We consider in this case the standards under the common law of Illinois governing the liability of independent auditors for failing to detect fraud by the management of the company audited; the standing of auditors under the RICO statute (Racketeer Influenced and Corrupt Organizations Act, 18 U.S.C. §§ 1961 et seq.) to sue such a company for damages; and — sua sponte, as we are obliged to do — questions of federal subject-matter jurisdiction.
Between 1970 and 1975, managerial employees of Cenco Incorporated engaged in a massive fraud. The fraud began in Cenco’s Medical/Health Division but eventually spread to the top management of Cenco, and by the time it was unmasked the chairman and president of Cenco plus a number of vice-presidents and other top managers were deeply involved. Not all the managers of Cenco were corrupt, however. Seven of the nine members of the board of directors were not in on the fraud, although there is evidence that they were negligent in allowing it to flourish undetected beneath their noses. The fraud was eventually discovered by a newly hired financial officer at Cenco who reported his suspicions to the Securities and Exchange Commission. Cenco’s independent auditor throughout the period of the fraud, the accounting partnership of Seidman & Seidman, either never discovered the fraud or if it did failed to report it.
The fraud primarily involved the inflating of inventories in the Medical/Health Division far above their actual value. This increased the apparent worth of Cenco and greatly increased the market price of its stock (when the fraud was unmasked, the market price plummeted by more than 75 percent). The inflated stock was used to buy up other companies on the cheap. Cenco further benefited from the fraud by being able to borrow money at lower rates than if its inventories had been honestly stated and by getting its insurers to pay inflated claims for inventory lost or destroyed, since Cenco’s insurance claims were based on inflated rather than actual inventory values. Thus, those involved in the fraud were not stealing from the company, as in the usual corporate fraud case, but were instead aggrandizing the company (and themselves) at the expense of outsiders, such as the owners of the companies that Cenco bought with its inflated stock, the banks that loaned Cenco money, and the insurance companies that insured its inventories.
The unmasking of the fraud led to the filing of a class action in federal district court against Cenco, its corrupt managers, and Seidman & Seidman. The class was composed of purchasers of Cenco stock during the period of the fraud, when the stock price was inflated. The suit charged the defendants with having violated various federal securities laws and SEC rules, notably Rule 10b-5, and common law strictures against fraud and related misconduct. Cenco — now under clean management — filed a cross-claim under Rule 13(g) of the Federal Rules of Civil Procedure against its codefendant Seidman & Seidman, alleging that Seidman was liable to Cenco under both federal and state common law for having failed to prevent the fraud by Cenco’s managers.- Seidman filed its own Rule 13(g) cross-claim against Cenco, alleging that Seidman had been one of the victims of Cenco’s fraud and was thus entitled to damages from Cenco under the RICO statute, state common law, and principles of indemnity.
Cenco’s and Seidman’s cross-claims against each other were set for trial together. The trial began on May 8,1980, one day before the last settlement in the class action, which happened to be the settlement with Seidman (for $3.5 million), was submitted to the district judge for his approval. The trial was separated into liability and damages phases; liability, of course, was tried first. At the close of all the evidence but before submitting the case to the jury, the judge granted a directed verdict in fa*452vor of Cenco dismissing Seidman’s cross-claim and a directed verdict in favor of Seidman on those counts in Cenco’s cross-claim that alleged that Seidman had aided and abetted the fraud of Cenco’s managers and had actually joined their conspiracy to defraud.
The case went to the jury on the three remaining counts in Cenco’s cross-claim, alleging respectively breach of contract, professional malpractice (negligence), and fraud. Cenco’s evidence tended to show that in the early stages of the fraud Seidman had been careless in checking Cenco’s inventory figures and its carelessness had prevented the fraud from being nipped in the bud; that as the fraud expanded, Seidman’s auditors became suspicious, but, perhaps to protect the very high fees that Seidman was getting from Cenco (about $1 million a year, which was 70 percent of Seidman’s total billings), concealed their suspicions and kept giving Cenco a clean bill of health at their audit reports; that one partner in Seidman, asked by Ceneo’s general counsel (who was not in on the fraud) whether Seidman suspected anything, answered: “No one suspects fraud. Dismiss that.” Seidman’s evidence tended to show, to the contrary, that Seidman had diligently attempted to follow up all signs of fraud but had been thwarted by the efforts of the large group of managers at all levels at Cenco who were in on the fraud to prevent Seidman from learning about it.
