On the evening of May 28, 1977, fire destroyed the Beverly Hills Supper Club in Southgate, Kentucky. One hundred sixty-five patrons and employees perished in the fire and many others were injured. Extensive litigation followed in both the State and Federal courts. Underlying this appeal is a class action commenced in the United States District Court for the Eastern District of Kentucky, based on diversity of citizenship. The class consists of the legal representatives of the persons killed and approximately thirty-five individuals who claimed to have been injured in the fire. Plaintiffs named as defendants several manufacturers of “old technology” aluminum branch circuit wiring, claiming those materials had been installed in the supper club and had caused the fire.1
Shortly before the trial was scheduled to begin, the trial judge ordered that it be bifurcated. The jury first would consider the question of “causation in fact.” If aluminum wiring were found to be a cause of the fire, the jury, would then determine questions of liability and damages.
Plaintiffs’ theory at trial was that the fire began in a “dead” or empty space within the north wall of a cubbyhole next to the Zebra Room, located on the first floor of the Supper Club.2 Plaintiffs asserted that the fire originated at an aluminum duplex receptacle. The receptacle, a standard electrical outlet into which electrical appliances are plugged, was allegedly located in the cubbyhole and connected to aluminum branch circuit wiring.
Plaintiffs claimed that, due to a number of physical characteristics of old technology wiring, heat developed at the connection of the aluminum branch circuit wiring to the receptacle,3 and that this heat eventually *211ignited the wooden studs and other building materials in the wall. Plaintiffs claimed that the heat finally caused an open flame which spread undetected within the wall for approximately one to one and one-half hours before flame broke through the wall and directly engaged the Zebra Room itself.
Defendants responded that the receptacle in the cubbyhole was not proved to have been wired with aluminum branch circuit wiring. They claimed that the fire more likely began due to copper wiring of an electrical pump that was connected to a water fountain located in front of a staircase on the north side of the north wall. The defendants also suggested that the fire started as a result of numerous fire code violations found to have existed in the club.
After twenty-two days of trial over a period of eleven weeks, the jury returned a special verdict answering in the negative the question whether the connection of old technology aluminum wired to an electrical device caused the fire at the Supper Club. Based upon that finding, the trial judge entered a general judgment in favor of the defendants. Plaintiffs moved for a mistrial, for judgment notwithstanding the verdict and for a new trial. The trial judge denied all motions. These appeals followed.
While several issues are raised in these appeals, one error, improper experimentation by a juror, is of such importance that it alone mandates vacating the judgment and remanding for new trial or other proceedings. Of the numerous other issues raised, therefore, we address only those relevant to the disposition of this appeal or those whose resolution may facilitate any proceedings on remand.
I
During the trial, one of the jurors performed an improper experiment when he investigated the condition of the aluminum wiring and connections in his home. He then reported his findings to other jurors, findings which were factually at odds with plaintiffs’ theory of how the fire began.
Plaintiffs sought to show throughout the course of the trial that aluminum branch wiring is more likely to overheat and cause fires than is copper wiring. Carl Duncan, qualified as an expert in electrical fire origin, testified that an early indication of the degenerative process leading to overheating is that binding screws holding the wire appear to be loose. This loosening, he testified, aggravates the inability of the wire to conduct electricity. As the process allegedly occurs over a period of 5 to 10 years, plaintiffs characterized aluminum wiring systems as “time bombs.”
Following this testimony, the juror examined receptacles in his home. He pulled receptacles from their boxes and checked the binding head screws for tightness. He also looked for receptacles manifesting later stages of degeneration. His experiment tended to contradict evidence presented by the plaintiffs on the hazards of aluminum wiring. The juror found nothing wrong *212with his own receptacles which, he claimed, had been installed eleven years earlier. The juror believed that, under plaintiffs’ theory, the receptacles would have been in place long enough for some degeneration to have occurred. He also found that the screws holding aluminum wiring to his electrical devices in his outlets were tight.
After the verdict for the defendants had been rendered and the jury discharged, the juror wrote an anonymous letter to the Kentucky Enquirer, a newspaper of general circulation in northern Kentucky. In the letter, he explained the reasons for his decision and challenged the validity of the plaintiffs’ evidence when compared to findings in his own home.4 He also expressed the view that the fire was caused by numerous fire code violations found in the Supper Club.
In the post-trial evidentiary hearing, the trial judge learned that the juror communicated this information to at least six other jurors during the course of the trial. At least one juror recalled having privately discussed this matter with the experimenting juror during the course of the jury deliberations. Thereafter, plaintiffs moved for a mistrial.
The trial judge denied plaintiffs’ motions for mistrial, for judgment notwithstanding the verdict, and for a new trial based upon the juror’s conduct, observing:
In the context of the trial length, the quantity of evidence presented and the number of witnesses called by each side, this action by a juror appears to be of minor consequences and not a sufficient intrusion upon the deliberative process as to require the setting aside of the jury verdict.
*213In Re: Beverly Hills Fire Litigation, C. No. 77-79 at p. 8 (E.D.Ky. April 7, 1980).
Plaintiffs claim that the juror’s investigation was an impermissible experiment requiring that the verdict be set aside. Defendants respond that the juror’s action was harmless. They further stress that the inflammatory language regarding “time bombs” made it almost inevitable that the experiment would occur. Since the trial was expected to be a long one, the defendants claim that it is unreasonable to expect that a juror would wait until the end of the trial to inspect his own home for dangers which the plaintiffs had so characterized. Defendants contend that the juror’s investigation was not in the nature of an “experiment” nor a purposeful attempt to develop information about the case being tried, but was more in the nature of a personal and unrelated experience which could not have affected the judgment of that juror or those to whom he communicated that information.
Upon a careful examination of the record and the applicable law, we regretfully conclude that the jury verdict was impermissibly tainted by what can only be characterized as an improper juror experiment.
The general rule is that a juror may not impeach his verdict. Fed.R.Evid. 606(b). See McDonald v. Pless, 238 U.S. 264, 35 S.Ct. 783, 59 L.Ed. 1300 (1915); Womble v. J. C. Penney, 431 F.2d 985 (6th Cir. 1970); Gault v. Poor Sisters of St. Frances, 375 F.2d 539 (6th Cir. 1967). The rule ensures that jurors will not feel constrained in their deliberations for fear of later scrutiny by others. Further, it guarantees that jurors cannot manipulate the system when their views are in the minority by repudiating an earlier verdict and obtaining a mistrial.5 It thus advances important policy considerations.
An exception to the general rule has developed where external factors are shown to have existed. As stated by Judge Peck in Womble, supra, “the [general] rule does not preclude inquiry into any extraneous influences brought to bear upon the jury in order to show what the influences were and whether they were prejudicial.” 431 F.2d at 989. See Stiles v. Lawrie, 211 F.2d 188 (6th Cir. 1954). Rule 606(b) specifically allows inquiry into external influences upon a jury.6 This exception is necessary to assure that the parties receive a fair trial and that the integrity of the system itself is maintained. See United States v. Ferguson, 486 F.2d 968 (6th Cir. 1973); Briggs v. United States, 221 F.2d 636 (6th Cir. 1955).
Defendants contend that the juror’s act in checking his own wiring was “too simple and natural to be deemed in any sense an experiment.” Stone v. City of Florence, 203 S.C. 527, 28 S.E.2d 409, 410 (1943). Rather, they suggest, knowledge gleaned from that activity is in the nature of general knowledge or common experience, which a juror is entitled to consider in his deliberations. In our view, the juror experimentation here was more than a mental or emotional reaction or expression. The experiment, in fact, injected extraneous information into the trial. While the juror’s con*214duct here may very well have been “simple and natural,” it was, under any test, an experiment.
