would have been entitled to reimbursement of their costs, and maybe the class counsel would try to stick the named plaintiff with that bill. But that's a huge "maybe," even if we disregard the practical problems of adapting the 10-day deadline for responding to a Rule 68 offer to the requirement that the judge approve any settlement of a class action. Cf. Rand v. Monsanto Co., supra. The plaintiff has failed to prove his entitlement to a fee.

The case must be remanded, however, for a redetermination of the fee due class counsel. This is a regrettable outcome in light of the time that the district judge has already devoted to this manner, and we shall therefore suggest (not order) an alternative method of determination that might save time and expense for everyone. The object in awarding a reasonable attorney's fee, as we have been at pains to stress, is to give the lawyer what he would have gotten in the way of a fee in an arm's length negotiation, had one been feasible. In other words the object is to simulate the market where a direct market determination is infeasible. It is infeasible in a class action because no member of the class has a sufficient stake to drive a hard — or any — bargain with the lawyer. So the judge has to step in and play surrogate client. Apparently judges do this with fair success. A recent study finds that "the federal courts appear to have applied, at least implicitly, principles parallel to the market's in determining and awarding attorney fees" in class actions. William J. Lynk, "The Courts and the Market: An Economic Analysis of Contingent Fees in Class-Action Litigation," 19 J.Legal Stud. 247, 260 (1990). But efforts to play the role conscientiously can sometimes backfire, entailing the protracted and ultimately futile expenditure of judicial time that we have reviewed in this opinion.

The judicial task might be simplified if the judge and the lawyers bent their efforts on finding out what the market in fact pays not for the individual hours but for the ensemble of services rendered in a case of this character. This was a contingent fee suit that yielded a recovery for the "clients" (the class members) of $45 million. The class counsel are entitled to the fee they would have received had they handled a similar suit on a contingent fee basis, with a similar outcome, for a paying client. Suppose a large investor had sued Continental for securities fraud, and won $45 million. What would its lawyers have gotten pursuant to their contingent fee contract? We know that in personal-injury suits the usual range for contingent fees is between 33 and 50 percent but we also know that in large commercial litigation with prospects of multimillion dollar recoveries the percentage frequently is tapered — it might be 33 percent of the first million, 25 percent of the next million, and so on down. The class counsel did not present and the judge did not ask for testimony or statistics concerning the fee arrangements in commercial litigation comparable to the present suit. Yet it might be quicker and easier to generate such evidence than it would be to hassle over every item or category of hours and expense and what multiple to fix and so forth. We also remind bench and bar of our approval in the Evans case of Judge Zagel's sampling approach to the auditing of lawyer's hours.

Our emphasis on the contingent factor is not inconsistent with the growing judicial skepticism about the use of risk multipliers in statutory fee-shifting cases in which the plaintiff (not a class) is seeking substantial monetary relief. See, e.g., Burdett v. Miller, 957 F.2d 1375, 1384 (7th Cir. 1992); King v. Palmer, 950 F.2d 771 (D.C. Cir. 1991) (en banc). The problem with risk multipliers in such cases is that the lawyer has two sources of fees: the statutory fee award, and the contingent-fee contract that he negotiates with his client. The latter takes care of the risk of loss and thus makes the inclusion of a risk multiple in the statutory award redundant, assuming as the cases do that the purpose of fee-shifting is to assure competent representation for plaintiffs rather than to make the cost of litigation to the successful plaintiff zero. Pennsylvania v. Delaware Valley Citizens' Council for Clean Air, 483 U.S. 711, 725, 107 S.Ct. 3078, 3086, 97 L.Ed.2d 585 (1987) (plurality opinion); Lenard v. Argento, 808 F.2d 1242, 1247 (7th Cir. 1987); Gekas v. Attorney Registration Disciplinary Commission, 793 F.2d 846, 853 (7th Cir. 1986). In a class action suit, however, there is no contract between lawyer and client — no source other than the judicial award of fees to which the lawyer can look for compensation for risk of loss. The situation is thus parallel to that in an ordinary damages case in which the prevailing plaintiff has no entitlement to a judicial award of fees. The difference of course is that instead of the market's dictating the terms on which the lawyer is retained, the judge must do so. But he must try as best he can to simulate the market, and one way to do this is by obtaining evidence about the terms of retention in similar suits, suits that differ only because, since they are not class actions, the market fixes the terms. It is true that in some of those comparable suits a prevailing plaintiff may be entitled to statutory fees, and that cushion may induce the lawyer to accept retention for a smaller contingent fee than he would demand without the cushion. This is a complication to be weighed against the disadvantages of more conventional, but also more cumbersome, methods of determining a reasonable attorney's fee.

A word finally on the lack of adversary procedure in this case. (On the general issue, see Court Awarded Attorney Fees: Report of the Third Circuit Task Force, 108 F.R.D. 237, 255 (1985).) Since the defendants were out of the case by virtue of their settlement — it being agreed that the lawyers; fees were to come out of the settlement amount — they had no incentive to oppose the request for fees, and they did not. No class member objected either — but why should he have? His gain from a reduction, even a large reduction, in the fees awarded the lawyers would be minuscule. So the lawyers had no opponent in the district court and they have none here. This put more work on the district judge and on us than in a case where there is an adversary to keep the plaintiff and appellant honest. The lawyers are not to blame. They are entitled to seek fees, and entitled to appeal if the district judge cuts them down. But judges in our system are geared to adversary proceedings. If we are asked to do nonadversary things, we need different procedures. When lawyers request fees from a class settlement fund, they are not like adversaries in litigation; they are like artists requesting a grant from the National Endowment for the Arts. Grant-making organizations establish nonadversarial methods for screening applications; perhaps we need something like that for cases like this. The appointment of a special master to advise the court is an obvious possibility, one frequently used in fee matters and especially appropriate in a case such as this that lacks an adversary setting. Of course the master is limited to assessing objective criteria of cost and performance. The judge may have insights to add by virtue of having observed the lawyers in action. But this would not deprive the special master's recommendation of its value to a prompt and accurate determination of the fee to which the class counsel are entitled.

The judgment is reversed, and the matter is remanded to the district court for further proceedings consistent with this opinion.

REVERSED AND REMANDED.