Mr. Gregory Gilbert and Mrs. Tamara Gilbert, the Appellants, brought a products liability action against Cosco Inc. claiming that defects in the child restraint device manufactured by Cosco Inc. caused the death of their infant child, Deric Gilbert. Mr. and Mrs. Gilbert appeal the jury verdict alleging the district court erred in certain evidentiary decisions and erred in instructing the jury. We disagree with Mr. and Mrs. Gilbert and affirm the district court.

The facts involved in this dispute are as follows. Mrs. Tamara Gilbert was driving home in her Bronco II with her three-month-old infant, Deric Gilbert. Deric was riding in a Model 582 child restraint designed and manufactured by Cosco Inc. During the drive home, Mrs. Gilbert was distracted by her child and accidently drove the vehicle off the highway at a speed of fifty to fifty-five miles per hour. The vehicle, which remained upright at all times, bounced down a steep embankment roughly 200 to 215 feet before coming to rest. As a result of the accident, Deric suffered a severe head injury from which he died nearly one week later.

Mr. Gregory Gilbert, both as father and as administrator of Deric's estate, and Mrs. Tamara Gilbert (hereinafter referred to as the Gilberts) brought suit against Cosco Inc. (Cosco) alleging the Model 582 child restraint was defective and that the defect was the direct and proximate cause of the fatal injuries sustained by Deric in the accident. At trial, the Gilberts alleged two defects in the Model 582 child restraint: (1) the type of foam padding, and (2) the design, which caused a "springboard effect." Cosco denied that its Model 582 child restraint was defective and put forth alternative explanations for the injuries sustained. The case was tried before the United States District Court for the Eastern District of Oklahoma based on diversity jurisdiction. The jury returned a verdict in favor of Cosco and the Gilberts appeal.

The substance of this appeal concerns certain evidentiary rulings made by the district court. At trial, over objection by the Gilberts, the district court allowed Cosco to introduce as evidence film depicting sled tests it conducted with the Model 582 child restraint. The court also granted Cosco's Motion in Limine to preclude the admission of evidence relating to other child restraint models manufactured by Cosco which were not a subject of this lawsuit.

The Gilberts allege that the trial court: (1) erred in admitting the sled tests Cosco conducted with the Model 582 child restraint; (2) erred in excluding evidence which demonstrated decisions made by Cosco in manufacturing other models of child restraints; and (3) erred in failing to instruct the jury, as requested, that upon proof that Cosco withheld evidence, such evidence is presumed unfavorable to Cosco.

We first address the Gilberts' contention that the trial court erred in admitting the sled tests into evidence. The Gilberts argue that the sled tests were not conducted under conditions substantially similar to the accident and thus were irrelevant or prejudicial.1 Cosco contends that the sled tests were not meant to simulate the accident, nor were they admitted into evidence as such, instead, the tests were designed to illustrate scientific principles. After careful consideration of the record, we affirm the district court.

The district court is given broad discretion in deciding evidentiary matters including the admissibility of experiments.2 Four Corners Helicopters, Inc. v. Turbomeca, S.A., 979 F.2d 1434, 1441-42 (10th Cir.1992); McEwen v. City of Norman, 926 F.2d 1539, 1553-54 (10th Cir.1991). We review such evidentiary rulings for an abuse of discretion. Durtsche v. American Colloid Co., 958 F.2d 1007, 1011 (10th Cir.1992). "Under the abuse of discretion standard, a trial court's decision will not be disturbed unless the appellate court has a definite and firm conviction that the lower court made a clear error of judgment or exceeded the bounds of permissible choice in the circumstances." McEwen, 926 F.2d at 1553-54 (quoting United States v. Ortiz, 804 F.2d 1161, 1164 n. 2 (10th Cir.1986)).