It is not uncommon for a court to instruct a jury that jurors may consider the evidence in light of their general knowledge and experiences of life, and the trial judge gave a similar instruction here.7 One function of the jury is to infuse a practical sense into the legal theories offered at trial. Courts therefore have generally found discussions regarding such experiences quite unobjectionable. See Stephens v. City of Dayton, 474 F.2d 997 (6th Cir. 1973); Womble, supra, 431 F.2d at 989.
On the other hand, our court has not hesitated to declare mistrials where the activity went beyond mere general knowledge and was instead a response to the facts of the case at hand. An investigation is improper where it “amount[s] to additional evidence supplementary to that introduced during the trial.” Womble, supra, 431 F.2d at 989.8 Thus, our court ordered a new trial where a juror brought into the jury room a manual published by the Highway Department, but not introduced at trial, purporting to show the length of skid marks made by automobiles at different speeds. Stiles v. Lawrie, 211 F.2d 188 (6th Cir. 1954). Similarly, our court held that a new trial was properly granted where during the trial a juror had travelled to plaintiff’s.property to view certain cattle which were the subject of litigation, reporting back to the jury that he thought that the cattle looked like “about the best looking he had seen in a long time.” Aluminum Company of America v. Loveday, 273 F.2d 499 (6th Cir. 1959), cert. denied, 363 U.S. 802, 80 S.Ct. 1236, 4 L.Ed.2d 1146 (1960).
In the present case the juror’s investigation had the effect of putting both himself and the other jurors with whom he discussed his findings in possession of evidence not offered at trial. The juror tested the outlets for tightness and compared his results to the evidence in the case, observing in his letter to the Enquirer that “my outlets are still tight after 11 years of use, how come these are not loose?” It is clear that additional evidence was before at least one member of the jury.
In fairness, it is apparent that the juror was troubled by references made to the dangerous propensities of aluminum wire.9 The juror indicated his concern at the evidentiary hearing following the trial:
Q: Approximately when did that [his check of the fuse box] occur?
A: Sir, the first part of the trial, it come out that they they was talking about aluminum wire wired to outlets that was like a time bomb; it could go off any time, and they brought out slides and they was showing the outlet glowing and charring and they showed the studs burning; and that’s why I went home and checked them that night. I didn’t check the aluminum for tests or experiments or to see if it was safe. I was concerned with my family.
Transcript of Post-trial Examination of Jurors at 39. Whatever the juror’s motives, however, the potential for prejudice inherent in an out-of-court experiment in this case remains. It is impossible to determine without the benefit of a vigorous cross-examination following formal introduction of evidence whether an experiment duplicated *215what actually occurred in the case. Highly misleading results can follow. Here, the juror assumed his electrical outlets were constructed in the same manner as those offered by plaintiffs. He then made conclusions directly related to the outcome of the case, and he communicated these facts to other jurors. No opportunity was afforded either litigant to determine whether the juror’s wiring was aluminum and, if so, whether it was “old technology” wire. They further were unable to consider other conditions that may have accounted for the differing results. In short, the juror considered evidence from an experiment which was not subject to scrutiny or cross-examination by any party.
Our circuit has recognized that such an error can rarely be viewed as harmless. The jury’s receipt of such extraneous information “requires that the verdict be set aside, unless entirely devoid of any proven influence or the probability of such influence upon the jury’s deliberations or verdict.” Stiles, supra, 211 F.2d at 190. Here, the juror’s letter made clear that the results of his investigation were a factor in his decisionmaking. While influence on a single juror may be enough to necessitate reversal and remand, see Loveday, supra, 273 F.2d at 500, the error is particularly grievous because a unanimous verdict was required and the juror communicated his findings to other jurors.10 We conclude that the controlling law permits no alternative to reversal.11
In remanding, the question remains whether anything can be done to avoid repetition of this incident. The trial judge instructed the jury not to seek outside information. Obviously, the juror’s concern for his safety played a substantial part in the experiment which was conducted despite the instruction. While due regard ought to be given to the customary latitude which counsel need in order to bring life and meaning to the case, we suggest that the court on retrial may wish to confer with counsel on the means by which they can avoid alarming the jury. It would probably not be inappropriate under the circumstances if counsel were cautioned in the use of inflammatory language. Perhaps the trial court itself could give more explicit precautionary instructions, provided, of course, that they themselves were balanced and did not create the very hazard they might be designed to avoid. All of this, however, we leave to the good judgment and discretion of the trial court.12
*216II
Two claimed errors concern the severance of causation from other issues in the trial.
Plaintiffs first assert that the trial judge had no authority to isolate the issue of causation. It is well settled that the ordering of separate trials is within the sound discretion of the trial judge. Kosters v. Seven-Up Co., 595 F.2d 347 (6th Cir. 1979); Crummet v. Corbin, 475 F.2d 816 (6th Cir. 1973); Moss v. Associated Transport, 344 F.2d 23 (6th Cir. 1965). Plaintiffs argue, however, that no case law supports severance of the issue of causation; rather, only severance of liability and damages has been allowed. This view-point is inconsistent with the language of Fed.R.Civ.P. 42(b), which provides in part:
Separate Trials. The court, in furtherance of convenience or to avoid prejudice, or when separate trials will be conducive to expedition and economy, may order a separate trial ... of any separate issue
(emphasis added). There is thus no reason to adopt a different standard with regard to severing causation.
As the Rule indicates, and as our circuit has recognized, the court in ordering separate trials must consider several issues such as potential prejudice to the parties, potential confusion to the jury, and the relative convenience and economy which would result.13 See, e.g., Roster, supra, 595 F.2d at 355-56. A balance of these concerns led the Eighth Circuit to approve bifurcation of the causation issue in Beeck v. Aquaslide ’N’ Dive Corp., 562 F.2d 537 (8th Cir. 1977). The court observed:
Evidence of plaintiffs’ injuries and damages would clearly have taken up several days of trial time, and because of the severity of the injuries, may have been prejudicial to the defendant’s claim [that it did not manufacture the product that injured plaintiff] .... Judicial economy, beneficial to all the parties, was obviously served by the trial court’s grant of a separate trial.
Id. at 542.
The Beeck court also recognized that whether resolution of a single issue would likely be dispositive of an entire claim is highly relevant in determining the efficacy of bifurcation. The trial judge in the present case considered both the projected length of the trial and the likelihood that a resolution of the causation issue could shorten it.14
The conclusion of the trial judge has support in the record. The trial on causation alone took thirty-two days. Proof regarding the further issues of liability among the numerous defendants and of damages would be extensive and expensive.15 It therefore was reasonable for the *217trial judge to conclude that litigation of those issues should be avoided if they might be mooted by an adverse finding on the causation issue. The value of severance in expediting this case has already been proved. Had the juror experiment required a mistrial after the entire case had been tried, many more weeks of effort would have to be repeated.
A strong argument can, it is true, be made against the bifurcation of a trial limited to the issue of causation. There is a danger that bifurcation may deprive plaintiffs of their legitimate right to place before the jury the circumstances and atmosphere of the entire cause of action which they have brought into the court, replacing it with a sterile or laboratory atmosphere in which causation is parted from the reality of injury. In a litigation of lesser complexity, such considerations might well have prompted the trial judge to reject such a procedure. Here, however, it is only necessary for us to observe that the occurrence of the fire itself, a major disaster in Kentucky history by all standards, was generally known to the jurors from the outset. Further, the proofs themselves, although limited, were nonetheless fully adequate to apprise the jury of the general circumstances of the tragedy and the environment in which the fire arose. As a result, we hold that the trial judge did not abuse his discretion in severing the issue of causation here. In so ruling, however, we emphasize that the decision whether to proceed in the same manner on any retrial is within the discretion of the trial judge, who will undoubtedly be benefitted from the experiences in the first trial as he makes his decision.
