In this products liability action, plaintiff appeals from a grant of judgment as a matter of law in favor of the defendant pursuant to Federal Rule of Civil Procedure 50(a). The primary issue raised is whether the district court erred by striking the testimony of plaintiff's expert witness in its entirety on the ground that the expert altered the allegedly defective product during the course of his examination. Because we conclude that the district court's decision constituted reversible error, we will reverse and remand for further proceedings.

On August 31, 1990, a co-worker of plaintiff Eric Schmid was using a circular saw manufactured by defendant Milwaukee Electric Tool Corporation ("Electric Tool") to cut wooden forms for concrete steps. Schmid was holding the board as his co-worker cut. Allegedly, with no warning, the saw "kicked back" and severely cut Schmid's hand, causing permanent injury. Schmid theorizes that the saw suffered from a design defect: the guard was prevented from properly snapping back over the blade by debris collected in the guard mechanism.

Following the accident, Schmid's counsel sent the saw in question to an expert, Dr. Jeffrey Bratspies, for examination. According to his videotaped deposition, Dr. Bratspies noted that the guard was closing sluggishly and that there was a grating noise. He disassembled the guard to find out why. This disassembly, according to Bratspies, revealed particles trapped in the guard mechanism, some of which fell out during the course of the examination. Some of the trapped particles had caused scoring of the metal on the mating areas of the blade guard. Dr. Bratspies took photos of the assembled saw, the disassembled saw, the scoring, and the debris and the disassembled saw was then forwarded to Electric Tool's expert. Upon reassembly by that expert, the saw apparently worked properly; the guard snapped back as it was designed to do.

During cross examination of Dr. Bratspies, the following colloquy occurred:

Q. Now, as I understand it, you--you took the blade off and you--you took the guard apart. Am I correct?

A. I removed the guard, yes.

Q. And there were some particles that fell out and that sort of thing?

A. Yes.

Q. Were you--were you aware that if any other expert was to look at the saw that they would be unable to examine the saw as you examined it because of the way you disassembled the saw?

A. Yes. That was a--a judgment on my part. I felt that it was important to determine what was causing the--the problem. And after I disassembled the saw I left it in a disassembled condition. I did not reassemble it. And it was then taken back by other people, presumably, to investigate in--in--in exactly the condition that I had disassembled it.

Q. Okay. You did--you did realize that by disassembling it you were, in effect, altering the condition of it and its performance as compared with the condition in which you had received it?

A. Yes. Except that it was a transient condition at best because even as I was handling it there were particles that were falling out of it and its performance--its behavior was changing.

App. 209-10.

At trial, Dr. Bratspies's videotaped deposition, as well as the photos that he took, were initially introduced into evidence. Dr. Bratspies testified that the guard design allowed particles to interfere with the proper closing of the guard and that a seal over the guard would have prevented this problem. He opined that the saw was therefore defective and that absent this defect, Schmid would not have been injured.

Electric Tool moved to strike Dr. Bratspies's testimony, the photographs and the saw. It argued that, by disassembling the guard and not preserving the particles which fell out of the guard mechanism, Dr. Bratspies destroyed evidence, thereby prejudicing the defense. Electric Tool also complained that Dr. Bratspies neither timed how long it took the allegedly sluggish guard to close nor videotaped it closing before he disassembled it.

The district court found merit in these arguments and granted the motion:

[T]he defendant was denied any opportunity for a fair chance to examine the saw as to its condition at the time of the accident.... [T]he expert testified as to the condition of the saw at the time of the accident and the defendant was never in a position where it could respond to that testimony.... [I]t would be grossly unfair ... to allow one side to present that kind of testimony and not have the other side be in a similar position to do so.

App. at 153-55.

Because this evidentiary decision deprived plaintiff of sufficient evidence to support his theory of the case, the district court granted judgment as a matter of law in favor of the defendant. Schmid filed this timely appeal.1

There is some disagreement in this case as to whether the district court's decision to strike the saw and Dr. Bratspies's testimony from evidence was one made as a matter of substantive Pennsylvania products liability law or federal evidentiary law. There is some authority suggesting that spoliation of evidence, and the sanctions that such spoliation may give rise to, are matters appropriately governed by state law. See, e.g., Mensch v. Bic Corp., 1992 WL 236965 (E.D.Pa.1992) (applying Pennsylvania law); Lee v. Boyle-Midway Household Products, Inc., 792 F. Supp. 1001, 1005 (W.D.Pa.1992) (applying Pennsylvania law). On the other hand, there is also authority suggesting that the possible preclusion of evidence in cases such as this is governed by federal law as part of the inherent power of a district court to sanction parties. See, e.g., Chambers v. NASCO, Inc., --- U.S. ----, ---- - ----, 111 S. Ct. 2123, 2132-33, 115 L. Ed. 2d 27 (1991) (recognizing inherent power of a district court to sanction parties in appropriate cases); Dillon v. Nissan Motor Co., Ltd., 986 F.2d 263, 267 (8th Cir.1993) (recognizing the inherent power of a district court to sanction parties for destruction of evidence by, inter alia, excluding testimony); Unigard Sec. Ins. Co. v. Lakewood Engineering & Mfg. Corp., 982 F.2d 363, 368 (9th Cir.1992) (same). The parties have referred us to no case law, however, suggesting that the applicable federal and state law are materially different. Because we conclude that the district court's decision exceeded the appropriate bounds of its discretion under either federal or state law, we need not resolve the choice of law question.

