Burlington Northern Railroad Company (Burlington) appeals from an amended judgment of $320,450 in favor of former Burlington employee Silas Counts, following a bifurcated jury trial of Counts' claims under the Federal Employers' Liability Act (FELA), 45 U.S.C. §§ 51-60. Burlington appeals on grounds that Counts' theories for invalidating a release he entered into with Burlington were unprecedented or insufficient as a matter of law, and that Counts was allowed to use a jury that was tainted by prejudicial evidence from Burlington's claims file to determine damages. Because there was error in some of the several grounds for liability considered by the jury in returning a general verdict in favor of Counts, we reverse the entire case and remand it for a new trial. We also find error in the damages trial because of prejudice to Burlington due to failure to segregate or redact the exhibits from the first trial that were submitted to the jury in the second trial or, alternatively, to try the damages phase to a separate jury.

Counts injured his back and left leg in an accident in October, 1984, while he was working as a material handler for Burlington. Counts was walking on a wheel flatcar when a loose grating gave way and he fell. Burlington's records showed the flatcar had been sent for repairs the month before the accident, but the repairs were never done.

Burlington Claims Representative Ward Maser investigated Counts' accident and was assigned to negotiate with Counts. Burlington began advancing money to Counts for living expenses in November, 1984. Maser advised Counts the advances would not continue after settlement negotiations began or when Counts retained an attorney. On a few occasions, Maser told Counts there would be no more advances.

Maser's investigation revealed Counts' injuries were serious, that Burlington's liability for the accident was great, and that Counts was not at fault. Despite this information, Maser told Counts he was not that seriously injured. At Maser's suggestion, Counts attended the FELA trial of another Burlington employee, Bruce Jones. The case resulted in a verdict for Burlington because its negligence was not established. Maser called Counts after the defense verdict and told Counts the result could be the same for him.

Counts represented himself in negotiations with Maser and testified that he was afraid to retain an attorney because his advances would be cut off. Counts obtained from Maser's office a brochure entitled "Injured on the Job?," which explains the employee is free to hire a lawyer at any time, and warns of the risks of litigation. Counts spoke with, but did not retain, three lawyers about his claim prior to signing the release.

Counts had job protection in the form of a guaranteed job with Burlington until he was age 65, following the 1980 merger of the Burlington and Frisco railroads. During the negotiation of Counts' claim, Maser suggested Counts consider taking a sedentary yard office position. Maser contacted Maxine Timberman in the Billings labor relations department and learned that Counts would lose his guaranteed job protection if he were to take the yard job. After discussing this with Counts, they decided to settle the claim "out-of-service," and the yard job idea was rejected.

In February, 1986, Maser told Counts the claims office in his area was closing, that there would be no more offers or advances and that this was the last chance to take Burlington's settlement offer of $138,550. Counts was not told that of this sum, $32,127 was being paid by labor relations as a buy-out of his guaranteed job rather than as compensation for his FELA claim. Counts asked Maser at this time about his lifetime job guarantee and Maser told him to take it up with labor relations. Counts accepted the net settlement offer and signed a release of all claims, which stated he was forever and permanently disabled from returning to work for Burlington. After signing the release, Counts contacted labor relations and was informed that if he signed the release, he would not have a guaranteed job any longer.

On November 7, 1986, Counts filed this FELA suit. Counts' earlier diversity action alleging a state law claim for fraud in the inducement against Burlington was held to be preempted by the FELA. Counts v. Burlington Northern R.R. Co., 896 F.2d 424 (9th Cir.), cert. denied, --- U.S. ----, 111 S. Ct. 54, 112 L. Ed. 2d 30 (1990). The district court granted Counts' motion for summary judgment on the issue of Burlington's liability. The validity of the release and the amount of damages remained to be tried.

During discovery, Burlington produced the contents of the Counts' claims file up to the date the release was signed. The court later ordered Burlington to produce the entire file, including six documents prepared after the release was signed. The file contained a "close-out" sheet prepared by Maser, which contained his assessment that "Counts could have had an easy chance at a large verdict in court. I feel that $138,550 is a good settlement for the BN in that Counts is finally off the books."