The plaintiffs next claim that severance, even if not in itself error, nonetheless resulted in the improper exclusion of evidence relevant to the issue of causation. Plaintiffs offered several documents as evidence supporting their contentions that aluminum wiring has a greater propensity to cause fires and that some defendants knew of this propensity. The trial judge excluded some documents in their entirety,16 and allowed certain evidence admitted upon deletion of references to individual defendants.17 In Re: Beverly Hills Fire Litigation, C. No. 77-79 (E.D.Ky. Jan. 7, 1980).
In reviewing the documents at issue, the trial judge determined that several documents, though admissions as to some defendants, were hearsay as applied to other defendants. See Fed.R.Evid. 802; 801(d)(2). He determined further that documents indicating knowledge of some defendants were perhaps relevant to liability but were not relevant to causation, which alone was before the jury. See Fed.R.Evid. 401. Finally, he determined that those documents which were relevant to the issue of propensity of aluminum to cause fires were nonetheless inadmissible because their probative value was outweighed by the potential for prejudice if a. specifically identified defendant were singled out. See Fed.R. Evid. 403.
Evidence is not admissible in all eases where it is relevant. Fed.R.Evid. 403 provides:
Although relevant, evidence may be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or mis*218leading the jury, or by considerations of undue delay, waste of time, or needless presentation of cumulative evidence.
Under this rule, admission of such evidence is placed within the sound discretion of the trial court. United States v. Brady, 595 F.2d 359 (6th Cir. 1979).
[H] We have reviewed the documents deemed inadmissible by the trial judge and find that he did not abuse his discretion in excluding them. To the extent that the documents did not refer only to the knowledge of the defendants but also to the propensity of aluminum to cause fires, they were relevant. They were, however, largely cumulative of plaintiffs’ evidence respecting the characteristics of aluminum wire.18 Moreover, as the trial judge observed, they may have improperly prejudiced the jury.
The trial judge on remand may decide not to isolate causation from liability. In that case, the evidence could in the trial court’s discretion be offered against some defendants with an appropriate limiting instruction. See Fed.R.Evid. 105. If the case is again bifurcated, it again will be within the trial judge’s discretion whether and under what circumstances to admit the documents. Although we see the dilemma caused by these exhibits as one argument working against the employment of bifurcation here, it does not constitute a basis for denying it altogether, if in the discretion of the trial judge it remains the most efficient method of managing this complex litigation on remand.
Ill
The plaintiffs in their appeal and the defendants in their cross-appeal each argue that their respective motions for a directed verdict should have been granted at the conclusion of the trial. Our careful review of the extensive record reveals a hotly contested issue of causation that properly was submitted to the jury. We do, however, believe that the motion of defendants is worthy of extended consideration.
Defendants claim that plaintiffs offered no proof tending to show that the duplex receptacle actually failed. They admit that evidence was offered which, if believed, tended to show that aluminum wire failed more often than did copper wire, that the fire started in the north wall of the cubbyhole, and that a duplex receptacle was mounted on that wall. Defendants assert, however, that plaintiffs’ only proof that aluminum wiring actually failed is that it is statistically more likely that aluminum will fail. They add that plaintiffs offered no proof that the symptoms of aluminum wiring failure actually occurred, nor did they show that the duplex failed to function.
Defendants also claim that the jury had no method to determine that aluminum rather than copper wiring failed. They thus liken the present case to Rollins v. Avey, 296 S.W.2d 214 (Ky.1956). In Rollins, plaintiff had a furnace installed in her living room. Thirty or forty minutes thereafter, a gas explosion occurred in the house. The plaintiff had three other gas appliances in her home, and there was no evidence that the furnace exploded or was destroyed. The Kentucky Supreme Court affirmed a directed verdict for defendants, finding that plaintiff’s failure to determine which instrumentality failed was fatal to plaintiff’s case. Similarly, defendants assert, plaintiffs have not pinpointed the instrumentality causing the fire here.
We are, of course, bound to apply Kentucky law to the substantive issues in this diversity case. Erie R.R. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938). In determining whether plaintiffs presented enough evidence of cause in fact to avoid a directed verdict, the relevant standard in consideration of such a motion is that
plaintiff is entitled to the most favorable inferences and construction attributable *219to the evidence. If such evidence is so regarded and substantially tends to support the cause of action, the verdict should not be directed against him.
Fields v. Western Kentucky Gas Co., 478 S.W.2d 20, 22 (Ky.1972).19
The Kentucky view regarding when evidence of causation is sufficient to reach a jury is discussed in Huffman v. SS. Mary & Elizabeth Hospital, 475 S.W.2d 631 (Ky. 1972):
[Ljegal causation may be established by circumstantial evidence. While reasonable inferences are permissible, a jury verdict must be based on something other than speculation, supposition or surmise. The type of evidence that will support a reasonable inference must indicate the probable as distinguished from a possible cause .... There must be sufficient proof to tilt the balance from possibility to probability.
Id. at 633 (citations omitted) (emphasis in original).
Where an incident could result from more than one cause, plaintiff tips the balance from possibility to probability only by ruling out other theories of causation:
[Wjhere an injury may as reasonably be attributed to a cause that will excuse the defendant as to a cause that will subject it to liability, no recovery can be had.
Sutton’s Adm’r. v. Louisville & N. R. Co., 168 Ky. 81, 181 S.W. 938, 940 (1916). That is, it must be more likely than not that the incident in question occurred as the result of the cause posited by plaintiff.
Thus, defendants invariably receive a judgment in their favor as a matter of law in Kentucky where plaintiffs are unable to isolate one cause, either by direct evidence or, more relevant to the present case, by eliminating other possible causes. For example, summary judgment for defendant was affirmed where a pool of gas exploded but it was unclear whether mechanical defects, human intervention or other causes ignited the pool. Highway Transport Co. v. Daniel Baker Co., 398 S.W.2d 501 (Ky.1965). Similarly, an argument that brake failure occurred because the brakes at issue did not conform to factory specifications was rejected where other causes such as dirt in the brake lining, improper adjustment of the brake drum or improper tire pressure were not eliminated as causes. Midwestern V. W. v. Ringley, 503 S.W.2d 745 (Ky.1973). In Rollins, supra, plaintiffs offered no evidence that excluded other gas appliances as possible causes.20
Conversely, the court in Fields v. Western Kentucky Gas Co., 478 S.W.2d 20 (Ky.1972), found that plaintiff had presented a causation theory sufficient to go to a jury. An explosion occurred in a manhole, and plaintiff’s witness testified that a seepage of natural gas was the cause. More important, he was able to state that the other proposed causes (sewer gas or a solvent used to seal pipes) were improbable ones. In reversing a directed verdict for defendant, the court stated:
Although a jury may not be permitted to speculate when the probabilities of an event’s having happened in one of two or more ways are equal and there is no evidence as to which way it did happen, it is neither legal “speculation” nor “conjecture” when a jury finds as a fact that an event happened by reason of a particular cause when the evidence on behalf of a *220party, if believed, is sufficient to show that it is more likely than not that the event occurred as a result of the cause so found.
Id. at 22. Thus, a directed verdict for defendants would be improper where plaintiffs’ evidence, if believed, could establish that other possible causes could not have precipitated the Beverly Hills fire, thereby making it more likely than not that it was caused by aluminum wiring.
