This case reaches us for the fourth time. See Nemir v. Mitsubishi Motor Sales of Am., Inc., 6 Fed. Appx. 266 (6th Cir.2001) (unpublished) (per curiam) ("Nemir I"); In re. Michael A. Nemir, No. 01-2260 (6th Cir. Oct. 12, 2001) ("Nemir II"); Nemir v. Mitsubishi Motor Sales of Am., Inc., No-02-1780 (6th Cir. Feb. 3, 2003) ("Nemir III"). Plaintiff-Appellant Michael A. Nemir brought a diversity products liability suit against Defendants-Appellees Mitsubishi Motors Corporation and its parent company, Chrysler Corporation, alleging that his seatbelt's failure to latch caused injuries that left him brain-damaged and paralyzed following a car accident. After the jury rejected Nemir's claims, the district court entered judgment for, and awarded costs to, Mitsubishi, and Nemir appealed both decisions. For the following reasons: (1) the district court's judgment for Mitsubishi is REVERSED and the case is REMANDED for a new trial; (2) the district court's award of fees to Mitsubishi is VACATED; and (3) the case is to be assigned to a different district judge.

Given this case's dense history and its relevance to today's appeal, we will narrate the background in some detail. On December 14, 1993, near his home in Maryland, Nemir drove his car into a tree on the side of the road, causing his head to strike a pillar on the car's passenger side. He was found lying in the car's backseat, and the driver's side seatbelt was unlatched. The accident damaged Nemir's brain stem, leaving him wheelchair bound, cortically blind, and unable to continue practicing medicine. On November 22, 1996, Nemir filed suit against Mitsubishi, the maker of the 1991 Dodge Stealth that Nemir was driving, and Chrysler, of which Mitsubishi is a subsidiary. (For simplicity's sake, we will refer to the defendants collectively as "Mitsubishi.")

Nemir's complaint alleged that at the time of the accident, he had been wearing his seatbelt—the Takata 52—but that it had only "partially latched," such that it appeared to be properly fastened when in fact it was still prone to unlatching. Seeking both compensatory and punitive damages, Nemir asserted that the seatbelt was defectively designed because of its potential to unlatch, and because of the vehicle's failure to alert the driver that the seatbelt was not fully latched. The failure to warn and punitive damages claims were subsequently dismissed, and we affirmed those dismissals in Nemir I, 6 Fed.Appx. at 277.

Prior to trial, Nemir retained an expert witness, Dr. Thomas Horton, the former Director of Engineering for Takata, Inc., the seatbelt's manufacturer. Horton planned to testify that Nemir's seatbelt suffered from a design defect, known as "partial-latching," which caused the seatbelt to unlatch during the accident. However, the district court ruled that Horton was unqualified to testify as an expert witness, due to purported deficiencies in his methodology.

In December 1994, based on complaints it had received, the National Highway Traffic Safety Administration (NHTSA) opened an investigation into "[a]lleged defect[s] of Takata seat belt buckles includ[ing] failure to latch, unlatch, or remain latched." NHTSA requested information from Chrysler, including complaints received from its customers, pertaining to "all model years 1986 through 1991 Dodge Colt[s] and any other passenger motor vehicles equipped with Takata [52] series." (In 1996, owing to a manufacturing defect, some Takata 52 seatbelts—but not the seatbelt at issue in our case-were recalled.)

Aware of this investigation, Nemir sought discovery from Mitsubishi of "[a]ll documents containing, referring to, or relating to any claims, assertions, or reports by an individual or entity that any Dodge vehicle using the same or similar seatbelt system wherein a front seat belted passenger received a serious or fatal injuries." On October 31, 1997, Mitsubishi responded that it "has no documents responsive to this request." On September 11, 1998, Mitsubishi reaffirmed that it "currently is unaware of any materials responsive to this request." When Nemir further pressed the issue at a hearing held on November 17, 1998, Mitsubishi insisted that "[w]ith respect to all of the requests that are contained in the order, we produced what documents exist and what documents were able to be located. . . we searched, as we did for all of the other requests, and found nothing that was responsive to the request."

Nemir eventually filed a Freedom of Information Act request with NHTSA, and subsequently obtained 200 to 300 complaints of partial latching in Mitsubishi and Chrysler automobiles. The documents included evidence of "[Mitsubishi's] own employees going out and confirming partial latch in the field." When asked by the district court whether Mitsubishi had any additional documents similar to the ones that Nemir obtained from the NHTSA, Mitsubishi's counsel told the court that "I don't know of any other documents, Your Honor."

The district court ordered Mitsubishi to search again for additional complaints in its possession. On April 13, 1999, Mitsubishi admitted that it possessed, but had not previously disclosed, approximately 4,000 customer contact reports documenting complaints, along with approximately 25,000 warranty claims about the partial latching of the seatbelts. Mitsubishi—who claimed to have previously withheld these documents because "they do not show claims that partial latching occurred in the manner alleged to have occurred in this case"—continued to maintain that the documents should not be disclosed to Nemir.

Although the district court disagreed, it also prohibited Nemir from "conduct[ing] further discovery by attempting to interview or contact complainants listed in these documents." Moreover, the district court scheduled argument on Mitsubishi's motion for summary judgment for May 21, 1999—just seventeen days after the 29,000 documents were ordered disclosed. On July 30, 1999, the district court granted summary judgment to Mitsubishi, holding that Nemir had failed to present evidence by which a reasonable jury could find (under Maryland law, which the parties agree controls) that the buckle suffered from a design defect. Nemir v. Mitsubishi Motors Sales of America, 60 F. Supp. 2d 660, 673-77 (E.D.Mich.1999).

In Nemir I, we reversed. First, we noted that the district court had erroneously stated that Horton had not examined the actual buckle in Nemir's seatbelt—an omission that the district court had dubbed "inexplicable"—when in fact the record revealed that Horton had testified that he had examined the actual buckle and that it had partially latched twice out of twenty attempts. Id. at 274. Second, we held that the district court erroneously excluded the entirety of Horton's testimony, and specifically held that "Horton could use [the abovementioned testing] to testify that the partial latch of Dr. Nemir's seatbelt caused the damage in question." Id. at 275. Third, we held that Nemir had produced sufficient evidence for a jury to conclude that a design defect rendered the buckle unreasonably dangerous. Id. at 277. We remanded for trial.