The jury found for Seidman on all three counts. Cenco appeals both from the dismissal of its aiding and abetting and conspiracy counts and from the judgment entered against it on the jury’s verdict, which it contends was based on erroneous instructions. Seidman appeals from the dismissal of the RICO and common law counts in its cross-claim. Several other rulings by the district judge are also before us on these cross-appeals.
We begin our analysis with the directed verdict in favor of Seidman on Cenco’s aiding and abetting and conspiracy counts. Like the rest of Cenco’s cross-claim these counts are based solely on Illinois law. Although the cross-claim purports to base federal jurisdiction on 28 U.S.C. § 1331, the general federal-question jurisdictional statute, and on section 27 of the Securities Exchange Act of 1934, 15 U.S.C. § 78aa, a general conferral of federal jurisdiction over cases arising under the federal securities laws, these statutes are not pertinent to Cenco’s cross-claim since Cenco does not allege any violation of a federal statute or rule. One sentence in Cenco’s reply brief in this court, citing an interstate pollution case, asks us to create a rule of federal common law to govern auditors’ liabilities, but we decline so casual an invitation to so ambitious an exercise of judicial creativity.
The only basis for federal jurisdiction over Cenco’s cross-claim is the concept of ancillary jurisdiction, which allows state-law claims to be litigated in federal court in some circumstances even if the requirements of diversity jurisdiction are not met (as they are not here). A Rule 13(g) cross-claim is within the ancillary jurisdiction of the federal courts if the main claim is within their jurisdiction, as the claim of the class plaintiffs was; and the dismissal of the main claim does not require the dismissal of the cross-claim, at least if the main claim is not dismissed before the trial of the cross-claim. See Federman v. Empire Fire & Marine Ins. Co., 597 F.2d 798, 811 (2d Cir. 1979). Since the class action settlement was not approved by the district court until June 30, 1980, which was after the start of the trial on the cross-claims, and did not become final until upheld by this court a year later sub nom. Helfand v. Fromkin, 645 F.2d 76 (7th Cir. 1981) (mem.), the district court had jurisdiction to adjudicate Cenco’s cross-claim against Seidman even though it was based purely on state law and Cenco and Seidman are not citizens of different states.
Coming to the merits, we can easily dispose of the charge that Seidman aided and abetted the fraud by Cenco’s managers. There is no tort of aiding and abetting under Illinois law or, so far as we know, the law of any other state; all the cases that Cenco has cited with regard to this count are criminal cases. This is not a gap in tort *453law. Anyone who would be guilty in a criminal proceeding of aiding and abetting a fraud would be liable under tort law as a participant in the fraud, since aider-abettor liability requires participation in the criminal venture. E.g., United .States v. Beck, 615 F.2d 441, 448-49 (7th Cir. 1980). The only utility of a separate tort of aiding and abetting in the commission of a tort would be to give plaintiffs’ lawyers one more charge to fling at the jury in the hope that if enough charges are made the jury may accept at least one. In any event, creating a new Illinois tort is something for the Illinois courts or legislature to do rather than the federal courts.
We have much the same reaction to the conspiracy count, though the fact that Seidman does not argue that there is no tort of conspiracy makes us diffident about disposing of it on this ground. Tort law, unlike criminal law, does not punish “inchoate,” which is to say purely preparatory, conduct; for wrongdoing to be actionable as a tort there must be an injury, see, e.g., Town of Thornton v. Winterhoff, 406 Ill. 113, 119, 92 N.E.2d 163, 166 (1950), as Cenco itself argues in defending the dismissal of Seidman’s cross-claim. So while some torts, such as boycotts in restraint of trade, by their nature involve concerted activity, there is no basis in common law thinking for a tort of conspiracy to commit a tort. If there is a conspiracy and it fails, there is no injury and hence no tort liability; if it succeeds, the damages are fully recoverable in an action on the underlying tort. See Prosser, Handbook of the Law of Torts 293 (4th ed. 1971).