The evidence here, if believed, could convince a jury that the fire began within the north wall of the Zebra Room cubbyhole.21 It also tended to show that the fire started at a duplex receptacle located on that wall.22 The jury could have found from the evidence that the duplex receptacle was wired with aluminum branch wiring,23 and that aluminum wiring failed fifty to fifty-five times more often than did copper wiring. Dr. Jesse Aronstein, a mechanical engineer, testified that other aluminum devices in the building were found in various stages of deterioration. Additionally, Eileen Druckman and Michael Sims testified regarding their perceptions of heat and odor, which could have buttressed plaintiffs’ theory regarding the symptoms of electrical overheating.24 Evidence was presented indicating that receptacles may continue to be operable while overheating, which the jury could have found was a response to defendant’s assertion that the wire did not fail because electrical devices continued to function.25
The evidence, if believed, also could have eliminated other causes for the fire such as *221devices in the adjoining alcove area,26 short-circuiting,27 negligent workmanship, the pump referred to by defendants,28 or code violations existing in the Club. Contrary to defendant’s assertions, the jury could have found that a light switch located nearby was not the source of the fire;29 further, they could have found that there were no other electrical devices in the cubbyhole area.30 Evidence regarding exclusion of these causes, if believed, tended to make it more probable than not that the aluminum wiring attached to the duplex actually failed. Unlike Rollins, the jury could have found that the instrumentality causing the fire had been identified. A directed verdict therefore would have been improper here.
Defendants argue that plaintiffs engaged in a “compounding of inferences” in proving causation, a method not condoned in Kentucky courts as it leads to verdicts based on surmise or speculation. See, e.g., Rollins, supra;31 Briner v. General Motors Corp., 461 S.W.2d 99 (Ky.1970). Those cases are inapposite to our result here.
Kentucky courts generally are convinced that the plaintiff is stacking inferences where plaintiff constructs a theory with several hypotheses, none of which is supported by either direct or circumstantial evidence.32 In the present case, each fact depended on by plaintiffs is supported by some evidence in the record for the jury to *222assess. Electrician William White testified that he placed aluminum wire in the receptacle. As suggested above, there was further circumstantial evidence which could have supported a finding that the fire originated there. Absent proof of the existence of the wire, the verdict may have been based on an impermissible stacking of inferences. White’s direct testimony avoided this result.
We obviously do not by our analysis mean to suggest that the Beverly Hills fire necessarily was caused by aluminum wiring. An extraordinary amount of circumstantial evidence was considered by the jury, which was subject to differing, yet reasonable interpretations. For example, the credibility of White, who defendants claim was induced to change his original story, was clearly at issue.
It may be that the vigorous and effective cross-examination of White induced the jury to disbelieve him. Also, defendants presented a great deal of information regarding other theories of causation, which the jury may have found more persuasive. It is not for us to pass on the credibility of White’s testimony, or of that of any other witness, as defendants would seem by their arguments to urge. We merely find that, giving the evidence a construction favorable to plaintiffs, enough evidence was presented to avoid a directed verdict in favor of defendants. See Fields, supra, 478 S.W.2d at 22.
IV
Plaintiffs also claim that the trial court erred in the form of special questions submitted to the jury.
The jury was provided with two alternative verdict forms, which stated:
We, the jury, unanimously find that connection of old technology aluminum wire to an electrical device caused the fire at the Beverly Hills Supper Club on May 28, 1977.
We, the jury, unanimously find that connection of old technology aluminum wire to an electrical device did not cause the fire at the Beverly Hills Supper Club on May 28, 1977.
Transcript at 7351-52. As noted earlier, the jury adopted the second statement.
Plaintiffs assert that the trial judge improperly charged the jury because he failed to define causation as a “substantial factor” in bringing about injury. Kentucky law, they assert, requires that the jury be so charged, e.g., Deutsch v. Shein, 597 S.W.2d 141 (Ky.1980).
Plaintiffs fail to note that the court expressly included such an instruction.33 We do not think it was necessary for the judge once more to have incorporated the term “substantial factor” in the verdict form, having carefully instructed the jury in that regard already. It might have been preferable to have done so, and we would see no harm on remand if the trial judge should incorporate such language. The trial judge may decide to revise the form in such a way as to meet the approval of all parties. Again, the question is properly within the discretion of the trial judge.
V
No other errors asserted by the plaintiffs merit extended discussion. The *223court did not err in permitting the employment of external portions of a model of the north wall. It was not found to be inaccurate, and it was used for demonstrative purposes by both plaintiffs and defendants. It further was not error for the trial judge to preclude the plaintiffs on rebuttal from eliciting proof that the interior construction of the model did not conform to the actual construction of the wall. Because the parties learned of the inaccuracies early in trial, the judge never allowed the jury to view the interior of the model. It would not have aided the jury’s consideration at all to have expanded the proofs further by developing this irrelevant information.
Neither did the trial court err in permitting cross-examination of Thomas McClory regarding numerous fire code violations found in the Beverly Hills Supper Club. McClory offered an opinion on the cause of the fire, thereby making other potential causes an appropriate subject of cross-examination. Similarly, Mr. Horton’s testimony regarding fire and smoke in another room of the club was relevant to whether the fire originated in the north wall and was properly admitted.
Finally, the plaintiffs assert that defense counsel was guilty of at least six instances of improper remarks during closing arguments. We deem it unnecessary to comment on any of these. They are unlikely to recur at retrial, at least in their original form. Suffice it to say that nothing called to our attention would have warranted reversal. Much of that complained of was responsive to equally robust argument by opposing counsel.
VI
In their cross-appeal, the defendants assert that the trial court erred in denying their motion to dismiss and for partial summary judgment on the basis that the action was barred by Ky.Rev.Stat.Ann. § 413.135 (Baldwin). That statute, referred to as Kentucky’s “no action” statute, was enacted in 1966 and provides in part as follows:
(1) No action to recover damages, whether based upon contract or sounding in tort, resulting from or arising out of any deficiency in the design, planning, supervision, inspection or construction of any improvement to real property, or for any injury to property, either real or personal, arising out of such deficiency, or for injury to the person or for wrongful death arising out of any such deficiency, shall be brought against any person performing or furnishing the design, planning, supervision, inspection or construction of any such improvement after the expiration of five years following the substantial completion of such improvement.
The Supper Club was substantially rebuilt in 1970 and 1971 following a previous fire in 1970. According to the evidence at the trial, no significant improvements were made to the Supper Club since that time. If controlling, section 413.135 would bar any claim resulting from allegedly defective materials installed before June 2,1972, five years before the first complaint was filed.34 On the other hand, if section 413.135 does not apply, the action would be timely as all claims were commenced within one year of the fire as required by the general tort statute of limitations. See Ky.Rev.Stat. Ann. § 413.140(l)(a) (Baldwin). Because the improvements occurred in 1970 and 1971, plaintiffs’ claims against these defendants are time-barred if section 413.135 is applicable, and an alternate basis for affirmance would exist.
The trial judge denied defendants’ motions, finding the statute did not apply to those providing materials for construction projects. He determined that the statute should be narrowly construed because application of the statute would work a harsh result on those whose claim was barred without notice, and because the class of persons to whom the statute applies is un*224certain.35 Since the judge did not view defendants as among the class of persons either named or intended to be protected by the statute, he found it unnecessary to decide whether the statute was offensive to Ky.Const. §§ 14, 54 and 241. At least at this juncture of the proceedings, we agree with the result reached by the trial judge, but not necessarily with his reasoning.
The Kentucky courts have not considered whether “materialmen” who design products for construction projects are within the contemplation of the statute. Where state law is unclear, we must “make a considered ‘educated guess’ ” as to how Kentucky courts would view the statute. Ann Arbor Trust Co. v. North American Co., 527 F.2d 526 (6th Cir. 1975).
We are not nearly so confident as the trial court that the statute ought to be narrowly construed. The statute requires that “[n]o action to recover damages . . . arising out of any deficiency in the design ... of any improvement to real property ... shall be brought against any person performing or furnishing the design .... ” Ky.Rev.Stat.Ann. § 413.135(1) (Baldwin) (emphasis added).36 Whether or not we may agree with the results, we cannot alter the language of the statute, which by its terms covers all persons furnishing such designs.