We also ordered the district court to afford Nemir sufficient time and opportunity to investigate the consumer complaints that Mitsubishi had turned over at "the eleventh hour." Id. at 275. Although Mitsubishi argued that none of the consumer complaints contained the words "partial latch," and thus were not discoverable, we rejoined that "the term `partial latch phenomenon' is a term of art employed by engineers, seatbelt designers, and lawyers. The layman with a concern that his seatbelt will not remain fastened will not use this term." Id. at 276. After reviewing the sample of 188 complaints that Nemir had included in the appellate record, we noted that "[a] complete review of the record indicates that phraseology analogous to `intermittent partial latching' occurred 188 times in the customer complaints submitted as evidence. To summarily dismiss this evidence without affording the plaintiff time to investigate and further elicit information regarding the exact nature of these incidents is clear error." Id.

On remand, the district court issued a new protective order. The order limited Nemir's discovery to "the 188 complainants referred to by" our decision in Nemir I. Although Nemir argued to the district court that Nemir I singled out 188 complaints only because that was the number of complaints that had been included in the appellate record—in other words, that we had not manually sifted through each and every of the 29,000 complaints and determined that only 188 were relevant—the district court nonetheless refused to allow Nemir to investigate the other 28,000—plus complaints, insisting that "I'm going to stay literally with what the panel said."

In addition, the order prohibited Nemir from directly contacting any of Mitsubishi's consumers outside of defense counsel's presence. Of the 188 customers that Nemir was permitted to contact, he was able to locate only thirteen. Prior to the depositions of those witnesses, Nemir asked the district court for permission to contact additional complainants beyond the original 188; the district court told Nemir that he should conduct the first set of depositions, and then "if you need to expand this discovery further, come back to me on it." Nemir never did so.

Nemir did ask the district court to sanction Mitsubishi for its refusal to disclose the 29,000 documents. He appended to his motion evidence that the improper withholding of relevant discovery was not an uncommon practice of both Mitsubishi and its lead counsel. Nemir attached an affidavit from a plaintiff's attorney, who had litigated against Mitsubishi in a prior case and averred that "Mitsubishi [had] repeatedly refused to turn over properly requested information and respond to interrogatories," and "did not produce certain materials until it was threatened with daily fines." Also submitted by Nemir was a copy of the decision in Baker v. General Motors Corp., 159 F.R.D. 519 (W.D.Mo.1994), reversed on other grounds, Baker v. General Motors Corp., 86 F.3d 811, 817 (8th Cir.1996), in which Mitsubishi's lead counsel had defended General Motors against lawsuits arising from electrical fires in certain vehicles. There, the court noted that "[e]very time a request was made for a document, General Motors took it upon itself to determine what scope it thought was relevant," id. at 523, (emphasis in original) the court found numerous examples "of the types of critical documents that were `dumped' on plaintiffs the week before trial," id. at 522, and the court concluded that there was a "deliberate, willful policy on the part of General Motors to stonewall discovery as much and as long as the patience of the Court would tolerate." Id. at 524.

When Nemir pressed for sanctions at a hearing conducted on March 16, 2001, the district court refused to "reach the motion for sanctions at this point," and complained that "I just don't want to be bothered with sanctions." Instead, the district court decided to "stay that matter and return to it when the time is appropriate." The district court never revisited the motion, and also prohibited Nemir from informing the jury that Mitsubishi had initially withheld the consumer complaints.

In addition, despite our holding in Nemir I that its namesake had "demonstrate[d] a triable issue of fact with regard to the design defect condition and the unreasonably dangerous condition of the product," the district court again ruled that "[n]othing in the record before [us] indicates that Dr. Nemir's seatbelt caused an unreasonably dangerous condition." Convinced that the district court had ignored our prior decision to the contrary, Nemir sought from us a writ of mandamus compelling the district court to "submit to the jury the issue of `inherently unreasonable risk' in accord with [Nemir I]." The district court responded to the petition for mandamus by assuring us that its ruling was "subject to change based on the proofs that are presented." Nemir II, slip op at 5. Relying on the district court's assurances, we denied the petition for mandamus, concluding that "[t]he case is proceeding to trial in accordance with our determination that the plaintiff has presented sufficient proof to create a question for the jury regarding his design defect claim." Nemir II, slip op. at 2.

Notwithstanding our instructions in Nemir II, the district court, pursuant to Federal Rule of Evidence 706, appointed Eddie P. Cooper and Lindley Manning as experts to help the court "determine whether the seatbelt buckle presented an inherent unreasonable risk of danger to the plaintiff." Cooper-who was employed by a consulting company that had performed significant work for Mitsubishi and Takata—concluded that the seatbelt was not inherently dangerous. Manning, echoing Horton's conclusions, testified that he too had been able to partially engage the Nemir buckle during laboratory testing. Manning also opined that "the defect that we found is unreasonably dangerous. . . . The fact that it would happen even once or a few times makes it unreasonably dangerous." On May 22, 2002, relying on the testimony of Cooper and discounting that of Manning, the district court again held that the seatbelt "does not present an inherently unreasonable risk of danger," thereby reiterating its unwillingness to instruct the jury on Nemir's theory of strict liability. See Nemir v. Mitsubishi Motors Corp., 201 F. Supp. 2d 779, 783 (E.D.Mich.2002).

The trial was fueled by three disputed issues. First was whether Nemir was wearing his seatbelt at the time of the accident. Nemir testified generally that he almost always wore his seatbelt while driving, and specifically that he was wearing it at the time of the accident. In support of Nemir's testimony, another of his experts, Dr. Joseph L. Burton, testified that cotton fibers resembling those from Nemir's jacket were embedded along the edge of the seatbelt; that the seatbelt appeared to have been stretched; and that the buckle had "loading marks" consistent with wear. Mitsubishi submitted evidence indicating that Nemir's injuries had diminished his memory of the accident, and produced expert witnesses who down-played the conclusiveness of Nemir's fiber evidence and surmised that Nemir's seatbelt lacked the markings that would have appeared had it been worn during the accident.

Second was the question of the cause of Nemir's injuries. Burton testified that had he been belted, Nemir's collision with the pillar would have been much slower, and Nemir would have suffered milder injuries. Mitsubishi's expert, Dr. Charles Hatsell, concluded that the injuries to Nemir's brain stem resulted from head rotation that even a fully functioning seatbelt would have failed to prevent. Through cross-examination, Nemir's attorney attempted to impeach Hatsell by confronting him with a contradictory excerpt from a book that had been co-authored by Dr. Werner Spitz, another of Mitsubishi's experts. The district court prohibited this questioning, however, concluding that because Dr. Spitz would be testifying later in the case, Nemir should simply elicit the information from Dr. Spitz when he testified.