Since the early 17th century, courts have admitted evidence tending to show that a party destroyed evidence relevant to the dispute being litigated. Jamie S. Gorelick, Steven Marzen and Lawrence Solum, Destruction of Evidence, Sec. 2.1 (1989). Such evidence permitted an inference, the "spoliation inference," that the destroyed evidence would have been unfavorable to the position of the offending party. As Judge Breyer put it in Nation-wide Check Corp. v. Forest Hills Distributors, Inc., 692 F.2d 214, 218 (1st Cir.1982), "the evidentiary rationale [for the spoliation inference] is nothing more than the common sense observation that a party who has notice that [evidence] is relevant to litigation and who proceeds to destroy [evidence] is more likely to have been threatened by [that evidence] than is a party in the same position who does not destroy the document." As Judge Breyer also noted, the spoliation inference is also seen as having "prophylactic and punitive effects." Id. The admissibility of spoliation evidence and the propriety of the spoliation inference is well established in most jurisdictions, including Pennsylvania. See e.g., Nation-wide Check Corp., 692 F.2d 214 (1st Cir.1982); Mensch v. Bic Corp., 1992 WL 236965 (E.D.Pa.1992) (citing Pennsylvania cases); Gorelick, et al., supra, Sec. 2.24.

The district court here invoked a sanction far more serious than the spoliation inference. It barred all evidence emanating from Dr. Bratspies's observations and thereby deprived Schmid of any opportunity to prove his case. While we do not doubt the inherent authority of a district court to impose such a drastic sanction in an appropriate case, we conclude that this was not such a case. We believe the key considerations in determining whether such a sanction is appropriate should be: (1) the degree of fault of the party who altered or destroyed the evidence; (2) the degree of prejudice suffered by the opposing party; and (3) whether there is a lesser sanction that will avoid substantial unfairness to the opposing party and, where the offending party is seriously at fault, will serve to deter such conduct by others in the future. See e.g., Dillon v. Nissan Motor Co., Ltd., 986 F.2d 263 (8th Cir.1993); S.D.I. Operating Partnership, L.B. v. Neuwirth, 973 F.2d 652 (8th Cir.1992); Gorelick, et al., supra, Sec. 3.16, p. 117 ("To fulfill the purposes of discovery sanctions ...--that is, to restore the accuracy of the trial, compensate innocent victims, and punish guilty spoliators--courts select the least onerous sanction corresponding to the willfulness of the destructive act and the prejudice suffered by the victim.") An analysis in this case sensitive to these considerations does not support the judgment of the district court.

Dr. Bratspies did not destroy the allegedly defective product. This fact alone distinguishes most of the cases in which plaintiffs have been deprived of the use of evidence necessary to prove their claims.2 All Dr. Bratspies did was disassemble the saw in an attempt to determine why the guard was closing slowly. Disassembly was necessary to determine whether Schmid had a meritorious claim against the manufacturer of the saw. Moreover, when Dr. Bratspies decided to go forward with his investigation, he was confronted with a "transient" situation in which any handling of the saw altered its condition to some degree. This meant that some alteration over time was inevitable and made it important that an investigation be conducted promptly. Finally, we note that Dr. Bratspies did take photographs of the saw prior to disassembly, the parts of the disassembled saw, and the debris. These photographs corroborate Dr. Bratspies's observation regarding the presence of debris within the guard housing. They also allow a defense expert some, though admittedly limited, visual access to the saw prior to and immediately following disassembly.

Turning from Dr. Bratspies's conduct to the extent of the prejudice to the defendant, it is important to note at the outset that this is a design defect case. This is not a case in which the plaintiff contends that the particular saw causing his injuries was defectively manufactured; there is no claim that the saw had a defect not shared by all saws of the same model. Rather, Schmid's claim is that the design of saws of this model permitted particles to infiltrate the guard mechanism, thereby impairing its effectiveness and that those saws could have been designed with a seal that would have prevented the infiltration.