If that language is plain and unambiguous its meaning should be upheld as so expressed, uninfluenced by any unwise or unusual result that might follow the upholding of the plainly expressed writing or statutes ....
Reynolds Metal Co. v. Glass, 302 Ky. 622, 195 S.W.2d 280, 283 (Ky.1946).
Plaintiffs urge that the statute must be interpreted in light of its purpose, which was to protect those in the building industry from unlimited liability following collapse of the rule requiring privity of contract to recover.37 See generally Comment, Limitation of Action Statutes for Architects and Builders — Blueprints for Non-action, 18 Catholic U.L.Rev. 361 (1969). The Kentucky Court has suggested that “the intention to be gathered from employed language is one that it plainly expresses, and not the one that may have been in the mind of the composer, but which he failed to express.” Reynolds, supra, 195 S.W.2d at 283. See also Kentucky Ass’n of Chiropractors, Inc. v. Jefferson County Medical Soc’y, 549 S.W.2d 817 (Ky.1977) (intention ascertained from words employed in enacting the statute, not by surmising what was intended but not expressed). We see no evidence in the language of the statute indicating an intent to exclude materialmen from protection of the statute.
It is therefore our “educated guess” that a Kentucky Court would disagree with those cases reviewing similar statutes which held that materialmen were not covered by such statutes. See, e.g., Kittson City v. Wells, Denbrook & Assoc., Inc., 308 Minn. 237, 241 N.W.2d 799 (Minn.1976). Review of this history of similar “no action” statutes in several states and in Kentucky itself discloses no reason to exclude materialmen from the statute’s coverage. The general trend to eliminate privity as a requirement of an action for breach of implied warranty *225or for negligence opened a broad new field of potential liability.38 Additionally, courts found that a cause of action did not arise, nor did the statute of limitations commence to run, until plaintiff was injured. Engineers, contractors and architects were thus confronted with potential liability for conduct which may have occurred many years earlier, well after any opportunity to provide a reasonable defense or to have preserved any evidence had passed. See generally Comment, supra, 18 Catholic L.Rev. 361. Consequently, those groups strongly urged adoption of “no-action” statutes pertaining to improvements to realty. Because those designing parts intended to become part of the realty were affected similarly by these changes, it can safely be assumed absent exclusionary language that they are protected.39
Assuming that Kentucky’s “no action” statute was intended to extend to a class of persons including defendants, we must consider whether Kentucky would view this statute as consistent with Ky.Const. §§ 14, 54 and 241. We believe it would not.
The Kentucky Supreme Court considered the constitutional defects of section 413.135 in Saylor v. Hall, 497 S.W.2d 218 (Ky.1973). In Saylor, the court declined to apply section 413.135 to bar suit against a home-builder for personal injuries caused by defects in construction where construction occurred before the statute was passed. The court noted that several state courts have considered similar statutes to be valid. It also observed that our court in Lee v. Fister, 413 F.2d 1286 (6th Cir. 1969), had applied the Kentucky statute without considering the effect of several Kentucky constitutional provisions upon its validity. Saylor, supra, 497 S.W.2d at 222.
The Kentucky court then determined that the Kentucky constitution required a different analysis than that undertaken in other *226states. It found that the Kentucky constitution precludes the legislature from abolishing common law rights of action for injuries or death caused by negligence.40
The court noted that a cause of action does not accrue until the time of injury. It concluded that, because section 413.130 could cut off a common law right to recovery before a cause of action even accrues, it violated Ky.Const. §§ 14, 54 and 241. The court stated:
The statutory expressions as they relate to actions based on negligence perform an abortion on the right of action, not in the first trimester, but before conception.
Saylor, supra, 497 S.W.2d at 224.
The court took care to write narrowly and to limit its holding to the particular facts of that case. It is thus possible to limit Saylor to those cases where construction occurred before the statute was promulgated. We think the more reasonable interpretation of Saylor, however, is that it applies to the present case, even though improvements to the club were made following promulgation of the statute. If the statute were applied to bar plaintiffs’ claim, it would have cut off their right of suit before it accrued. We thus find the Saylor rationale applicable here.41
This case demonstrates once more the fundamental problem federal courts encounter in diversity cases. Jurisdiction conferred upon them by act of Congress forces upon the federal courts decisions of uniquely state law, which they are ill-equipped to render in any authoritative way. Thus, in the absence of any authoritative decision of the Kentucky Supreme Court on this issue, we are left with only the ability to guess what that court would do, given its limited rulings in Saylor. Upon the basis of that rationale, we conclude that, were it faced with the circumstances of this case, the Kentucky Supreme Court would probably rule the entire statute violative of the above cited Kentucky constitutional provisions. We therefore conclude that the trial judge’s denial of the motions for partial summary judgment and for dismissal on this issue was proper, and we are unwilling *227to affirm judgment for defendants on this alternative basis.
To recapitulate, it is our tentative conclusion that the trial judge in this case did not err in refusing to apply the Kentucky “no action” statute. We are of the opinion that it is doubtful that the statute ought to be narrowly construed given language of the statute and the history of such legislation generally. We are nonetheless tentatively of the opinion that the Kentucky constitution precludes application of the statute to bar suit against defendants, as it would have extinguished a common law right of action before injury occurred and before plaintiffs had any reasonable opportunity to seek redress in court. Absent any further expression on the subject by the Kentucky Supreme Court, or an authoritative decision by an intermediate appellate court, it would be our opinion that the statute should not apply. If the Kentucky courts render an authoritative decision bearing upon the question while this case is open, the district court remains free to reconsider its position both in the light of that change and of any changes in the posture of the case on remand.
CONCLUSION
Our decision to reverse is most regretfully made, as the length of time it has taken to reach it may suggest. The trial was generally a fair one, vigorously and effectively presented by able counsel before a skillful and experienced trial judge who cannot be faulted for the events which have occasioned the reversal. We are mindful of the trial judge’s observation, earlier stated in an unpublished opinion of this court, that “[ejxperience teaches that while every additional day of trial increases the possibility of error, it correspondingly reduces the risk that any single error may have prejudicial effect upon the ultimate result.” In Re: Beverly Hills Fire Litigation, C. No. 77-79 (E.D. Ky. April 7, 1980), quoting United States v. Arvant, No. 78-5345 (6th Cir. August 27, 1979). Nonetheless, the recited facts of the improper experiment and its use in the jury deliberations are too compelling and too fraught with potential for prejudice to be ignored.
Reversed and remanded.
ORDER
. Plaintiffs alleged three theories of liability in their complaint: concert of action, alternative liability, and enterprise liability. The trial judge granted summary judgment in favor of defendants on the issues of alternative and enterprise liability, stating that Kentucky recognized neither theory as a basis for liability. He allowed plaintiffs to go forward on a theory of concerted action. In Re: Beverly Hills Fire Litigation, C. No. 77-79 (E.D.Ky. Nov. 14, 1979).
In his November 14 order, the trial judge outlined his analysis of Kentucky law of concerted action. He indicated that “[tjhis doctrine imposes joint liability against ‘[a]ll those who, in pursuance of a common plan or design to commit a tortious act, actively take part in it.’ ” Id at 8 (citations omitted). He determined that three elements must be satisfied in order to prevail on this theory: a causal relation between the act and injury; cooperative or concerted activities by defendants; and violation thereby of a legal standard of care. Id at 8-9. He added that concerted activity can be proved in Kentucky either by explicit or tacit agreement among defendants. Id. at 12. This order is not at issue in this appeal.