As a general rule, the proponent seeking to admit out-of-court experiments into evidence must demonstrate a " 'similarity of circumstances and conditions' " between the tests and the subject of litigation. Jackson v. Fletcher, 647 F.2d 1020, 1027 (10th Cir.1981) (quoting Navajo Freight Lines v. Mahaffy, 174 F.2d 305, 310 (10th Cir.1949)). See also Robinson v. Audi NSU Auto Union Aktiengesellschaft, 739 F.2d 1481, 1484 (10th Cir.1984); Brandt v. French, 638 F.2d 209, 212 (10th Cir.1981). "The purpose of this rule is to prevent confusion of the jury." Robinson, 739 F.2d at 1484 (citing Jackson, 647 F.2d at 1027). A recognized exception to this rule exists when the experiment merely illustrates principles used to form an expert opinion. In such instances, strict adherence to the facts is not required. Four Corners Helicopters, 979 F.2d at 1442; Robinson, 739 F.2d at 1484; Brandt, 638 F.2d at 212. See also Nachtsheim v. Beech Aircraft Corp., 847 F.2d 1261 (7th Cir.1988).3 Therefore, experiments which purport to recreate an accident must be conducted under conditions similar to that accident, while experiments which demonstrate general principles used in forming an expert's opinion are not required to adhere strictly to the conditions of the accident. Furthermore, "when experiments do not simulate the actual events at issue, the jury should be instructed that the evidence is admitted for a limited purpose or purposes." Robinson, 739 F.2d at 1484 (citing Brandt, 638 F.2d at 212); Bannister, 812 F.2d at 1270.

Cosco argues that since the "tests were offered to demonstrate scientific principles underlying Dr. McElhaney's testimony and opinions," the tests need not simulate the conditions of the accident.4 Cosco further contends that the tests were conducted specifically to refute the Gilberts' expert witness who testified that Ensolite padding was superior to the urethane foam used by Cosco, and that the design of the seat caused a "springboard effect."

It is clear from the record that the trial court's decision to admit the sled tests into evidence was for the limited purpose of illustrating certain scientific principles and not to recreate the accident. The court heard testimony by Dr. McElhaney discussing how the tests were conducted and the purposes for which the tests would be used if offered as evidence. Specifically, Cosco proposed to use the tests to demonstrate the physical principles upon which Dr. McElhaney would base the following opinions: (1) in a relative comparison of energy absorption properties, Ensolite performed no better than the urethane foam used in the Model 582 child restraint; (2) contact between the child's head and the padded restraint device could not generate sufficient force to cause a significant head injury; and (3) the alleged rebound effect identified by the Gilberts' expert witness was due to applying breaks to the sled after impact, and not from any design flaw in the child restraint.

The trial court did not make a hurried decision. Before admitting the sled tests into evidence, the court considered arguments from both sides, heard testimony from Dr. McElhaney concerning the tests and the relevancy of the data, allowed the Gilberts' attorney to question Dr. McElhaney, and in some instances, actively questioned the expert witness to discern the potential relevancy and prejudicial impact of the tests. The sled tests were ultimately used to assist Dr. McElhaney in demonstrating physical principles which formed the basis of his expert opinion and thus it was unnecessary for the tests to replicate conditions of the accident. The tests were clearly relevant, and while somewhat prejudicial to the Gilberts, we do not find that the trial court abused its discretion by allowing the tests into evidence. Moreover, the Gilberts were given ample opportunity to attack the credibility of Dr. McElhaney's conclusions by pointing out inconsistencies and shortcomings in the sled test design and such factors were considered by the jury in weighing the evidence.

The Gilberts contend that despite Cosco's assurances that the tests were not meant to recreate the accident, Dr. McElhaney actually utilized the tests at trial to show how the accident happened. Essentially, the Gilberts argue that even if the sled tests were admissible for the limited purposes proposed by Cosco, the tests were actually used at trial to recreate the accident. We disagree.

The Gilberts refer to two instances in particular where they contend the tests were used to demonstrate more than basic principles: (1) Dr. McElhaney testified that the forces applied to the mannequin's head during the sled test could not cause a skull fracture, and since the sled test was "more severe" than the accident, there was no potential for injury in the accident when the child's head struck the restraint device; and (2) Dr. McElhaney testified that a Ford Bronco seat was used in the sled tests because "we're looking at what happened in the accident" and the performance of a child restraint depends upon the cushioning of the adult seat.

Sometimes the distinction is not clear between tests that recreate an accident and tests that demonstrate scientific principles for the purpose of forming expert opinion. An expert's opinion will invariably pertain to some aspect of the accident and thus the tests, while not designed to simulate the accident, will be related to that accident during testimony. The more blurred that distinction becomes, the greater the risk for prejudice.

Trial court judges are, necessarily, afforded "considerable discretion in determining the conduct of a trial, including the orderly presentation of evidence." Thweatt v. Ontko, 814 F.2d 1466, 1470 (10th Cir.1987). The sled tests were relevant evidence because they demonstrated physical principles that formed a basis for Dr. McElhaney's expert opinion and contradicted the opinion of the Gilberts' expert witness. The mere fact that on a few instances the tests were referred to in a way that may have suggested comparison with the accident does not by itself indicate that the Gilberts suffered impermissible prejudice. The instances referred to by the Gilberts were not of the extent or magnitude necessary to require reversal.