Because this was Schmid's theory, Electric Tool's need for immediate access to the particular saw involved in the accident was greatly diminished. The ability to determine whether the design of its saw permits sufficient infiltration to impair the guard's effectiveness and whether a safer design is feasible are matters that can be determined as well or better by inspecting and testing multiple saws of the same design than by inspecting the particular saw involved in the accident. In Quaile v. Carol Cable Co., Inc., 1993 WL 53563 (E.D.Pa.1993), for example, the court declined to grant summary judgment for the defendant even though the plaintiff had disposed of an allegedly defective lamp. Because plaintiff's theory of the case postulated that all defendant's lamps were defectively designed, the court reasoned that the defendant could examine its other lamps and, accordingly, was not prejudiced by the missing lamp. See also Lee v. Boyle-Midway Household Products, Inc., 792 F. Supp. 1001, 1005 (W.D.Pa.1992) ("Under Pennsylvania law, in a case in which plaintiff does not allege a defect in all of the defendant's products, a defendant in a products liability case is entitled to summary judgment when loss or destruction of evidence deprives the defense of the most direct means of countering plaintiff's allegations.") (emphasis added).

We do not suggest, of course, that a defendant in the position of Electric Tool has no legitimate interest in inspecting an injury-causing product in a condition as close as possible to its condition at the time of the accident. The plaintiff in a design defect case, in addition to proving a design defect, must prove that that defect caused his injury and the defendant will want as much information as possible relevant to the issue of causation. We do suggest, however, that the potential for prejudice is much less in a design defect case. We also note that, in this particular case, it appears unlikely that an inspection of the saw by a defense expert contemporaneous with Dr. Bratspies's inspection would have produced evidence of importance to the defense on the causation issue. There appears to be no dispute that Schmid's hand was cut by the blade of the saw, and Schmid testified that he saw the saw "kick back" and the guard fail to close at the time of the accident. While Electric Tool speculates that an earlier examination of the saw might have provided some evidence helpful to it on the causation issue, it has not come forward with plausible, concrete suggestions as to what that evidence might have been.3

The district court made no finding that Dr. Bratspies intended to impair the ability of the potential defendant to defend itself and we do not think the record before the district court would support such a finding. The most that can be said based on the evidence before the district court is that Dr. Bratspies failed to time or videotape the closing of the guard and did not attempt to arrange for the potential defendant to have its own expert present during his investigation.

While Dr. Bratspies's investigation and the documentation of his findings would have been more complete had he timed and videotaped the closing of the guard, we think this is an insufficient degree of fault to justify a dismissal in a case where, as here, the defendant has not shown substantial prejudice that would have been avoided by those additional steps. Given the "transient condition" of the saw, even if these steps had been taken, the result would not have been an indisputable account of what took place at the time of the accident. It was impossible for Dr. Bratspies or anyone else to duplicate the events that resulted in Schmid's injuries and the unavailability of the information that a precise duplication would have afforded was a fact of life with which both sides had to live.

In the final analysis, it seems to us that we could affirm the district court only if we were willing to hold that an expert in Dr. Bratspies's position has an affirmative duty not to conduct an investigation without affording all potential defendants an opportunity to have an expert present. We decline to so hold.

When Dr. Bratspies conducted his investigation, no suit had been filed and Schmid did not know whether he had a basis for instituting suit. Thus, no defendant had been identified and, a fortiori, no defense expert had been engaged. An across-the-board rule that would require an identification of all potential defendants at this nascent stage of the potential controversy and an invitation to each of them to attend an exploratory investigation would be inefficient, if not altogether unworkable. Many accident investigations do not lead to litigation and many narrow the field of potential defendants. There are still more situations in which the resources necessary to assemble the experts of potential defendants would be invested with no significant return on the investment. On the other hand, in cases like this one, the attendant delay would pose a significant risk of prejudice to the plaintiff.

Rather than endorse Electric Tool's novel rule, we consider it more prudent to rely on the traditional case by case approach keyed to the degree of fault on the part of the party accused of spoilation and the degree of prejudice to the opponent. In this case, there was very little fault that could be laid at Dr. Bratspies's feet, very little nonspeculative prejudice to defendant, and a sanction that deprived Schmid of any opportunity to prove his claim. That sanction was not commensurate with the limited fault and prejudice present in this case. As a result, we can sanction neither the district court's striking of Schmid's expert evidence nor its granting judgment to Electric Tool.

The judgment of the district court will be reversed, and this case will be remanded for a new trial.

Before: SLOVITER, Chief Judge, BECKER, STAPLETON, MANSMANN, GREENBERG, HUTCHINSON, SCIRICA, COWEN, NYGAARD, ALITO, ROTH, and LEWIS, Circuit Judges.

The petition for rehearing filed by appellee in the above-entitled case having been submitted to the judges who participated in the decision of this Court and to all the other available circuit judges of the circuit in regular active service, and no judge who concurred in the decision having asked for rehearing, and a majority of the circuit judges of the circuit in regular active service not having voted for rehearing by the court in banc, the petition for rehearing is denied.