. The Zebra Room, comprised of the Room proper, an alcove section and a cubbyhole section, is an “L-shaped” room in the front, or southeast section of the building. It is located on the first floor of the building. At approximately 18' X 28', the Zebra Room proper is one of the smallest rooms in the large supper club, which occupies over an acre in area. Immediately to the west of the Zebra Room proper is the alcove section, approximately 10' X 10'. The cubbyhole section is to the west of the alcove section, approximately 7' X 10'. It is separated from the alcove by double doors. Immediately west of the cubbyhole is the main bar. Approximate measurements were determined from a floor plan of the Beverly Hills Supper Club submitted in evidence. All directional references throughout this opinion are based on an “assumed north” indicated on the floor plan.
The Zebra Room proper, the alcove and the cubbyhole all are bordered by a common “north wall.” A staircase is located on the north side of the north wall. A fountain is located directly to the north of the staircase.
. Carl Duncan, qualified as an expert in electrical fire origin, testified that there were a number of characteristics of aluminum wire, in comparison to copper wire, which made it subject to overheating. First, large size wire was used because aluminum wiring is less conductive than copper wiring; yet, the screws used to hold the aluminum wiring in place were disproportionately small. Duncan claimed this caused a lesser percentage of conducting material to be in contact with the binding screw than would have occurred with copper wire, which made it more difficult for electricity to *211be transferred. Duncan claimed this phenomenon itself caused overheating. Second, he found that aluminum was more easily nicked, fractured or broken during installment than was copper; he claimed that damage to the wire reduced its conductivity at various spots. Additionally, an oxide film (rust) immediately forms on exposed aluminum when exposed to the atmosphere. A similar film forms on copper, but on copper it is conductive whereas on aluminum wire it is not. He concluded this made it more difficult for electricity to flow from wire to receptacal, therefore causing a heat buildup. Finally, Duncan testified that aluminum wiring has a tendency to “creep,” which decreases further the wire’s ability to conduct electricity. He defined “creep” stating:
Creep is that phenomenon of a material to flow away from pressure, and it’s also referred to as cold flow. What happens is as you are torquing the screw on the material, the material itself has a tendency to flow away from pressure, and that is a physical phenomenon of the material.
The net effect of that creep or cold flow is that additional surface area of the conductor is exposed to oxidation and when relaxation occurs oxide film is formed and additional heat generated because of the, again, oxidation formation on the conductor and the energy having to break that oxide film down to maintain continuity and conductivity.
Transcript at 728-29.
. The letter states:
I want to let the people know how I reached a decision on the Beverly Hills Trial. This is the first time I have ever been called for Jury Duty ... it was something I think everyone should do just to find out how the system works. It could have been shortened — it was rather repetitious, and a two hour lunch — I’m used to a half hour.
The receptacles that were taken out of the front part of the building (not in the fire) were wired with aluminum branch circuit wiring.
These were sent up state and put on a test rack. The Jury never had an opportunity to see these receptacles even though they were in the court room. New receptacles however were passed showing how aluminum could be wired — some correctly & some wrong.
In our deliberations, the receptacles that were put on test were with the many artifacts of the case. This is the first time I got to see them, (about 10 or 12 receptacles)
Two of these receptacles had the [aluminum] wire counter clockwise under the binding head screw which is wrong.
When tightening a screw clockwise the wire under the screw should be clockwise also.
Two more outlets or receptacles, the [aluminum] wire was loose under the binding head screw — I could move the wire.
In another outlet, I wanted to see if I could tighten a binding head screw that appeared to be tight. I got about Vio" turn on the screw.
The first week of the trial the binding head screw came up about 500 times — even [the trial judge] started to get upset with the repetition.
I went home one night, pulled about 15 outlets from their boxes and wanted to see how loose the connections were.
I could not turn any of the screws one bit. My home is wired with [aluminum], I bought the house 11 years ago in 1969. The plaintiffs talk about stress relaxation and creep which would cause the [aluminum] wire to loose its torque after a short period of time. My outlets are still tight after 11 years of use, How come these are not loose?
Another part of the trial, a list of code violations were read off as long as your arm. A witness asked if they contributed to the fire, the answer was no.
These were violations that were found in most of the unburned portions of the fire what about the violations that were destroyed by the fire no one knows about?
This is what caused the fire!! In my opinion one violation of code is too many, They should close a place down and keep it closed until the violations are corrected. When I go out on a weekend to eat or take in a show for enjoyment, I want to come home [safely].
There was a flash of fire seen in the Garden room about the same time there was smoke in the Zebra Room. Where did it start? God only knows where that fire started and how it started. Why is it that a person has to be hurt or maybe killed before a problem is corrected?
I also want to let the people who lost family & friends in the tragic fire of Beverly Hills that my prayers are with them.
one Juror on the Beverly Hills trial
. As the Advisory Committee on the Proposed Rules of Evidence noted:
The mental operations and emotional reactions of jurors in arriving at a given result would, if allowed as a subject of inquiry, place every verdict at the mercy of jurors and invite tampering and harassment. See Grenz v. Werre, 129 N.W.2d 681 (N.D.1964). The authorities are in virtually complete accord in excluding the evidence ....
. Rule 606(b) provides:
Upon an inquiry into the validity of a verdict or indictment, a juror may not testify as to any matter or statement occurring during the course of the jury’s deliberations or to the effect of anything upon his or any other jur- or’s mind or emotions as influencing him to assent to or dissent from the verdict or indictment or concerning his mental processes in connection therewith, except that a juror may testify on the question whether extraneous prejudicial information was improperly brought to the jury’s attention or whether any outside inñuence was improperly brought to bear upon any juror. Nor may his affidavit or evidence of any statement by him concerning a matter about which he would be precluded from testifying be received for these purposes.
Fed.R.Evid. 606(b) (emphasis added).
. The trial judge instructed the jury in part as follows:
You are expected in deciding this case to use your judgment and common sense. Give the evidence and the testimony of the witnesses a reasonable and fair interpretation based upon your knowledge and experience of the natural tendencies of human beings, that is what you bring with you into the courtroom, your lifetime of experience, your experiences with human beings and how they react.
Transcript at 7344.
. See also United States v. Beach, 296 F.2d 153 (4th Cir. 1961) (“the law is well settled that a case must be decided upon evidence submitted in court during the trial and not upon private experiments of the jurors”).
. Defendants point to four occasions when counsel for plaintiffs used the phrase “time-bomb” and two other occasions when witnesses for plaintiff lent support to that characterization.
. In Stiles, the court explained why a presumption of prejudice is necessary. It stated:
The foreman of the jury was asked by the trial court whether [the extraneous evidence] played any important part in the final verdict of the jury, and his reply was: “I don’t know, Judge.” Neither did the district court know, nor could it have known, whether this evidence, introduced without the knowledge of court or counsel after the retirement of the jury, influenced the verdict; and, certainly, this court, farther removed from the trial, could have no way of knowing what effect this extraneous evidence produced upon the verdict.
Stiles, supra, 211 F.2d at 190.
. Because the juror discussed his findings with other jurors, we need not decide whether an uncorroborated claim that an experiment was conducted, which potentially could be used merely as a tool to manipulate the verdict, would support a mistrial.
. Plaintiffs add that a second piece of extraneous information was improperly before the jury. Some jurors viewed a publication entitled “Reconstruction of a Tragedy,” even though the trial judge refused to admit the document into evidence on the grounds it was hearsay. A copy of the publication had been admitted inadvertently as part of a deposition to which it was appended. Plaintiffs assert their counsel simply failed to note the appendages to the exhibit, which is not surprising given the great quantity of evidence introduced at trial. Our examination of the document in question convinces us that the inadvertent introduction of the documents was not prejudicial as it was probably more damaging to the defendants than to the plaintiffs. The pamphlet indicated that the fire was electrical in nature and started in a concealed space in the Zebra Room. It further ruled out defendants’ theory that the fire started at the water pump, and it contained several explicit pictures of the tragedy. In any event, it is highly unlikely that this circumstance will recur.