Burlington's motion for a bifurcation of the trial to determine validity of the release and then, if necessary, damages, was granted. On the first day of trial, Burlington filed a motion requesting separate jury panels for the two trials. Burlington was concerned about the prejudicial impact of the information contained in its claims file during the damages trial. The motion, renewed several times, was denied.

Burlington's motion for a directed verdict on the validity of the release was denied. The jury returned a general verdict finding the release invalid. Burlington's motion for judgment notwithstanding the verdict was denied. Following the damages trial, the jury awarded Counts $459,000. The district court entered an amended judgment, offsetting the amount previously paid to Counts by Burlington. Burlington timely appeals and we have jurisdiction pursuant to 28 U.S.C. § 1291.

Whether or not this case was submitted to the jury on incorrect theories of law is a question of law, reviewable de novo. See United States v. McConney, 728 F.2d 1195, 1201 (9th Cir.) (en banc), cert. denied, 469 U.S. 824, 105 S. Ct. 101, 83 L. Ed. 2d 46 (1984). The district court's determination on bifurcation of trials is reviewed for abuse of discretion. See Hirst v. Gertzen, 676 F.2d 1252, 1261 (9th Cir.1982); United States v. 1,071.08 Acres of Land, 564 F.2d 1350, 1352 (9th Cir.1977).

A. The Inadequate Consideration Theory for Invalidating the Release

The jury was instructed that they could find the release was not supported by adequate consideration if it found a large disparity between the amount of the release and Counts' actual monetary loss. Counsel for Burlington preserved its objection to this instruction for appeal. Counsel specifically objected to the following language in the consideration instruction: "A release is not supported by adequate consideration is [sic] there is a large disparity between the amount of the release and the actual monetary loss which the injured party eventually incurred."

This instruction does not correctly state FELA law on challenging a release for want of consideration. The Supreme Court has held that adequate consideration exists for a release of FELA claims when there are mutual concessions by the employee and the employer, and if the employee who gives the release receives something of value to which he had no previous right. Maynard v. Durham & S. Ry. Co., 365 U.S. 160, 163, 81 S. Ct. 561, 563, 5 L. Ed. 2d 486 (1961).

In Maynard, the employee claimed that all he received in exchange for his release was his paycheck, to which he had an absolute right, and nothing more. Id. The Supreme Court held there was a genuine issue of fact concerning the presence of consideration for the release by Maynard. Id. In this case, however, according to Counts' own computations, he received $70,423 in addition to his wage entitlement in exchange for signing the release. Under Maynard, the release in this case was supported by adequate consideration and the issue should not have been submitted to the jury. The instruction stated incorrect law and the error was not harmless.

Generally, where there is an error of law in submitting one of a number of theories to the jury, the general jury verdict is tainted because it is impossible for the reviewing court to determine with certainty whether the theory upon which the jury based its verdict is a proper theory. Sunkist Growers, Inc. v. Winckler & Smith Citrus Products Co., 370 U.S. 19, 29-30, 82 S. Ct. 1130, 1136, 8 L. Ed. 2d 305 (1962); Kern v. Levolor Lorentzen, Inc., 899 F.2d 772, 777 (9th Cir.1990). However, this court "may construe a general verdict as attributable to one of several theories if it was supported by substantial evidence and was submitted to the jury free from error." Kern, 899 F.2d at 777, citing Traver v. Meshriy, 627 F.2d 934, 938 (9th Cir.1980).

The factors we must consider in deciding whether to exercise this discretion are: (1) the potential for confusion of the jury; (2) whether the losing party's defenses apply to the count upon which the verdict is being sustained; (3) the strength of the evidence supporting the count relied upon to sustain the verdict; and (4) the extent to which the same disputed issues of fact apply to the various legal theories.

Id., citing Traver, 627 F.2d at 938-39.