In any event, Cenco could not have been prejudiced by the judge’s dismissal of the conspiracy count. The judge allowed Cenco’s fraud count to go to the jury, and Cenco has pointed us to no evidence that pertained only to Seidman’s alleged participation in the conspiracy and not to its alleged participation in the fraud as well. In fact the fraud was a conspiracy, a conspiracy among numerous managers at Cenco; and by allowing the jury to consider whether Seidman had participated in the fraud the judge necessarily allowed the jury to consider whether Seidman had participated in a conspiracy to defraud. If it was error to withdraw the conspiracy count from the jury, which we doubt, it was harmless error.
This brings us to the main issue in the case — whether the district judge gave erroneous instructions to the jury. The challenged instructions relate to the question whether Seidman was entitled to use the wrongdoing of Cenco’s managers as a defense against the charges of breach of contract, professional malpractice, and fraud. Despite the plurality of charges it is one question because breach of contract, negligence, and fraud, when committed by auditors, are a single form of wrongdoing under different names. The contract in question here (really a series of contracts) consists of the letters between Seidman and Cenco outlining the terms of Seidman’s annual retention to audit Cenco’s books. The material part of the letters is the incorporation by reference of general accounting standards which, so far as pertinent to this case, require the auditor to use his professional skill to follow up any signs of fraud that he discovers in the audit. The tort of negligence in the context of auditing is likewise a failure to use professional care and skill in carrying out an audit. And if such care and skill are not used, then the audit reports to the client will contain misrepresentations, either negligent or, if the auditor knows that the representations in the reports are untruthful or is indifferent to whether or not they are truthful, fraudulent.
Because these theories of auditors’ misconduct are so alike, the defenses based on misconduct of the audited firm or its employees are also alike, though verbalized differently. A breach of contract is excused if the promisee’s hindrance or failure to cooperate prevented the promisor from performing the contract. See Restatement (Second) of Contracts § 245 (1979). The corresponding defense in the case of negligence is, of course, contributory negligence. We need not consider to what extent that defense in an auditors’ liability case may *454have been modified by the recent decision of the Illinois Supreme Court in a personal-injury ease to replace contributory by comparative negligence. The court held that this change was not to apply to trials begun prior to the date of the decision, which was June 8,1981. Alvis v. Ribar, 85 Ill.2d 1, 28, 52 Ill.Dec. 23, 35, 421 N.E.2d 886, 898 (1981). The decision whether to make a change of law prospective or retrospective, and if the former as of what date, is a matter of substance within the broad meaning assigned the term in cases (e.g., Walker v. Armco Steel Co., 446 U.S. 740, 751, 100 S.Ct. 1978, 1985, 64 L.Ed.2d 659 (1980)) following Erie R. R. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938). Hence it binds the federal courts in applying state law, whether in a diversity suit or in any other suit where, as here, state law supplies the rule of decision.
Negligence is not a defense to an intentional tort such as fraud. E.g., Mother Earth, Ltd. v. Strawberry Camel, Ltd., 72 Ill.App.3d 37, 52, 28 Ill.Dec. 226, 238, 390 N.E.2d 393, 405 (1979). But a participant in a fraud cannot also be a victim entitled to recover damages, for he cannot have relied on the truth of the fraudulent representations, and such reliance is an essential element in a case of fraud. See, e.g., Broberg v. Mann, 66 Ill.App.2d 134, 139, 213 N.E.2d 89, 92-93 (1965). If the misrepresentation is negligent rather than intentional, contributory negligence plays the same role it would play in an ordinary negligence case. See Prosser, supra, at 706.
The jury instructions in this case stated these defenses accurately, but Cenco contends that the instructions should not have been given, because they related not to Cenco’s conduct but to that of its managers. The judge was aware of the distinction but instructed the jury that the acts of a corporation’s employees are the acts of the corporation itself if the employees were acting on the corporation’s behalf. If this instruction was correct, then the instructions which allowed the jury to consider Cenco’s misconduct as a defense to Seidman’s alleged wrongdoing were proper.
To determine the correctness of the instruction requires us to decide in what circumstances, if any, fraud by corporate employees is a defense in a suit by the corporation against its auditors for failure to prevent the fraud. Illinois precedent allows us to reject one extreme position on this question, which is that the employee’s fraud is always attributed to the corporation by the principle of respondeat superior. This position, which would exonerate auditors from all liability for failing to detect and prevent frauds by employees of the audited company, was rejected in Cereal Byproducts Co. v. Hall, 8 Ill.App.2d 331, 132 N.E.2d 27 (1956), where a company’s independent auditors were held liable for negligently failing to detect embezzlement by the company’s bookkeeper. Auditors are not detectives hired to ferret out fraud, but if they chance on signs of fraud they may not avert their eyes — they must investigate. The references to keeping an eye out for fraud that appear in the accounting standards incorporated (by reference) in the retention letters between Cenco and Seidman would have little point if not interpreted to impose a duty on auditors to follow up any signs of fraud that come to their attention.