The next day, when Nemir's attorney attempted to question Dr. Spitz about his book, the district court sustained Mitsubishi's objection to the questioning, as illustrated by the following exchange:

Nemir's Counsel (questioning Dr. Spitz): [Your book] talks about the brain stem contusions and I'm looking at the chart that Dr. Hatsell has that we all saw . . .

Mitsubishi's Counsel: Your Honor, I would object to any questioning of this witness on the issue of brain injury and injury causation. Although he is qualified, he's not given any testimony on direct examination on injury causation, and we're not offering him for that.

Nemir's Counsel: Well, I—I specifically . . .

Judge Feikens: If he's not being offered for that, then why are you going into it?

Nemir's Counsel: Because you asked me to do it, your Honor. I asked to be able to cross-examine with his book yesterday with Dr. Hatsell, and you said the proper place to do it was with this witness.

Judge Feikens: On . . . on brain injury even though he's not being offered from that reason?

Nemir's Counsel: Your Honor, I tried to cross-examine to show this man had written—or shown that Hatsell's opinion is impossible. I tried to cross-examine. You said, no, I should do it with...

Judge Feikens: I'll sustain the objection.

Needless to say, the impeaching passage never reached the jury.

Third was whether the seatbelt was capable of partially latching, such that it could malfunction during an accident. Dr. Horton testified that the seatbelt buckle was unreasonably dangerous because it allowed for "partial engagement." But he was unable to explain the basis for his conclusion because the district court prohibited Horton from testifying that he was able, twice out of twenty attempts, to create a partial latch of Nemir's seatbelt buckle, opining that Horton's method-in which he manipulated the buckle at varying speeds and angles—was scientifically unsound. Nemir, 200 F.Supp.2d at 778. The district court also prohibited Horton from testifying about causation, reiterating its prior concern that Horton "does not demonstrate how he arrived at his conclusion that partial latch caused Nemir's injuries or how he eliminated other possibilities, chief among which is the possibility that Nemir was not wearing a seat belt." Nemir, 200 F.Supp.2d at 776. Moreover, although Federal Rule of Evidence 706 provides that a court-appointed witness "may be called to testify by the court or any party," the district court refused to allow Nemir to call Manning at trial.

Taking advantage of the gaps in Horton's testimony and the absence of Manning's testimony, Mitsubishi argued to the jury, among other things, that partial latching had never been demonstrated, that Horton's testimony was "sheer speculation," and that "[y]ou can't prove your case with just the words of a hired expert." Moreover, although the other court-appointed witness, Cooper, had testified prior to trial that were he a manufacturer that had received complaints of unlatching seatbelts, he would want to examine the buckles in question so that he could pinpoint the cause, Mitsubishi claimed that it had not performed any such tests. The district court prohibited Nemir from introducing—in order to explain why there was not more evidence of partial engagement during actual use—evidence that Mitsubishi had failed to test the buckles flagged by its customers. And Mitsubishi's expert testified that it was impossible to achieve partial engagement in actual use.

The jury's Verdict Form posed three questions:

1. Was Nemir wearing his seatbelt at the time of the accident?

2. If yes, was the seatbelt defective?

3. If yes, was the defect the proximate cause of Nemir's injuries?

The jury answered that Nemir was not wearing his seatbelt at the time of the accident, and therefore did not reach the other questions. On July 4, 2002, the district court entered judgment for Mitsubishi.

On June 12, 2002, Nemir appealed, arguing that the district court erred in excluding portions of Horton's testimony, restricting his access to customers who had filed complaints, refusing to instruct the jury on strict liability, refusing to allow him to call Dr. Manning, and inhibiting his impeachment of Dr. Hatsell. After Nemir submitted his appellate brief, the district court ordered Nemir to post a bond of $20,000 to secure the costs of the appeal.

In the meantime, Mitsubishi asked the district court to award it $48,000 in costs from Nemir. Pursuant to 28 U.S.C. § 1920, the Clerk of the district court entered an order awarding $24,000, refusing to award Mitsubishi all of its request for court reporter and duplication fees. Indeed, the Bill of Costs Handbook states that the latter are "not recoverable within the discretion of the taxation clerk unless counsel has previously secured an order authorizing the recovery of these costs," which Mitsubishi had not. Mitsubishi petitioned the district court to reinstate $21,000 of the $24,000 in denied costs.

On February 3, 2003, we reversed the district court's setting the surety bond at $20,000, holding that a $1,000 bond "will be adequate to protect the defendant's interests." Nemir III, slip op. at 1. The day after we issued our opinion, the district court vacated the Clerk's award of costs and awarded Mitsubishi the entirety of its original request—that is, $3,000 more than even Mitsubishi had asked to be added to the Clerk's initial award. Nemir separately appealed the district court's award of costs.

Nemir asserts that the district court erred in: (1) refusing to instruct the jury on Nemir's theory of strict liability; (2) limiting Nemir's discovery; (3) excluding an array of evidence; and (4) awarding excessive costs to Mitsubishi.

Nemir contends that the district court's refusal to instruct the jury about his theory of strict liability violated the mandate that we issued in Nemir I. As we explained above, in originally granting summary judgment to Mitsubishi, the district court held that Nemir had failed to demonstrate that the seatbelt suffered from an unreasonably dangerous "design defect" that could underlie a cause of action based on strict liability. In Nemir I, we reversed and identified numerous items of evidence that "demonstrate[] a triable issue of fact with regard to the design defect condition and the unreasonably dangerous condition of the product." Id. at 276.

Our prior holding that Nemir had created a triable issue as to strict liability bound the district court. "[U]pon remand of a case for further proceedings after a decision by the appellate court, the trial court must proceed in accordance with the mandate and the law of the case as established on appeal. The trial court must implement both the letter and the spirit of the mandate, taking into account the appellate court's opinion and the circumstances it embraces." United States v. Moored, 38 F.3d 1419, 1421 (6th Cir.1994). As we made clear in Nemir I, the district court was required to allow Nemir to present his strict liability theory to the jury.

Mitsubishi maintains, however, that our prior decision did not reach this issue, contending that "[Nemir] is under the mistaken belief that this Court's opinion established the law of the case on the issue of which Maryland law standard of proof would apply, `inherently unreasonably dangerous' strict liability or the risk/utility . . . balancing test. However, it is clear from the language of [Nemir I] that this Court did not address that issue. . . ." Mitsubishi's argument misrepresents the repeated and unambiguous language of our prior decision. We stated plainly, more than once, that Nemir had met his burden on the unreasonably dangerous standard. See Nemir I, 6 Fed.Appx. at 276 ("Plaintiff demonstrates a triable issue of fact with regard to the design defect condition and the unreasonably dangerous condition of the product."); id. ("The plaintiff's evidence could lead a reasonable jury to conclude that the product poses an inherently unreasonable risk."); id. at 277 ("[Horton's] testimony creates a jury issue with regards to whether or not the Takata 52 seatbelt is `unreasonably dangerous' under [Maryland law]."). The district court ignored our clear and repeated instructions regarding the submission to the jury of Nemir's theory of strict liability, and Mitsubishi's argument to the contrary is plainly without merit.