We do not find this an appropriate case for the exercise of discretion to avoid reversing the release phase of the case. Unlike Kern and Traver, the facts used by Counts to support the theories of recovery (lack of consideration, fraud in the inducement, mutual mistake of fact and economic duress) were not largely identical. Unlike Benigni v. City of Hemet, 879 F.2d 473 (9th Cir.1988), the theories in this case are not "practically identical." Id. at 478. As was the case in Syufy Enterprises v. American Multicinema, Inc., 793 F.2d 990, 1002 (9th Cir.1986), cert. denied, 479 U.S. 1031, 107 S. Ct. 876, 93 L. Ed. 2d 830 and 479 U.S. 1034, 107 S. Ct. 884, 93 L. Ed. 2d 838 (1987), this case involves claims sharing common issues of fact, but also involving distinctive legal and factual components. Furthermore, as will be discussed below, of the four theories that were submitted to the jury, only one was submitted to the jury free from error.

Contrary to Counts' argument, the failure of Burlington to request a special verdict in this case does not affect the appealability of this issue. This issue involves legal claims rather than factual claims and does not involve the sufficiency of the evidence or calculation of damages. Compare McCord v. Maguire, 873 F.2d 1271, 1274 (9th Cir.), amended, 885 F.2d 650 (9th Cir.1989); Landes Constr. Co. v. Royal Bank of Canada, 833 F.2d 1365, 1373 (9th Cir.1987).

B. The Fraud Theory for Invalidating the Release

FELA cases permit invalidation of a release on the ground of fraud in the inducement. Callen v. Pennsylvania R.R. Co., 332 U.S. 625, 630, 68 S. Ct. 296, 298, 92 L. Ed. 242 (1948); see also South Buffalo Ry. Co. v. Ahern, 344 U.S. 367, 372, 73 S. Ct. 340, 342, 97 L. Ed. 395 (1953); Graham v. Atchison, T. & S.F. Ry. Co., 176 F.2d 819, 826 (9th Cir.1949). The employee in a FELA case bears the burden of showing a release is invalid on the ground of fraud. Callen, 332 U.S. at 630, 68 S.Ct. at 298.

The jury was instructed in this case on the five elements of fraud and that it could find the release invalid if Counts was induced to sign the release by deliberately false and material statements of the railroad. The jury was also instructed it could find the release invalid based upon fraud if its execution was materially induced by Burlington's failure to explain to Counts the full scope of FELA rights and benefits. This second, alternative instruction on fraud was based upon Apitsch v. Patapsco & Back Rivers R.R. Co., 385 F. Supp. 495, 505-07 (D.Md.1974), a case involving a railroad that was systematically soliciting defective releases by giving deceptive descriptions of FELA rights versus worker's compensation rights.

While there is evidence of fraud and overreaching in this case, it was error to instruct the jury that Counts could show fraud either by proving the elements of fraud or by simply showing failure to fully explain his FELA rights. The second fraud instruction relieved Counts of his burden to prove the elements of fraud. Apitsch is not binding on this court and is factually distinct from this case. A failure to fully explain FELA rights to an employee is a consideration in the fraud determination, but does not automatically constitute fraud. The "Apitsch " instruction relieved Counts of the burden of establishing fraud in this case and this error was not harmless.

C. The Mutual Mistake Theory for Invalidating the Release

A FELA release may be set aside on the basis of mutual mistake of fact in executing the release. Callen, 332 U.S. at 630, 68 S.Ct. at 298; Graham, 176 F.2d at 823-25. The burden is on the employee to show invalidity based upon mutual mistake. Callen, 332 U.S. at 630, 68 S.Ct. at 298. Burlington argues that there was no mistake as to the job guarantee associated with the potential yard office job and if there was any mistake at all, it was not the type of mistake recognized in FELA case law as requiring invalidation of a release. We disagree.

The release signed by Counts stated he was forever and permanently disabled from returning to work for Burlington. The document therefore released not only his personal injury action, but also his right to work for Burlington. The testimony at trial showed that both Maser and Counts were operating under the assumption that Counts would give up his guaranteed status with Burlington if he took the yard job. However, the testimony of Mike Crilly, a general claims manager for Burlington, indicated that this was not necessarily the case. Counts testified that he did not believe he would have signed the release and given up his employment if he knew he could work a guaranteed job in the yard office.