But this does not tell us what the result should be if the fraud permeates the top management of the company and if, moreover, the managers are not stealing from the company — that is, from its current stockholders — but instead are turning the company into an engine of theft against outsiders — creditors, prospective stockholders, insurers, etc. On this question the Illinois cases on auditors’ liability provide no guidance. In fact, to our knowledge the question has never been the subject of a reported case. Leeds Estate, Bldg. & Inv. Co. v. Shepherd, 36 (L.R.) Ch.D. 787, 802, 809 (1887), described by Cenco in its main brief as “the one squarely applicable common law decision on accountants’ liability,” is nothing of the sort. The auditor in that case had failed to discover that the company’s manager, by misrepresenting the profits of the company, had caused the company to pay out dividends, directors’ fees, and *455bonuses for himself — all in violation of the charter — as a result of which the company went broke. This was stealing from, not for, the company.
In predicting »how the Illinois courts might decide the present case, we assume they would be guided by the underlying objectives of tort liability. Those objectives are to compensate the victims of wrongdoing and to deter future wrongdoing. With regard to the first, we must refine our earlier statement that the “victim” of Seidman’s alleged laxity was Cenco Incorporated. A corporation is a legal fiction. The people who will receive the benefits of any judgment rendered in favor of Cenco on its cross-claim against Seidman are Cenco’s stockholders, comprising people who bought stock in Cenco before the fraud began, people who bought during the fraud period and either sold afterwards when the stock price fell or continue to hold the stock at a loss, and people who bought after the fraud was unmasked. A judgment in favor of Cenco on its claim against Seidman would not differentiate among these classes, but would benefit every stockholder as of the date of the judgment (or the date when a judgment was anticipated with some precision) in proportion to the number of shares he owned.
Once the real beneficiaries of any judgment in favor of Cenco are identified, it is apparent that such a judgment would be perverse from the standpoint of compensating the victims of wrongdoing. Among the people who bought stock in Cenco before the fraud began are the corrupt officers themselves. To the extent they are still stockholders in the company, they would benefit pro rata from a judgment in favor of Cenco. The other stockholders in this class are innocent in a sense, but of course it is they who elected the board of directors that managed Cenco during the fraud. The people who bought during the fraud period and either sold at a loss or continue to hold at a loss are the plaintiffs in the recently settled class action in which both Cenco and Seidman were defendants. Seidman has already paid $3.5 million to them. Those who continue to own stock in Cenco (as distinct from those who sold at a loss) would receive additional compensation if Cenco prevailed in this action against Seidman. This is not to say they would be overcompensated; but it seems odd that the same shareholders should be able to recover damages from Seidman twice for the same wrong — once directly and once, in this suit, indirectly. Finally, the shareholders who bought after the fraud was unmasked lost nothing. The unmasking of the fraud caused the price of Cenco’s stock to be bid down to reflect not only the true value of its inventories but also any anticipated injury to the company as a result of the fraud.
Because of shareholder turnover, there is always a potential mismatch between the recovery of damages by a corporation and the compensation of the shareholders actually injured by the wrong for which the damages were awarded. It is simply a more dramatic mismatch in this case than usual.
From the standpoint of deterrence, the question is whether the type of fraud that engulfed Cenco between 1970 and 1975 will be deterred more effectively if Cenco can shift the entire cost of the fraud from itself (which is to say, from its stockholders’ pockets) to the independent auditor who failed to prevent the fraud. We think not. Cenco’s owners — the stockholders — hired managers (directly, in the case of the president and chairman, who were both members of the board of directors, indirectly in the case of the others) who turned out to be thoroughly corrupt and to corrupt the corporation so thoroughly that it caused widespread harm to outsiders. If Seidman had been a more diligent auditor, conceivably if it had been a more honest auditor, the fraud might have been nipped in the bud; and liability to Cenco. would make Seidman, and firms like it, more diligent and honest in the future. But if the owners of the corrupt enterprise are allowed to shift the costs of its wrongdoing entirely to the auditor, their incentives to hire honest managers and monitor their behavior will be reduced. While it is true that in a publicly held corporation such as Cenco most share*456holders do not have a large enough stake to want to play an active role in hiring and supervising managers, the shareholders delegate this role to a board of directors, which in this case failed in its responsibility. And not all of Cenco’s shareholders were “little people.” During the period of the fraud Curtiss-Wright Corporation owned between 5 and 16 percent of Cenco’s common stock and had its own accounting firm conduct a study of Cenco’s operations — without discovering the fraud.