Nemir also argues that even if the strict liability instruction were not compelled by Nemir I, the district court erred by basing its strict liability decision on the testimony of Cooper, given that his employer counted both Mitsubishi and Takata among its past clients, and by misapplying the law governing unreasonably dangerous products. Given that the district court's decision to preclude the strict liability instruction was erroneous on a more basic level-that it violated our mandate in Nemir I—we need not address these arguments.

Nemir's claims that the district court improperly limited discovery also descend from our prior review of this case. In Nemir I, as part of our reversal of the district court's grant of summary judgment, we noted that "the district court perfunctorily disregarded hundreds of pages of consumer seatbelt complaints. This evidence was released to [Nemir] in the eleventh hour, after years of discovery disputes over whether or not such complaints were discoverable material." Nemir I, 6 Fed. Appx. at 275. Indeed, we explained that "[t]his issue potentially impacts the safety of millions of American consumers and [Nemir] should be granted sufficient time to fully investigate [his] case." Id. at 276. Nemir argues that the district court failed to do so in a number of ways.

Nemir contends that the district court erred in prohibiting Nemir's counsel from unilaterally contacting any of Mitsubishi's complaining consumers. Federal Rule of Civil Procedure 26(c) authorizes such a protective order only under circumstances "which justice requires to protect a party or person from annoyance, embarrassment, oppression, or undue burden or expense," the potential for which must be illustrated with "a particular and specific demonstration of fact, as distinguished from stereotyped and conclusory statements." Gulf Oil Co. v. Bernard, 452 U.S. 89, 102 n. 16, 101 S. Ct. 2193, 68 L. Ed. 2d 693 (1981). The district court asserted that the order "was necessary in this case to protect the defendants from being unduly burdened and unfairly prejudiced if the plaintiffs were able to . . . interview these customers ex parte."

Yet the district court offered no relevant explanation of how Mitsubishi would be burdened or prejudiced by Nemir's interviewing these consumers, nor is one apparent. Mitsubishi argues that the restrictions protected its business interests by shielding its customers from damaging innuendo about the safety of its products. For support, Mitsubishi relies upon two district court cases: Commodity Futures Trading Commission v. Rosenthal & Co., 74 F.R.D. 454 (N.D.Ill.1977), and Volkswagenwerk v. Westburg, 260 F. Supp. 636 (E.D.Pa.1966). Neither supports Mitsubishi's argument; the plaintiffs in each of those cases had demanded unfettered access to all of the defendant's customers, even those who had never complained about any defects in the defendant's products. In those situations, the courts hesitated to authorize fishing expeditions that might have tainted the defendants' relationships with their customers.

Conversely, the customers that Nemir sought to contact had already complained to Mitsubishi of seatbelt malfunctions. The relevance of these customers was already clear, and interviews by plaintiff's counsel would likely have served as little more than a reminder of the seatbelt problems that the consumers themselves had perceived and articulated. Mitsubishi's concerns, then, are speculative, and mere speculation is insufficient to warrant such stringent limits on pretrial investigation. See, e.g., Gottstein v. National Association for the Self Employed, 186 F.R.D. 654, 659 (D.Kan.1999) ("Defendants also suggest that they have a legitimate business interest in preventing unrestricted contact with their customers . . . . They present nothing but supposition and conjecture, however, to support finding future communications of counsel for plaintiffs will harm their reputation or profits."); Multi-Tech Systems v. Hayes Microcomputer Products, Inc., 800 F. Supp. 825 (D.Minn.1992) ("Granting [plaintiff] access to the customers of [defendant] that it has already discovered during depositions ... will level the parties playing field.... The court declines to adopt [guidelines that the parties could follow in contacting each others customers] because guidelines might place too great a burden on the parties' right to garner evidence in support of the prosecution or defense of the claims in this case."); cf. Williams v. Chartwell Financial Services, Ltd., 204 F.3d 748 (7th Cir.2000) ("The plaintiffs have a right to contact members of the putative class .... [t]he district court's decision as to [a] protective order must involve a careful balancing of the potential for abuse created by the class action and the right of the plaintiffs to contact potential class members.").

Moreover, the burden this protective order imposed on Nemir's investigation was considerable. Rather than being able to conduct informal interviews, Nemir was required to coordinate each and every witness interview with Mitsubishi, as well as bear the cost, in both hours and dollars, of formal depositions. Perhaps even more importantly, Nemir was robbed of the privacy of his attorney's trial preparation. As the Supreme Court declared when refusing to compel the discovery of counsel's notes stemming from his ex parte interview of witnesses, "the general policy against invading the privacy of an attorney's course of preparation is so well recognized and so essential to an orderly working of our system of legal procedure that a burden rests on the one who would invade that privacy." Hickman v. Taylor, 329 U.S. 495, 512, 67 S. Ct. 385, 91 L. Ed. 451 (1947). If Mitsubishi's unsubstantiated fears of prejudice justified a protective order, such orders would be justified in virtually every case—a result that would obliterate the essential principle articulated by the Court.

Nemir further argues that the district court misinterpreted our decision in Nemir I when it limited him to investigating 188 of the consumer complaints, rather than all 29,000. Mitsubishi argued, and the district court agreed, that our decision in Nemir I restricted Nemir to investigating only the 188 complaints that we explicitly mentioned in our decision. But nowhere did our decision limit Nemir to those 188 incidents. Rather, in describing the relevance of the consumer complaints to establishing a dispute of material fact, we noted that "[a] complete review of the record indicates that phraseology analogous to `intermittent partial latching' occurred 188 times in the customer complaints submitted as evidence. To summarily dismiss this evidence without affording the plaintiff time to investigate and further elicit information regarding the exact nature of these incidents is clear error." Nemir, 6 Fed.Appx. at 276.