Plaintiffs also claim it was error for the trial judge to have refused to ask certain voir dire questions of potential jurors. Plaintiffs did not object at the completion of voir dire; moreover, we find the trial judge did not abuse his discretion. See Eisenhauer v. Burger, 431 F.2d 833 (6th Cir. 1970).
. Plaintiffs also claim they were prejudiced by the lateness of a decision to bifurcate the proceedings. We admit some sympathy with this complaint, but we cannot say that any prejudice to the plaintiffs was not outweighed by other considerations of efficiency and convenience. In any event, it is not likely to recur.
. The trial judge described the potential advantages as follows:
This is a class action on behalf of approximately 200 persons against 23 defendants who have been grouped together as the “Aluminum Wire and Device Group.” The parties estimate that approximately 40 trial days will be required for the presentation of several hundred witnesses. Critical to plaintiffs’ case is the issue of causation. Before any determination of liability upon any of these defendants can be made, it must be established that “old technology” aluminum wire either caused or contributed to the fire at the Beverly Hills Supper Club May 28, 1977. While such establishment does not in and of itself determine liability of any or all of the defendants, a negative determination would free them from such liability. There is reason to believe that a determination of this issue would materially reduce the time required to try this case. While a determination for the defendants would as a matter of law end the proceedings, it is entirely possible that an affirmative determination might enhance the likelihood of settlement.
In Re: Beverly Hills Fire Litigation, C. No. 77-79 (E.D.Ky. Dec. 12, 1979).
. Certainly evidence regarding whether defendants breached a legal standard of care will be extensive. Moreover, because each of several defendants could insulate itself from liability by proving it was not part of any express or tacit agreement to produce defective wiring, it is anticipated that much evidence would be *217submitted regarding this issue. See note 1 supra. A defendant further may offer evidence tending to show its wiring, because of inherent metallurgical or other differences, would perform differently from that of other companies. Additionally, the damages issue remains unresolved.
. Those documents carried the Exhibit numhers:
#1182 #1928 #19,528
1291 15,310 19,542
1325 15,324 19,553
1471 15,372 19,557
15,552A 15,501 21,303
1733 15,557 25,001
1924 15,682
. gee specifically documents carrying exhibit numbers # 1920, 1926, 19,529, 1301, 1976, 3691, and 19,610. Plaintiffs also claim that some of these documents were not permitted to be submitted to the jury during deliberations. *218They do not tell us which documents were not considered, nor do they indicate which were considered with portions deleted.
. Three expert witnesses testified regarding the dangerous propensities of aluminum wire.
. We are bound by state law regarding the sufficiency of evidence. Standard Alliance Ind. v. Black Clawson Co., 587 F.2d 813, 823 (6th Cir. 1978); Chumbler v. McClure, 505 F.2d 489, 491 (6th Cir. 1974).
. See also Huffman v. SS. Mary & Elizabeth Hospital, 475 S.W.2d 631 (Ky.1972) (theory that plaintiff contracted serum hepatitis as a result of blood transfusion rejected where plaintiff unable to rule out other causes such as improperly sterilized needle or puncture wound occurring outside hospital); Briner v. General Motors Corp., 461 S.W.2d 99 (Ky.1970) (theory that defective air conditioner caused shimmying which in turn caused steering mechanism to freeze, thereby causing an accident, rejected where plaintiff unable to rule out possibilities that plaintiff struck an object or that tire was deflated, brakes malfunctioned or tie rod broke); American Insurance Co. v. Horton, 401 S.W.2d 758 (Ky.1966) (causation theory that ethylene gas used to ripen bananas caused an explosion rejected where plaintiff unable to exclude sewer gas or natural gas as causes).
. Duncan testified that the fire was able to ignite and spread without anyone learning of it because the “dead space” behind the wall offered an unobstructed fire path with fixed barriers. As further support for his hypothesis, he noted that the wails collapsed inward toward the room, indicating that structural support inside the wall had given way.
. Duncan was able to pinpoint the origin of the fire from its “burn pattern”, that is, marks indicating the path taken by the fire. He testified that the bum pattern showed the fire moved from west to east, originating approximately 48" east of the main bar room wall and 4' to 4V2' off the floor. James Donnelly, an expert witness employed at Systems Engineering Associates, concurred in this assessment.
. William White, the electrician who wired the supper club, testified that he wired the duplex receptacle in the cubbyhole with aluminum branch circuit wiring. Duncan opined that the wiring was aluminum because such wire was found in a panel box furnishing electricity to the Zebra Room. A report of the Consumer Product Safety Commission disclosed the same information. Thomas McClorey, an expert qualified in architecture and civil engineering, testified he found aluminum wire in the panel box. Dr. Aronstein ran tests which indicated that the club was wired with old technology aluminum wire.
Dr. Aronstein reasoned that one could infer from the fact that no wire was found on certain devices that it was made of aluminum. Aluminum, but not copper, wire would have melted at temperatures reached in the fire. Duncan testified that aluminum wire melts at approximately 1100° F, while copper melts at approximately 1950° F; further, Duncan stated that the Beverly Hills fire reached temperatures between 1500° and 1200° F.
. McClorey testified that a fire in the north wall would cause a stratification of the air in the heating ducts at the top of the ceiling. He stated that this stratification would cause one of the air vents in the Zebra Room to put out cold air, the air which was at the bottom of the air duct. The second vent would put out hot air because that would have been the air which was heated in the air duct as it passed the point of the fire’s origin, and that heating process would have caused the heated air to rise above the cooler air-conditioned air. Michael Sims, a guest at a wedding reception in the Zebra Room, gave testimony which could support this theory. He claimed he heard rumbling noises above and below the room, noticed that the room kept getting warmer on the side where the wedding party was located, but that the room was very cool on the other side. In fact, he stated, the room was so cool on the other side that women were putting on sweaters. Eventually, he also began to feel vibrations, and he saw the wedding cake begin to melt.
Both McClorey and Eileen Druckman testified they smelled strange odors. Druckman smelled something like candles that had been blown out; Sims smelled an odor like boiling beef or battery acid. The jury could have found those perceptions similar to plaintiffs’ argument that overheating produces an acrid smell.
. Patty Kolbinsky, a club employee, testified that a light and adding machine, attached to the cubbyhole, continued to work. Duncan explained this was consistent with symptoms of overheating. He testified that when heat buildup occurs, the insulation material vaporizes *221and there is deterioration of the thermal plastic component parts of the receptacle itself. Heat from this area is obvious, he stated, and there is charring at the point where the wire connects with the receptacle. He suggested that the receptacle is still operable at this point because the circuit has not been interrupted. Dr. Aronstein agreed the circuits can work while overheating.
. Duncan testified that any malfunction in an alcove device would have resulted in rapid flame propagation within the room. He concluded that the fire would have been evident to others at an earlier point had it not occurred within the concealed space.
. A short circuit, Duncan opined, would cause a different burn pattern. He also testified that a profile of the fire showed that there had been long-term heating deterioration of the fire area. Overheating, he suggested, is sustained over a long period of time. A short-circuit is an instantaneous release of energy which causes an immediate elimination of service rather than allowing a buildup of heat.
. Duncan testified that if the pump cord had started the fire, the burn pattern would have been different and the cord itself would have been totally consumed by the fire. He stated further that such origin could not have provided the consistent heating of materials that occurred in the area of origin.
. Duncan was asked on redirect examination about an earlier reference to the switch, and he stated that different bum patterns would have resulted from fire originating at the switch.
. Duncan testified that evidence of any further service in the cubbyhole was unapparent to him.