Thus, not only were some of Cenco’s owners dishonest (and, to repeat, to the extent they still own stock in Cenco they would benefit from any judgment in Cenco’s favor against Seidman), but the honest owners, and their delegates — a board of directors on which dishonesty and carelessness were well represented — were slipshod in their oversight and so share responsibility for the fraud that Seidman also failed to detect. In addition, the scale of the fraud — the number and high rank of the managers involved — both complicated the task of discovery for Seidman and makes the failure of oversight by Cenco’s shareholders and board of directors harder to condone.
Cenco tries to draw a sharp contrast between an innocent Cenco and a Seidman that was (or so the jury could have found) an intentional tortfeasor. But if Cenco may be divorced from its corrupt managers, so may Seidman from the members and employees of the firm who suspected the fraud. If Seidman failed to police its people, Cenco failed as or more dramatically to police its own.
Furthermore, we must assume that Cenco’s corrupt managers were acting for the benefit of the company, not against it as in the Cereal Byproducts case. The jury was instructed that it could attribute the fraud of Cenco’s managers to Cenco only if it found that the managers had been acting on Cenco’s behalf, and the verdict for Seidman implies that the jury either so found or found that Seidman had not even committed a prima facie breach of duty to Cenco. The former assumption is more favorable to Cenco.
Fraud on behalf of a corporation is not the same thing as fraud against it. Fraud against the corporation usually hurts just the corporation; the stockholders are the principal if not only victims; their equities vis-á-vis a careless or reckless auditor are therefore strong. But the stockholders of a corporation whose officers commit fraud for the benefit of the corporation are beneficiaries of the fraud. Maybe not net beneficiaries, after the fraud is unmasked and the corporation is sued — that is a question of damages, and is not before us. But the primary costs of a fraud on the corporation’s behalf are borne not by the stockholders but by outsiders to the corporation, and the stockholders should not be allowed to escape all responsibility for such a fraud, as they are trying to do in this case.
We need not go so far as to predict that the Illinois courts would hold that in any action by a corporation against its auditors an employee’s fraud intended to benefit the company rather than the employee at the company’s expense will be attributed to the corporation, however lowly the employee. It is true that the lower down the employee is in the company hierarchy, the less likely he is to commit fraud for rather than against the company. But there are overzealous employees at every level — many a corporation has paid heavy damages for antitrust violations committed by low-level sales managers who thought they were acting in the company’s best interests as well as their own — and we think it premature as well as unnecessary to decide that an auditor is never liable for the frauds of loyal but misguided company employees that he could have prevented by taking care. But here the uncontested facts show fraud permeating the top management of Cenco. In such a case the corporation should not be allowed to shift the entire responsibility for the fraud to its auditors.
Hence the challenged jury instruction was proper in the circumstances; and having thus disposed of Ceneo’s cross-claim, we turn to Seidman’s. The indemnity count in Seidman’s cross-claim was dismissed on the basis of this court’s decision *457in Heizer Corp. v. Ross, 601 F.2d 330, 334-35 (7th Cir. 1979), discussed later in this opinion; Seidman has not appealed from this dismissal. The RICO count was based on 18 U.S.C. § 1964(c), which gives a treble-damage remedy to “any person injured in his business or property by reason of a violation of” section 1962. Read in conjunction with section 1961, section 1962 forbids a person or firm that has income from certain types of criminal activity to use that income to acquire any interest in an enterprise affecting interstate commerce. Seidman alleges that Cenco’s corporate acquisitions with stock whose price had been inflated through violations of various federal criminal securities statutes violated RICO. The district court dismissed Seidman’s RICO claim, without reaching the merits, on the ground that Seidman lacked standing to maintain it.