Mitsubishi fails to mention that the reason we singled out 188 complaints in Nemir I was that the 188 complaints were the only ones in our appellate record, and served as a sample of the larger set of documents. In other words, we cited the 188 documents as examples of the 29,000-plus documents that we deemed clearly relevant to the case and worthy of further investigation. Mitsubishi's argument that "[Nemir] has given this Court no good reason to set aside its original ruling and expand investigation discovery even further" attempts to concoct a holding that is not even remotely discernible from our opinion. See, e.g., First National Bank of Cincinnati v. Pepper, 547 F.2d 708, 714 (2d Cir.1976) (Friendly, J.) ("Such a literal application of the language from this court's previous opinion ... is, we think, not at all what the earlier [panel in this case] had in mind."); United States v. Scheipe, 474 F.2d 720, 721 (3d Cir.1973) ("We are not prepared to give literal efficacy to isolated excerpts from language of that opinion which would otherwise [contradict its holding]"); Continental Motors Corp. v. Continental Aviation Corp., 375 F.2d 857, 861-62 (5th Cir.1967) (reversing decision in which district court "gave too much uncritical and literal significance to [an earlier decision]").

One caveat remains. When the district court initially limited Nemir to discovery of the 188 incidents, it gave Nemir the option of coming back to the district court to request that it be allowed to investigate additional complaints. Nemir never did so, and normally this would mean that he waived his opportunity to interview the other complainants. However, because we are already remanding for new trial, and because our decision has overturned a separate discovery restriction imposed by the district court, there is no reason to deny Nemir the opportunity to investigate further.

Nemir argues that the district court erroneously failed to sanction Mitsubishi for obstructing discovery. Federal Rule of Civil Procedure 37(b)(2) allows a district court to impose sanctions against a party that "fails to obey an order to provide or permit discovery." Nemir argued that his discovery requests included the 29,000 consumer complaints, but that Mitsubishi persistently maintained that it had no documents corresponding to his request. Nemir also asserts that he was prejudiced by the delayed production, arguing that the opportunity to contact many of these potential witnesses evaporated during the delay, although he is noticeably vague about the reasons for this: did most of the potential witnesses move in the interim? But see Computer Task Group, Inc. v. Brotby, 364 F.3d 1112, 1116 (9th Cir.2004) (per curiam) ("[F]ailure to produce documents as ordered is considered sufficient prejudice." (internal quotations and alterations omitted)).

Although Mitsubishi contends that the district court expressly denied Nemir's motion for sanctions, the record clearly reveals that the district court stayed consideration of the motion, never to return to it. The district court's failure to rule on the motion requires that we remand for consideration anew. See Bacou Dalloz USA, Inc. v. Continental Polymers, Inc., 344 F.3d 22, 31 n. 6 (1st Cir.2003) ("Whether to impose sanctions... is a question for the district court in the first instance. We are not equipped to make a fact finding about [the defendant's] alleged bad faith and failure to comply with discovery obligations." (internal citation omitted)); cf. Mutual Service Insurance Co. v. Frit Industries, Inc., 358 F.3d 1312, 1326-27 (11th Cir.2004) ("Without explanation, it is impossible for us to determine whether or not the district court acted within the bounds of its discretion.").

Mitsubishi also argues that our decision in Nemir I held that no sanctionable discovery violation occurred, such that a contrary determination is now precluded by res judicata. In support, Mitsubishi points to the following passage from Nemir I: "[T]he district court perfunctorily disregarded hundreds of pages of consumer seatbelt complaints. This evidence was released to the plaintiff in the eleventh hour, after years of discovery disputes over whether or not such complaints were discoverable material." Nemir I, 6 Fed.Appx. at 275. According to Mitsubishi, our references to "years of discovery disputes" constituted a "holding" that the withholding of the documents was the result of a good-faith disagreement, rather than misconduct.

This argument is baseless. Our decision in Nemir I did not purport to address any of the issues regarding sanctions or discovery misconduct. That we characterized, in passing, the belated disclosure as coming "after years of discovery disputes" does not mean that Mitsubishi's hands were clean: the fact that there were disputes does not automatically mean that both sides in the dispute were playing fairly. (If anything, the tenor of our opinion was critical of Mitsubishi, noting that Mitsubishi had failed to disclose the documents' existence until the "eleventh hour" and suggesting that Mitsubishi withheld otherwise discoverable material by excluding any documents that did not contain a set of legalistic magic words. See id. at 275-76.)

As for whether evidence of any discovery misconduct should have been admitted at trial, we cannot evaluate that claim without first knowing whether any discovery misconduct occurred, and whether and to what extent any misconduct prejudiced Nemir. To give one example, it may be the case that if Nemir can show that Mitsubishi's delays prevented him from contacting a large number of relevant witnesses who could testify to the partial-latching phenomenon in their own seatbelts, then evidence of such delay would be a proper rebuttal to Mitsubishi's arguments that Nemir has been unable to demonstrate that partial-latching actually occurs in real life. There may (or may not) also be other examples of prejudice, or other possible remedies. All of this is for the district court to consider in the first instance in conjunction with Nemir's motion for discovery sanctions.

Nemir highlights an array of allegedly erroneous evidentiary exclusions, contending that the district court: (1) improperly restricted the testimony of Horton; (2) unlawfully prevented Manning from testifying at trial; (3) mistakenly prohibited evidence about Mitsubishi's failure to investigate other similar consumer complaints; and (4) hamstrung Nemir's ability to impeach one of Mitsubishi's expert witnesses.

Nemir argues that the district court erred in prohibiting Horton from testifying about: (1) the cause of Nemir's injuries; and (2) his tests on Nemir's buckle. As to the former, Nemir submits that the district court's barring Horton from testifying about the cause of Nemir's injuries contravened Nemir I. After noting that "Horton tested Nemir's belt buckle and determined that it could be placed into a state of partial engagement by slightly depressing the release button, and by improper or incomplete insertion," id. at 274, we held that

an expert witness' conclusion regarding all admissible evidence need not eliminate all other possible causes of injury in order to be admissible on the issue of causation. Accordingly, Horton could use this evidence to testify that the partial latch of Dr. Nemir's seatbelt caused the damage in question. While other possibilities exist with regards to causation, the fact that several possible causes might remain uneliminated only goes to the accuracy of the conclusion, not the soundness of the methodology.... The district court, by requiring "specific knowledge of the precise physiological cause" of the accident held the expert up to entirely too strict a standard when considering the admissibility of their testimony. The testimony of Horton, establishing causation based upon the above referenced facts, would adequately shoulder the Michigan burden of proof that plaintiff show a reasonable likelihood ... that his explanation of the injury is correct.

Id. at 275 (emphasis added, internal quotations and citations omitted).

At various points in the opinion, moreover, we identified the evidence that would permit Horton to reach his conclusions. We noted, for instance, that

[d]uring his deposition, Horton testified that he had inspected the Nemir seatbelt, and while he did not "measure any loads," he did "check it for partial latching." . . . This document clearly presents evidence which would allow Horton to testify that Nemir's seatbelt could be put into a position of partial latch, and that as such, the seatbelt was defectively designed.