Finally, the Gilberts contend that the trial court erred in not giving the jury a limiting instruction concerning the extent to which the sled tests could be considered as evidence. Since the Gilberts failed to preserve this argument for review by objecting to the jury instructions used by the trial court, we will review for plain error. United States v. Sides, 944 F.2d 1554, 1561 (10th Cir.), cert. denied. --- U.S. ----, 112 S. Ct. 604, 116 L. Ed. 2d 627 (1991). See Fed.R.Civ.P. 51.

Plain error requires a finding of " 'fundamental error, something so basic, so prejudicial, so lacking in its elements that justice cannot have been done.' " Sides, 944 F.2d at 1561 (citation omitted). Ordinarily, a trial court's omission of a limiting instruction is not plain error unless the error may have caused a verdict not warranted under the law or "where it is apparent on the face of the record that a miscarriage of justice may occur." Herndon v. Seven Bar Flying Service, Inc., 716 F.2d 1322, 1330 (10th Cir.1983) (citation omitted), cert. denied, 466 U.S. 958, 104 S. Ct. 2170, 80 L. Ed. 2d 553 (1984).

A limiting instruction should have been given if the Gilberts had requested it. When tests are allowed into evidence for limited purposes, the jury should be "carefully instructed as to the extent to which they can use and consider [tests] of this kind." Robinson, 739 F.2d at 1485. Such an instruction helps prevent undue prejudice.

Nevertheless, the failure to give the limiting instruction was not plain error. In Robinson, we held that the trial court's failure to give a limiting instruction was not reversible error for two reasons: (1) appellants did not request the limiting instruction; and (2) appellants suffered no prejudice because "other evidence" informed the jury of the limitations. Robinson, 739 F.2d at 1485. Both of these factors are also present in this case. Not only did the Gilberts fail to request a limiting instruction, they were not clearly prejudiced by the court's omission since the jury was informed on numerous occasions that the sled tests were not intended to recreate the accident. The limited purposes of the sled tests were discussed in detail in the presence of the jury during Dr. McElhaney's voir dire and direct examination. Accordingly, the court's failure to instruct the jury is not grounds for reversal.

We next address the Gilberts' allegation that the district court erred when it excluded from evidence parts of certain Cosco memoranda pertaining to child restraint models which were not the subject of this lawsuit. The Gilberts argue that the memoranda contained important decisions by Cosco regarding cost cutting and types of foam which are relevant to the Model 580 child restraint.

The trial court decided that portions of Cosco's memoranda were "more prejudicial than probative" and thus excluded those portions from evidence. This court is generally reluctant to overturn a trial court's decision to exclude evidence deemed prejudicial. Marsee v. United States Tobacco Co., 866 F.2d 319, 321 (10th Cir.1989). A trial court is in a better position to balance, under Fed.R.Evid. 403, "the probative value of and need for the evidence against the harm likely to result from its admission." McAlester v. United Air Lines, Inc., 851 F.2d 1249, 1257 (10th Cir.1988); C.A. Associates v. Dow Chemical Co., 918 F.2d 1485, 1489 (10th Cir.1990). Thus, this court will reverse a trial court's decision to exclude certain evidence only upon a showing that the trial court clearly abused its discretion. C.A. Associates, 918 F.2d at 1489.

After reviewing the record, we are confident the trial court did not abuse its discretion. The trial court deleted portions from two Cosco memoranda, but allowed into evidence a third memoranda in its entirety. While the deleted portions of Cosco's memoranda contained some potentially relevant evidence on decisions concerning foam padding, these sections were likewise prejudicial because they involved different child restraint models with completely different designs. The court's decision that the deleted sections would have been more prejudicial than probative is well founded and not a clear abuse of discretion.