The question whether Congress intended to grant a treble-damages remedy to people or firms injured in the way Seidman was injured, that is, as a consequence of being used as a tool of the criminal enterprise, is apparently one of first impression. The language of section 1964(c) provides no answer, analogies to section 4 of the Clayton Act are forced, and there is no useful legislative history relating to the provision. We therefore ask whether a treble-damages action by auditors of criminal enterprises would contribute to the compensatory and deterrent objectives of RICO. We think it would not, though not because the draftsmen of RICO were concerned with the penetration of lawful enterprises by “organized crime,” a euphemism for what used to be called the Mafia, and there was nothing of that sort here. This court has interpreted the RICO statute, in light of the long list of criminal offenses in section 1962, to forbid penetration of business enterprises by any “pattern of racketeering activity” embraced by that section, whether or not “organized crime” is involved. United States v. Aleman, 609 F.2d 298, 303-04 (7th Cir. 1979). What is critical here, rather, is that “the primary purpose of RICO is to cope with the infiltration of legitimate businesses.” United States v. Turkette, 452 U.S. 576, 101 S.Ct. 2524, 2533, 69 L.Ed.2d 246 (1981). It is presumably on behalf of the owners, perhaps also the customers and competitors, of such businesses that the civil damages remedy was created, and not on behalf of the people who supply office equipment or financial or legal services to criminal enterprises that may be violating RICO. It is unlikely that Congress if it had adverted to the issue would have chosen to create in the wake of every RICO violation waves of treble-damage suits by all who may have suffered indirectly from the violation, especially when many of these would inevitably be, as here, the witting or unwitting tools of the violator. The RICO claim was correctly dismissed.
Seidman also asserted several common law claims (primarily fraud and breach of contract) which the judge dismissed for lack of proof of any injury. Because, as noted earlier, there is no concept of an inchoate tort, wrongdoing that has no impact is not tortious. The existence of impact or injury is therefore relevant to tort liability though the amount of injury is not. But the distinction is a fine one, and when a trial is separated into liability and damages phases the judge must be careful not to dismiss a complaint on liability grounds when the plaintiff was just reserving his damages evidence for the damages phase, as he had been told to do. Since the district judge in this case had, the day after trial began, tentatively approved the settlement under which Seidman agreed to pay $3.5 million to the class, Seidman naturally thought it had no need to present any evidence of injury at trial, and it did not. The district judge should have taken judicial notice of the settlement and held that Seidman’s obligation thereunder was evidence enough of injury to withstand a directed verdict on liability. Apparently he did not do so because he thought Seidman was just seeking indemnification from a codefendant, which this court has held is not allowed in a Rule 10b-5 case. Heizer Corp. v. Ross, supra, 601 F.2d at 334-35. But indemnity is a remedy of one wrongdoer against another; and Seidman’s claim is that it was a *458victim rather than a wrongdoer. It is true that it paid several millions to the class plaintiffs, but it never admitted wrongdoing or was adjudicated a wrongdoer. Therefore, if it can prove that Cenco defrauded it into issuing false audit reports which in turn exposed it to liability to the class plaintiffs, the amount it paid to settle with the class would be a permissible item of damages. (Since Cenco’s cross-claim was dismissed on liability grounds, we need not consider the possible implications of Heizer for Cenco’s damage claim based on its settlement with the class plaintiffs — Cenco, unlike Seidman, being an admitted wrongdoer.)
Heizer shows, incidentally, that creating federal common law to decide the liabilities inter se of violators of the federal securities laws is not out of the question; Heizer was a federal common law case. Its rejection of indemnity as being contrary to the principle that one who is guilty of serious wrongdoing should not be let off scot-free has possible, and it would seem negative, implications for the merits of Cenco’s claim against Seidman. But since the parties here do not seek to use Heizer beyond the indemnity issue, we leave for another day the question of the possible broader implications of that decision.
Since Cenco does not argue that the district judge’s dismissal of Seidman’s state-law counts can be supported on any alternative ground, we must vacate and remand this part of the judgment. But it does not follow that these counts must be retried. They may well be dismissed on remand for lack of federal subject-matter jurisdiction, though this is a question to be decided by the district court in the first instance.