Id. at 271-72 (internal quotations and citations omitted). See also id. at 274 ("Horton's examination of the buckle and the load marks on the jacket could lead a reasonable jury to conclude that the seatbelt remained attached until impact, when it released and retracted, causing damage to the jacket."). On remand, we instructed the district court that Horton's testimony regarding causation was admissible.

Despite our mandate, the district court again ruled that Horton "cannot testify that partial latching [was] the sole cause of Nemir's injuries [because he] does not demonstrate how he arrived at his conclusion ... or how he eliminated other possibilities, chief among which is the possibility that Nemir was not wearing a seat belt." Nemir, 200 F.Supp.2d at 776. Because we plainly held in Nemir I that the causation testimony was admissible, the district court's refusal to admit this testimony contravened our explicit mandate.

Similarly, the district court prohibited Horton from testifying about his testing of the buckle. Horton's testimony—that twice out of twenty attempts, using purposeful manipulation, he was able to cause Nemir's buckle to unlatch—was a critical basis for our decision permitting Horton to present his conclusions about causation. Although we did not explicitly state that "Horton's description of his testing of Nemir's buckle is admissible at trial," we could not have permitted Horton's causation testimony without permitting him to present the bases for his conclusions. See Jones v. Lewis, 957 F.2d 260, 262 (6th Cir.1992) ("[T]he trial court may consider [only] those issues not decided expressly or impliedly by the appellate court or a previous trial court.") (emphasis added).

Even if Nemir I did not directly foreclose the district court's decision to exclude Horton's testing of the buckle, the exclusion was nonetheless clearly erroneous. Federal Rule of Evidence 702 provides that "[i]f scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to demonstrate a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or otherwise." In Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 597, 113 S. Ct. 2786, 125 L. Ed. 2d 469 (1993), the Supreme Court held that before admitting expert scientific testimony, the district court must determine that the testimony "rests on a reliable foundation and is relevant to the task at hand." Daubert applies "not just to scientific evidence but to all expert testimony, including testimony based on technical and other specialized knowledge." Clark v. Chrysler Corp., 310 F.3d 461, 466 (6th Cir.2002), vacated on other grounds, Chrysler Corp. v. Clark, ___ U.S. ___, 124 S. Ct. 102, 157 L. Ed. 2d 12 (2003). Because we review the screening of expert evidence for an abuse of discretion, "the district court has wide latitude to determine the admissibility of expert testimony under Daubert"—but "its discretion is not unbridled." Hardyman v. Norfolk & Western Railway Co., 243 F.3d 255, 267 (6th Cir.2001).

The district court found that Horton's method—in which he manipulated the buckle at varying speeds and angles—was scientifically unsound because "[n]o reasonable driver purposefully manipulates the buckle at different speeds and different angles to achieve a state of partial latch as Horton does." Yet the point of the purposeful manipulation was to show that partial latching could occur under certain circumstances, not to show directly that Nemir's buckle partially latched during the accident—given the infinite possible variations, it would have been virtually impossible to determine the velocity and angle with which Nemir had actually buckled his belt on the day in question. Cf. Clark, 310 F.3d at 473 ("[I]t is not necessary to prove, as Chrysler argues, that the prior accidents involved a vehicle identical to the one driven by [plaintiff] or that all of the circumstances of the accident are identical."). That most of the combinations did not produce partial latching might affect how heavily the jury weighs the evidence, but not whether it should be admitted.

This purpose was narrow, but vital to Nemir's case. Although Horton was permitted to testify more generally as to his conclusion that the buckle was capable of partial latching, Mitsubishi's central refrain to the jury when disputing the buckle's propensity to partially latch was that Horton was just spouting words that were uncorroborated by any actual test or demonstration. Thus, before Nemir could convince the jury that partial latching occurred during the accident, he had to show the jury that partial latching was something that could have happened during the accident: in other words, that his buckle was capable of partially latching. Merely being allowed to announce his conclusion, sans support, was unlikely very helpful to Horton's testimony or Nemir's case. See Rosen v. Ciba-Geigy Corp., 78 F.3d 316, 319 (7th Cir.1996) ("[A]n expert who supplies nothing but a bottom line supplies nothing of value to the judicial process. . . ."). And because an individual driver will inevitably buckle his seatbelt at varying angles and with varying speeds, purposeful manipulation would have been the only way to demonstrate partial latching in a laboratory setting. See Hardyman, 243 F.3d at 262 (finding abuse of discretion when district court demanded a method of evaluation that was scientifically impossible); Guild v. General Motors Corp., 53 F. Supp. 2d 363, 366 (W.D.N.Y.1999) ("[P]erfect identity between experimental and actual conditions is neither attainable nor required."). Indeed, both of the experts that were appointed by the district court used the same technique as did Horton to test the buckle, a technique that was appropriate for the conclusion that Horton was asserting.

The exclusion of these tests was exploited by Mitsubishi. During closing argument, Mitsubishi argued that "they haven't shown partial latching to you except in a drawing. They haven't shown partial latching of an old buckle, a new buckle, a used buckle, Dr. Nemir's buckle, or any TK 52 buckle." Mitsubishi went on to argue that "[i]t's never been demonstrated. It's never been demonstrated. All you have is Horton's words of this. He has no basis to conclude the TK 52 buckle—can partial latch in a manner that's representative of actual use beyond his sheer speculation." Reinforcing the harm to Nemir's case caused by the prohibition, Mitsubishi argued that "[t]his is serious case. You can't prove your case with just the words of a hired expert ... Another federal court, we brought out, rejected Mr. Horton's testimony for something called, by the court, parlor tricks. I submit that surely that is why he didn't attempt to demonstrate anything here." Of course, Horton had attempted to demonstrate something, and the district court's exclusion of this demonstration was error.

Nemir also sought to present testimony from court appointed expert Lindley Manning, who, like Horton, was able to cause Nemir's buckle to partially latch. Federal Rule of Evidence 706 provides that a court appointed witness "shall advise the parties of the witness' findings, if any; the witness' deposition may be taken by any party; and the witness may be called to testify by the court or any party." (emphasis added) Yet the district court, without explanation, prohibited Nemir from introducing Manning's testimony. Mitsubishi nevertheless argues that the "plain language" of Rule 706 supports the district court's decision. Yet the plain language of the rule could hardly be more clear in its support of Nemir's position: either the court or any party "may" call the witness to testify. The rule simply does not condition a party's doing so on the district court's approval, nor does it maintain merely that the district court "may allow either party to call the witness."