Therefore, an issue of fact was raised as to mistake in the execution of the release. While this case does not involve confusion or mistake as to the nature and extent of the claimant's injuries, the mistake here did involve the contents of the release and went to the essence of the employment damages portion of the contract. The release combined the personal injury and employment damages and the jury could consider whether the release was executed under a presently existing mutual mistake of fact. FELA case law does not rule out this type of mutual mistake. See Callen, 332 U.S. at 628-30, 68 S.Ct. at 297-98; Graham, 176 F.2d at 823-26.

D. The Economic Duress Theory for Invalidating the Release

Burlington argues that the issue of economic duress was erroneously submitted to the jury because economic duress has never been recognized as a ground for invalidating a FELA release and, alternatively, because economic duress was not established in this case. Although Burlington offered the jury instructions on this issue, it did so only after the court had denied its motion for a directed verdict, which encompassed this issue. Therefore, Burlington properly preserved this issue for appeal. See City of St. Louis v. Praprotnik, 485 U.S. 112, 119-20, 108 S. Ct. 915, 921-22, 99 L. Ed. 2d 107 (1988).

The economic duress theory was presumably rooted in the threats by Burlington to cut off Counts' advances at various times during the negotiations between Maser and Counts. The parties discussed at length the elements of economic duress and the evidence in the case. However, our review of the case law demonstrates that this theory is not a separate ground for invalidating a FELA release.

FELA releases may be set aside on the grounds of fraud, Maynard, 365 U.S. at 161, 81 S.Ct. at 562; Dice v. Akron, Canton & Youngstown R.R. Co., 342 U.S. 359, 369, 72 S. Ct. 312, 318, 96 L. Ed. 398 (1952); Callen, 332 U.S. at 630, 68 S.Ct. at 298; Graham, 176 F.2d at 824, lack of consideration, Maynard, 365 U.S. at 161, 81 S.Ct. at 562; and mutual mistake, Callen, 332 U.S. at 630, 68 S.Ct. at 298; Graham, 176 F.2d at 824. The Supreme Court has also stated that "[u]ntainted by fraud or overreaching, full and fair compromises of FELA claims do not clash with the policy of the Act." South Buffalo Ry. Co., 344 U.S. at 372, 73 S.Ct. at 343. This court has also acknowledged that a release should not be upheld if " 'any element of fraud, deceit, oppression, or unconscionable advantage is connected with the transaction.' " Graham, 176 F.2d at 826, citing Kansas City M. & B. Ry. Co. v. Chiles, 86 Miss. 361, 38 So. 498 (1905). However, the fact remains that economic duress has not been recognized as a separate theory for invalidating a FELA release. In the interest of uniform application of FELA law, see Dice, 342 U.S. at 361, 369, 72 S. Ct. at 314, 318; Graham, 176 F.2d at 824, we reject the economic duress theory as a separate ground for invalidating a FELA release. This theory should not have been presented to the jury.

At a telephonic pretrial conference held two months before trial, the parties were informed by the court that the bifurcated issues of the validity of the release and damages would be tried to the same jury. Neither party objected. However, on the first day of trial, Burlington filed a motion requesting separate jury panels for the two trials. This motion, renewed several times, was denied. Counts argues that because Burlington did not object to the procedure of using the same jury at the time of the preliminary pretrial conference, or when the final pretrial order was entered one week before trial, it has waived any grounds for appeal of this issue. Burlington argues that it made the motion at the beginning of the release trial, "as soon as BN realized the full prejudicial impact of the evidence Counts intended to introduce in the release trial."

Burlington was ordered in July, 1989, to produce six documents from its claims file, including the settlement data "close-out" sheet containing Maser's assessment of the value of the case in comparison to the settlement. However, Burlington argues that even after the pretrial order was filed, the parties were still disputing what evidence would be admitted in the release trial. Burlington argues that it cannot be held to have waived its objection to the use of the same jury before it knew the evidence that Counts intended to introduce in the release trial and how prejudicial that evidence would be.

We hold that under the circumstances, Burlington preserved its objection to the use of the same jury for both phases of the trial. Early on, there may have been no reason to suspect separate juries might be called for. While Burlington's objection to using the same jury for both trials was made at the eleventh hour, it was made prior to the start of the first trial. However, as discussed below, Burlington's dispute in reality is also based on the court's ruling that the exhibits from the first trial would be submitted to the jury in the second trial. We must consider whether this objection was preserved.