The rule in pendent jurisdiction is that if the federal claim to which the state-law claim is pendent is dismissed before trial, the court will decline jurisdiction over the state-law claim and remit the claimant to the state courts. United Mine Workers of America v. Gibbs, 383 U.S. 715, 726, 86 S.Ct. 1130, 1139, 16 L.Ed.2d 218 (1966). The RICO claim, the only federal claim to which Seidman’s common law counts could be pendent, was not dismissed before trial, so the rule just stated is not strictly applicable. On the other hand, since Seidman is seeking a new trial on those counts, the dismissal of the RICO claim which we affirm today is a dismissal of the main count before trial of the common law counts, more precisely before retrial of them. The position is the same if jurisdiction to decide Seidman’s common law claim is based on the concept of ancillary rather than of pendent jurisdiction. Since the claims to which Seidman’s claim are ancillary, namely the claims of the plaintiffs in the original class action, have all been finally disposed of, there will be no federal question remaining in the case when Seidman’s common law counts are retried; and pendent and ancillary jurisdiction are treated the same so far as the rules governing retention of the nonfederal claim after dismissal or other final disposition of the federal claim are concerned. See Federman, supra.
But the Gibbs rule stated above obviously was meant for cases where the federal claim is dismissed early in the litigation, before there have been significant proceedings on either the federal or the pendent (or ancillary) state claim. Where, as in this case, both claims have been tried once, the considerations of judicial economy that underlie the doctrines of pendent and ancillary jurisdiction support retention of the state claim for retrial. But these considerations are not necessarily decisive. The exercise of pendent jurisdiction is not mandatory. See Gibbs, supra, 383 U.S. at 726-27, 86 S.Ct. at 1139. Since there must be a new trial from scratch in this case, since the litigation has now narrowed down to a dispute between two citizens of the same state over the application of state law to their dispute, since Seidman’s claim against Cenco involves novel and difficult issues of state law, and finally since the judge who presided over the first trial has resigned from the bench, it may well be that the second trial should be in a state court. The district court will have to decide this on remand, balancing the considerations we have just listed against any judi*459cial economies from retaining the matter in federal court, where it began.
The district court on remand will also have to reconsider the order awarding Seidman $76,000 in taxable costs. Seidman was entitled to costs only if it was the “prevailing party.” Fed.R.Civ.P. 54(d). It was the prevailing party on Cenco’s cross-claim, but not on its own. True, “a defendant who successfully fends off a large claim may be awarded costs despite failure to prevail on a counterclaim.” Scientific Holding Co. v. Plessey Inc., 510 F.2d 15, 28 (2d Cir. 1974). The court below thought this precept applicable here because it regarded Seidman’s claims against Cenco as in the nature of defenses to Cenco’s claim. But we cannot accept that characterization. Just as Cenco sought by its cross-claim against Seidman to shift the cost of its settlement with the class to Seidman, so Seidman sought by its cross-claim against Cenco to shift the cost of its settlement with the class to Cenco. And while Cenco’s settlement with the class was larger than Seidman’s — $11 million versus $3.5 million — Seidman had other damage claims against Cenco (and vice versa). Finally, although the court dismissed Seidman’s cross-claim before the case went to the jury, the trial had been as much concerned with Seidman’s as with Cenco’s claims — they are not separable.
If we were affirming the dismissal of Seidman’s cross-claim in its entirety, therefore, we would conclude that neither party had prevailed. But since we are reversing the dismissal of the common law counts in Seidman’s cross-claim, Seidman may yet prevail, so we shall vacate the order taxing costs to abide the event.
Both parties quarrel with the computation of costs by the trial judge, and the court below may want to reexamine that computation should an award of costs become appropriate. We decline to do so ourselves because the entire matter will be academic if the court below should decide not to exercise jurisdiction over Seidman’s
state-law claims and by dismissing them demonstrate that Seidman is no more the prevailing party in this litigation than Cenco.
Last, we reject Cenco’s complaint that the trial court committed reversible error in allowing Seidman’s expert witness to testify without — Cenco argues — an adequate foundation. This was a matter within the trial judge’s discretion. We do not think he abused it but in any event it is not a sufficiently prejudicial error standing alone to warrant undoing a seven-week trial.
The judgment appealed from is affirmed except that the dismissal of the common law counts in Seidman’s cross-claim, and the order taxing costs, are vacated and remanded for further proceedings consistent with this opinion. The parties shall bear their own costs in this court.
So Ordered.