Manning's testimony would have been important to Nemir, for several reasons. First, it would have corroborated Horton's conclusions that the Nemir buckle was capable of partially latching and had been demonstrated to partially latch. Given that Manning was court-appointed, rather than hired by Nemir, a jury would likely have considered his testimony to be more credible than that of Horton, whom Mitsubishi derided as a "hired gun." Second, Manning's testimony that even infrequent partial latching may constitute a serious defect would have bolstered the weight afforded to Horton's research, which found partial engagement in two out of twenty attempts. The erroneous exclusion of this evidence again undercut Nemir's case.

3. Mitsubishi's Failure to Investigate Consumer Complaints

Nemir argues that the district court erred in prohibiting him from introducing evidence about Mitsubishi's alleged failure to properly investigate seatbelt-related complaints it had received from other consumers. Evidence of Mitsubishi's failure to investigate would probably be inadmissible to prove directly the proposition that the buckles are capable of partially latching. And although such evidence would probably be admissible to support a case for punitive damages, see Clark, 310 F.3d at 481 (upholding award of punitive damages because although aware of possible safety problems with pickup truck, Chrysler failed to perform relevant tests), Nemir's punitive damages claim has been dismissed.

But at the first trial, Mitsubishi opened the door: Mitsubishi argued that Nemir was unable to unearth any examples of partial-latching occurring in real life. In response, it would have been important for the jury to know that one of the reasons that he had been unable to deliver such evidence was that Mitsubishi had failed to perform appropriate investigations and/or testing that could have yielded such evidence (with Cooper's testimony confirming that examination of the buckles would have been appropriate). Cf. Beck v. Haik, 377 F.3d 624 (6th Cir.2004) ("A party may also be entitled to special instructions if he can raise an issue of fact as to whether a party has failed to preserve relevant evidence."); cf. also General Motors Corp. v. Blake, 237 Ga.App. 426, 515 S.E.2d 166, 169 (1999) ("Fully aware that the issue was why did the seatbelt not restrain plaintiff, if it did not do so, GM apparently tactically decided not to conduct a fuller and more complete investigation of the apparatus. Defendant should not be permitted to avoid learning the facts by lack of thoroughness and then claim surprise."); Daniel v. Indiana Mills & Manufacturing, Inc., 103 S.W.3d 302, 315 (Mo.App.2003) ("Plaintiff was permitted, in rebuttal, to use [defendant's crash test videos] to demonstrate that initial unlatching could occur."); Wiitala v. Ford Motor Co., 2001 WL 1179610 at *7 (Mich.App. Oct.5, 2001) ("We are also not persuaded by defendant's argument that the test documents uncovered after the violation were irrelevant because all but one involved buckles other than that found in plaintiff's car .... given that defendant took the position at trial that inertial unlatching is simply a `parlor trick,' we believe the evidence was relevant in that it directly countered this assertion. . . .").

If, during the new trial, Mitsubishi argues that the absence of real-life incidents suggests that partial-latching is a laboratory-only phenomenon, Nemir must be permitted to introduce evidence that the absence of real-life incidents might also signal only that Mitsubishi failed to investigate such incidents.

Finally, Nemir seeks reversal based on the district court's interference with his impeachment of Dr. Hatsell. As we explained above, when Nemir attempted to impeach Hatsell on cross-examination by confronting him with an excerpt from a book that had been co-authored by Spitz, another expert retained by Mitsubishi, the district court instructed Nemir to save such questioning for Spitz's testimony. When Nemir questioned Spitz about the book, the district court ignored its prior instructions and prohibited such questioning as beyond the scope of his testimony. In the end, Nemir was altogether prohibited from introducing this concededly relevant impeachment evidence.

Mitsubishi's defense of the district court's decision is spurious. First, Mitsubishi argues that the district court correctly refused to allow the impeachment of Dr. Hatsell with the book authored by Dr. Spitz because "there is no indication from the record that plaintiff was doing anything other than attempting to elicit an opinion from Dr. Hatsell as to the possible opinions of Dr. Spitz regarding brain injuries caused by rotational forces applied to the brain." This is technically correct but highly misleading: Dr. Spitz's published opinions about brain injuries were offered to impeach the contrary conclusions of Hatsell, not to have Hatsell comment on Spitz's upcoming testimony. Nemir sought to impeach Dr. Hatsell with the writings of Dr. Spitz; his appearance on the witness list was beside the point.

Second, Mitsubishi argues that even though the district court prohibited Nemir from impeaching Hatsell with the writings of Spitz on the grounds that Nemir would have a chance to elicit the impeaching information directly from Spitz, "the district court reasonably and properly sustained an objection to this collateral inquiry because it was not a subject of Dr. Spitz's testimony in this case." Mitsubishi again flagrantly misrepresents the record. Whether the questions targeted Spitz's testimony was irrelevant: the questions were designed not to impeach Spitz, but rather to impeach Hatsell by evincing a contrary opinion. The only reason that Nemir was impeaching Hatsell with Spitz's testimony—rather than confronting Hatsell directly with the treatise written by Spitz—was that the district court ordered Nemir to proceed this way.

Thus, Nemir was erroneously prevented from introducing relevant impeachment evidence. And this impeachment evidence might have been particularly valuable to Nemir, given that it came from Mitsubishi's own expert.

Mitsubishi argues that any errors made at trial were harmless, given that the errors influenced only the jury's determination whether the seatbelt functioned properly, whereas the jury found that Nemir had not been wearing his seatbelt at all. Nemir argues, in contrast, that the jury's determination that he was not wearing his seatbelt was infected by the errors regarding the effectiveness of the seatbelt. In other words, posits Nemir, if the jury did not believe that the seatbelt was ever capable of partially latching, it would necessarily have attributed Nemir's injuries to his not having worn his seatbelt at all. When reviewing whether a claim that an error was harmless, we must reverse "[i]f we do not have a fair assurance that the trial's outcome was not altered by error." Beck, 377 F.3d at 635.

In arguing that Nemir was not wearing his seatbelt on the day of the accident, Mitsubishi relied primarily on evidence unrelated to the buckle. During closing argument, Mitsubishi argued that Nemir's memory of the accident was suspect, thereby casting doubt on his claim that he was indeed wearing the seatbelt. Moreover, Mitsubishi pointed to an array of physical evidence, independent of the buckle, which it argued showed that Nemir was not wearing his seatbelt, and also argued that the evidence of seatbelt-wearing introduced by Nemir did not show otherwise. It is clear, then, that the jury reasonably could have found that Nemir was not wearing his seatbelt at the time of the accident, and could have done so even if it believed that Nemir's buckle had the potential to partially unlatch.