Early on in the damages trial, there was an indication that under the circumstances, the damages trial might better be held before a different jury. The settlement data sheet, containing Burlington's internal exposure assessment, was admitted during the first trial. Counsel for Counts underscored this assessment in the damages trial, noting in opening argument that the jury, "like the BN, will feel that Silas could have had an easy chance at a large verdict in court." During closing argument, Counts' counsel reminded the jury that "[t]he railroad admits that [Counts] would be entitled to a large verdict in court." These statements were not objected to by Burlington.

At the conclusion of closing arguments in the damages trial, the court informed the jury that the exhibits introduced in the damages phase, "as well as any of the pertinent exhibits in the first part of the trial," would be permitted in the jury room. The court then reminded the jury of the instruction that the jury was "to disregard all of the exhibits in the former case with the exception of the medical records."

This prompted the following response from Counts' counsel:

It is my understanding that the Court has stated that it is contemplating not giving the jury all of the exhibits that have already gone into evidence and that they have seen in the--in validity of the release case. And it is my position that I think they ought to have all of the exhibits.

It is going to be very difficult to break out those exhibits from what contains medical and what contains mitigation of damages, what contains admissions by the railroad. And if we try to take those out now quickly, it's just--I think it is a problem.

And I just see no reason not to give them all of the exhibits which are already in evidence. Because they are relying on them and our proof in this damage case relies on those exhibits extensively....

Your Honor, we can do it quickly and we can do it completely. We can break out the reports of Smith and Noteboom and the one hospital report. When Zander now talks about the advances for and realizes it has no relevancy at all, no bearing at all on this damage case, Ward Maser's letter of June 17, 1985, when there is talk about Silas' calling Hoyt and the other lawyer and all of this, what relevancy does that have. And it's prejudicial. It is prejudicial to the defendant. [Emphasis added.]

And then you get to the mitigation. Mitigation is what was or was not done really after the release. Because he wasn't even released to go to work. So I think we can give them what the Court has already instructed them that they will consider. And I think that will--it should settle the issue.

Okay. I'm going to backtrack a little and I think I'll give them all of the exhibits. I have given that last instruction, it's a cautionary instruction that they are not to consider them. They have already seen them. They have already deliberated over them, and I think to go back and try to segregate it out probably would be an unjust comment on the evidence and unduly--if they remember some of them, they don't have them now, they are going to wonder why. So I'll give them ... all of [the] exhibits.

It appears that the court was initially not going to allow exhibits from the first trial, such as the settlement data sheet, to go to the jury in the second trial. Counsel for Counts convinced the court otherwise, and while Burlington did not make a clear objection to this change in procedure, its objections, couched in terms of relevancy and prejudice, are on the record. We hold that this issue has been preserved for appeal.

Under the circumstances of this case, the district court abused its discretion in permitting all of the exhibits from the first trial to go to the jury in the second trial. The court's cautionary instruction to ignore any of the nonmedical evidence was inadequate to cure the prejudice. The settlement data sheet was not reintroduced in the damages portion of the case and its submission to the jury was inconsistent with the court's cautionary instruction. No attempt was made to segregate or redact the documents from the first trial that went to the jury in the second trial. Even if we assume that Burlington's internal exposure assessment memorandum was admissible in the first trial, this information concerning what Burlington thought of the case was not relevant in the damages phase and was very prejudicial.

On remand, the district court should carefully screen and redact any exhibits from the first trial that are used in the damages trial and carefully instruct the jury to disregard the evidence presented in the first trial. The jury's mere exposure to the evidence in the first trial does not necessarily require using a separate jury for the damages trial. Redacting the exhibits and instructing the jury may minimize any prejudice resulting from the evidence presented in the first trial. However, if prejudice would result, the district judge should use a separate jury for the damages trial.

The validity of the release must be retried because the issue was submitted to the jury under erroneous theories of law. We do not find this a proper case in which to exercise our discretion to attribute the general verdict to the mutual mistake theory. The damages issue must be retried either with a separate jury or with careful screening and redacting of the exhibits that are submitted to the jury. We have considered the other contentions of the parties and find no additional discussion is necessary.

REVERSED AND REMANDED.