Moreover, the memoranda admitted in its entirety and the undeleted portions of the other two memoranda contained possibly damaging comments by Cosco concerning the superiority of Ensolite padding. The deleted portions added little to this already admitted evidence and, as the Gilberts indicated at trial, the deleted portions were mainly probative of Cosco's "corporate mentality." However, under Oklahoma law, corporate mentality is not an element in determining product liability, Woods v. Fruehauf Trailer Corp., 765 P.2d 770, 773-74 (Okla.1988), but instead pertains to punitive damages once liability is established. See Wagoner v. Bennett, 814 P.2d 476, 478 (Okla.1991) (defines punitive damages); Barnes v. McKinney, 589 P.2d 698, 700 n. 1 (Okla.App.1978) (punitive damages "recoverable only as an incident to the recovery of compensatory damages"). Thus, we believe that the Gilberts did not suffer a manifest injustice as a result of the trial court's decision. See Thweatt, 814 F.2d at 1470 (a trial court's decisions regarding the orderly presentation of evidence are generally not disturbed "absent a manifest injustice to the parties").

The Gilberts' final issue on appeal is whether the trial court erred when it refused to accept the Gilberts' proffered jury instruction concerning the adverse inference rule. Specifically, the Gilberts contend that evidence adduced at trial indicated the existence of memoranda pertaining to the Model 582 child restraint device, yet during discovery, Cosco produced only one such memorandum. The Gilberts requested the court to instruct the jury that if the jury determined Cosco could have produced material evidence within its control but failed to do so, the jury was permitted to infer the unproduced evidence was unfavorable to Cosco.5

When exercising diversity jurisdiction, state law determines the substance of the jury instructions, but the decision to grant or deny a jury instruction is procedural and, therefore, controlled by federal law. Farrell v. Klein Tools, Inc., 866 F.2d 1294, 1296 (10th Cir.1989); Brownlow v. Aman, 740 F.2d 1476, 1490 (10th Cir.1984). The substantive law of Oklahoma dictates that the jury instruction proffered by the Gilberts concerning the adverse inference rule is generally impermissible. "[N]o instruction on the inference arising from failure to produce evidence or a witness should be given. The main reason is that the matter is a subject for argument to the jury and should not be given undue recognition by the court." Witt v. Martin, 672 P.2d 312, 318 (Okla.App.1983); Oklahoma Uniform Jury Instruction, Civil No. 3.11 (1981). It is unnecessary for us to decide whether, under Oklahoma law, circumstances could exist which might permit a trial court to instruct a jury as to the adverse inference rule because the Gilberts' argument is also without merit under federal procedural law.

Jury instructions on the adverse inference rule are permissible in federal court when there exists an "unexplained failure or refusal of a party ... to produce evidence that would tend to throw light on the issues." Gumbs v. International Harvester, Inc., 718 F.2d 88, 96 (3d Cir.1983). See also 3 Edward J. Devitt et al., Federal Jury Practice and Instructions § 72.16 (4th ed. 1987). For the adverse inference rule to be applicable, certain factors must generally be present. The Eighth Circuit, for example, requires the following: "(1) it appears that the documentary evidence exists or existed; (2) the suppressing party has possession or control of the evidence; (3) the evidence is available to the suppressing party, but not to the party seeking production; (4) it appears that there has been actual suppression or withholding of evidence." Evans v. Robbins, 897 F.2d 966, 970 (8th Cir.1990) (citing 31A C.J.S. Evidence § 156(2) (1964)). See also Spesco v. General Electric Co., 719 F.2d 233, 239 (7th Cir.1983); Gumbs, 718 F.2d at 96.

A trial court's decision to give or refuse an adverse inference instruction is reviewed for an abuse of discretion. See Wilson v. Merrell Dow Pharmaceuticals Inc., 893 F.2d 1149, 1150 (10th Cir.1990) (discretionary decision by trial court not to give adverse inference instruction when party failed to produce a witness). See also United States v. St. Michael's Credit Union, 880 F.2d 579, 597 (1st Cir.1989). Furthermore, "[t]he evidence necessary to justify an instruction must be more than conjecture and speculation." Brownlow, 740 F.2d at 1490.

The Gilberts offered some evidence suggesting that Cosco may have possessed other memoranda concerning Model 582 which it did not produce and Cosco apparently introduced no evidence to the contrary.6 Even if we assume this evidence was sufficient to permit the trial court to instruct the jury as to the adverse inference rule in federal court, it is clearly not so compelling as to require such an instruction. Therefore, even if circumstances might exist where Oklahoma law does not prohibit a jury instruction on the adverse inference rule, the trial court's decision not to so instruct the jury was not an abuse of discretion in this case.

In summary, we AFFIRM the district court and hold that the court did not abuse its discretion by admitting evidence of the sled tests or by excluding portions of Cosco's memoranda. Furthermore, the court did not err in refusing to instruct the jury as to the adverse inference rule.