Yet the jury was not required to consider only arguments that the parties raised during summation. It is a familiar part of any judge's instructions to a jury that it is bound by the evidence, not by counsel's closing arguments. See Federal Jury Practice and Instructions, Civil, 3 Fed. Jury Prac. & Instr. § 101.02 (5th ed.) ("After the evidence is presented, the parties' lawyers make closing arguments explaining what they believe the evidence has shown. What is said in the closing arguments is not evidence."). Although Nemir's argument would certainly be stronger if Mitsubishi had more frequently argued that the lack of seatbelt deficiency was itself proof that Nemir was not wearing his seatbelt, that Mitsubishi's counsel did not stress this argument does not mean that the jury did not consider it.

In any event, during the opening statement, Mitsubishi told the jury that "at the end of the case, you must decide [whether Nemir] engaged in the very difficult machinations that the evidence will show are necessary to cause partial latching or partial engagement, or that he simply didn't put his belt on." Given that an opening statement is "an outline to help [the jury] understand what each party expects the evidence to show," id., the jury could very well have viewed the evidence at trial in light of the either-or framework presented to it by Mitsubishi's opening statement. And if the jury did view the evidence this way, deficiencies in evidence about the buckle's malfunction could have led the jury to conclude that the seatbelt simply could not have malfunctioned, compelling the conclusion that Nemir was not wearing it.

Although we are not to presume that a jury disregarded the district judge's instructions when answering the questions in a special verdict form, see Chute v. Roebuck and Co., 143 F.3d 629, 632 (1st Cir.1998), we need presume no juror malfeasance to conclude that the jury's views about the buckle's reliability may have influenced its determination that Nemir was not wearing his seatbelt. As a matter of logic, the jury could have reasoned that: (1) The only way Nemir could have received his injuries if he was wearing his seatbelt was if the seatbelt was defective; (2) the seatbelt was not defective; therefore (3) Nemir was not wearing the seatbelt. As the Seventh Circuit put it, when reviewing a jury's verdict in a seatbelt case with similar facts, "[e]ngineers trying to understand a disaster often follow causal chains ("failure trees") until they find one that can account for the calamity. Sherlock Holmes observed that `when you have eliminated the impossible, whatever remains, however improbable, must be the truth.' Courts need not disdain a method that both engineers and detectives find useful." Bammerlin v. Navistar Int'l Transportation Corp., 30 F.3d 898, 902 (7th Cir.1994) (Easterbrook, J.) (internal citation omitted).

Our concern that the errors were not harmless is heightened by the case of Babcock v. General Motors Corp., 299 F.3d 60, 62 (1st Cir.2002), a case in which the plaintiff alleged that his seatbelt—due to the same partial-latching phenomenon alleged by Nemir—unlatched during an accident. Like our case, the plaintiff in Babcock had to convince the jury both that the driver had been wearing his seatbelt and that the seatbelt was defective. As to the latter, the plaintiff was bolstered by the testimony of an expert who had tested the driver's buckle. See id. at 68. Although the evidence that the plaintiff was wearing a seatbelt was arguably weaker than it was in our case—the plaintiff's case was based only on testimony about the driver's habit of always wearing his seatbelt, whereas Nemir also had some supportive physical evidence—the jury found both that the driver was wearing his seatbelt and that the seatbelt was defective. See id. at 66. Of course, for a host of reasons, this does not compel the conclusion that in our case, the jury's determination whether Nemir wore his seatbelt was affected by the lack of evidence regarding the efficacy of the buckle. But Babcock does reinforce our conclusion that, as a matter of logic, erroneous evidentiary rulings pertaining to one question had the potential to affect the jury's resolution of the other.

Nor does Grimes v. Mazda North American Operations, 355 F.3d 566 (6th Cir.2004), suggest otherwise. In Grimes, a case based on similar facts, we held that any error that might have arisen from the exclusion of evidence about prior accidents involving defective seat belt latches was harmless, because "[t]he jury returned a special verdict that plaintiff was not wearing her seatbelt and that this, not design defect, may have been a cause of her injury." Id. at 573. Yet in Grimes, the plaintiff never argued to us that one set of errors affected the jury's analysis of another question, such that the question before us in today's case was simply not before us in Grimes. "Questions which merely lurk in the record, neither brought to the attention of the court nor ruled upon, are not to be considered as having been so decided as to constitute precedents." Webster v. Fall, 266 U.S. 507, 511, 45 S. Ct. 148, 69 L. Ed. 411 (1925). See also United States v. Rivera-Guerrero, 377 F.3d 1064 (9th Cir. Jul. 20, 2004) (relying on quoted language). In any event, the evidence pertaining to seatbelt wearing was far more one-sided in the defendant's favor in Grimes, given that the plaintiff had told an emergency technician and two nurses that she had not been wearing her seatbelt. Thus, even if we had expressly considered such an argument in Grimes, the facts in the present case much more strongly suggest that the errors affected the trial's outcome.

Ultimately, because we simply do not know whether the errors pertaining to the functioning of the seatbelt affected the jury's determination of whether Nemir was wearing the seatbelt, we cannot say with confidence that the errors at trial were harmless.

Mitsubishi challenges the costs awarded by the district court to Mitsubishi. Under Federal Rule of Civil Procedure 54(d)(1), "costs other than attorneys' fees shall be allowed as of course to the prevailing party." Because our reversal of the judgment for Mitsubishi strips it of its status as the "prevailing party," the district court's award of costs is set aside.

The judgment of the district court is REVERSED, the award of costs VACATED, and the case REMANDED for a new trial. Finally, the case is to be assigned to a different district judge for all further proceedings in this case. To minimize confusion about the required course upon remand, we summarize the specifics of our decision: (1) Nemir should have been permitted a reasonable opportunity to contact, free of the previous restrictions, the other consumers who filed complaints with Mitsubishi; (2) the district court should have ruled on Nemir's motion for sanctions based on Mitsubishi's alleged violations of discovery rules; (3) Nemir should have been permitted to present his strict liability theory to the jury; (4) Horton should have been permitted to testify both as to causation and to his testing of Nemir's buckle; (5) Nemir should have been permitted to call Drs. Cooper and/or Manning to testify at trial; (6) Nemir should have been permitted to impeach the testimony of Dr. Hatsell with the writing and/or testimony of Dr. Spitz; and (7) Nemir should have been permitted to offer, in response to Mitsubishi's arguments that partial-latching has never happened in the real world, evidence that it declined to perform tests when alerted by consumers of the potential for its buckles to unlatch.