. The Rollins court was primarily concerned with whether plaintiff offered evidence sufficient to submit her case to the jury using a res ipsa loquitor theory as a basis for liability. Because liability is not yet at issue in the present case, that discussion in Rollins is not relevant here.
. For example, in LeSage v. Pitts, 311 Ky. 155, 223 S.W.2d 347 (1949), a loose wall fell on plaintiff. There was no direct evidence that defendant left the wall blocks in an unfixed position. The court rejected plaintiff’s characterization of events, stating:
Appellant, from the fact that mortar was attached to the block which rolled with him, infers that the block was anchored safely in its place in the original erection of the wall. He then theorizes, without even a scintilla of evidence to support the theory, that, in constructing the wall, appellee forgot, or at least failed, to leave openings in which to place the steel structure. This theory calls for an inference to be drawn upon the first inference. He further theorizes that, having forgotten to leave the openings in the first instance, appellee cut the openings after having erected the wall in solid formation. This is an inference drawn upon the second inference. Finally, he theorizes that, in cutting the openings in the solid wall, the block causing the injury was loosened and left in an unsafe condition in its original position. This is an inference based upon the third inference.
Id. at 223 S.W.2d at 352.
In Briner, supra, plaintiff was unable to show her car had defective brakes at the time defendant serviced the car; thus, any further proof that investigation should have occurred was an unconnected inference built on an earlier inference. When the Kentucky court speaks of “inferences,” therefore, it refers not to conclusions drawn inductively from circumstantial evi*222dence, but rather to conclusions drawn from a series of statements unsupported by any independent evidence. Miller v. Watts, 436 S.W.2d 515, 517 (Ky.1969) (“The kind of speculation that is not allowable occurs when the probabilities of an event’s having happened in one or two or more ways are equal and there is no evidence as to which way it happened.”) (emphasis in original); see also Klingenfus v. Dunaway, 402 S.W.2d 844, 846 (Ky.1966).
. The trial judge instructed the jury as follows:
In order for the plaintiffs to prevail, they must establish by a preponderance of the evidence that the fire was caused by the connection of old technology aluminum wire to electrical devices. This is known as causation.
The connection of old technology aluminum wire to electrical devices is a cause of the fire at the Beverly Hills Supper Club if it was a substantial factor in bringing about such fire.
Transcript at 7350 (emphasis added).
. There were numerous complaints filed by several plaintiffs against several defendants following June 2, 1977. Because those complaints were filed later, those suits similarly would be barred.
. The judge determined that the statute should be read narrowly in order to avoid the constitutional question raised in Saylor v. Hall, 497 S.W.2d 218 (Ky.1973). A narrow reading of the statute to exclude materialmen may raise another constitutional problem. As the judge noted, section 413.135 creates immunities. Some state courts have found that creation of statutory immunities in favor of contractors while excluding materialmen from similar protection violates state equal protection guarantees. See Pacific Indemnity Co. v. Thompson-Yeager, Inc., 260 N.W.2d 548 (Minn.1977). But see Carter v. Hartenstein, 248 Ark. 1172, 455 S.W.2d 918 (1970). This dilemma makes the argument for narrow construction less compelling.
. As used in the statute, “person” means “an individual, corporation, partnership, business trust, unincorporated association or joint stock company ....” Ky.Rev.Stat.Ann. §413.-135(5) (Baldwin).
. Kentucky rejected the defense of privity of contract and the general rule of manufacturer non-liability as early as 1956 in C. D. Herme, Inc. v. R. C. Tway Co., 294 S.W.2d 534 (Ky. 1956). See also Allen v. Coca-Cola Bottling Company, 403 S.W.2d 20 (Ky.1966).
. The Kentucky legislature evidenced further its intention to protect defendants against what it viewed as unlimited liability when it promulgated its Product Liability Act in 1978. Ky. Rev.Stat.Ann. § 411.300 et seq. (Baldwin). The purpose of the Act is indicated in the preamble to the Senate Bill, which provides:
WHEREAS, in recent years there has been an increasing public awareness of a need for safer products for the consumer with a resulting increase in the size and number of product liability claims, and
WHEREAS, this dramatic increase of product liability claims is causing an increase in the cost to the consumer of purchasing products, and an increase in the cost of manufacturing, wholesaling, retailing, and insuring said products, and
WHEREAS, a portion of said increased costs is caused by the absence of clear legal precedents and guidelines which govern the rights and liabilities of consumers, manufacturers, wholesalers, retailers, and insurers in the litigation of product liability claims, and
WHEREAS, it is in the interest of the public, consumers, manufacturers, wholesalers, retailers, and insurers, for the General Assembly to codify certain existing legal precedents and to establish certain guidelines which shall govern the rights of all participants in product liability litigation ....
Senate Bill No. 119 (June 17, 1978). See also Ky.Rev.Stat.Ann. § 413.120(14) (precluding action against “builders of a home or other improvements” more than five years after improvement).
It should be noted that we are not persuaded that section 411.300 et seq. is contradictory to section 413.135 as plaintiffs assert, their argument being that the former section is superfluous if the latter applies to materialmen. Section 413.135 applies only to constructions of real estate; it does not apply to manufacturers of personal property.
. Considering a similar statute, the New Jersey court stated:
As best we can perceive, the intent of the language of the statute was to protect those who contribute to the design, planning, supervision or construction of a structural improvement to real estate and those systems, ordinarily mechanical systems, such as heating, electrical, plumbing and air conditioning, which are integrally a normal part of that kind of improvement, and which are required for the structure to actually function as intended.
Brown v. Jersey Central Power & Light, 163 N.J.Super. 179, 394 A.2d 397, 405 (1978).
In this connection, it should be noted that whether a design is an “integral” part of a structure is not relevant to whether a certain class is protected as plaintiffs suggest; rather, it concerns whether the design is an “improvement” to realty. It would be difficult to argue that an electrical system is not an “improvement,” which is defined as “a permanent and necessary part of the building.” Menne v. American Radiator Co., 150 Ky. 151, 150 S.W. 24, 25 (1912).
. The Kentucky court reasoned:
Our state Constitution, however, has been held to prohibit the legislative branch from abolishing common-law rights of action for injuries to the person caused by negligence or for death caused by negligence.. ..
Section 14 of the Constitution of Kentucky states:
“All courts shall be open and every person, for an injury done him in his lands, goods, person or reputation, shall have remedy by due course of law, and right and justice administered without sale, denial or delay.”
This section was held to apply to the legislative branch of government as well as to the judicial in Commonwealth v. Werner, Ky., 280 S.W.2d 214 (1955). Section 54 of the same Constitution states:
“The General Assembly shall have no power to limit the amount to be recovered for injuries resulting in death or for injuries to person or property.” The Kentucky Constitution in Section 241 states:
“Whenever the death of a person shall result from an injury inflicted by negligence or wrongful act, then, in every such case, damages may be recovered for such death from the corporations and persons so causing the same....
This court construed section 241 in 1911 to mean that "... it is not within the power of the legislature to deny this right of action. The section is as comprehensive as language can make it The words ‘negligence’ and ‘wrongful act’ are sufficiently broad to embrace every degree of tort that can be committed against the person....” Britton’s Adm’r v. Samuels, 143 Ky. 129, 136 S.W. 143.
Saylor v. Hall, 497 S.W.2d 218, 222 (Ky.1973) (emphasis in original).
Kentucky has applied these constitutional requirements in other contexts. See, e.g., Ludwig v. Johnson, et al., 243 Ky. 533, 534, 49 S.W.2d 347 (1932) (guest statute).
. A decision of an intermediate appellate court in Kentucky applying a medical malpractice statute of limitations without discussion of the potential constitutional infirmities outlined by the Kentucky Supreme Court (formerly the Kentucky Court of Appeals) does not convince us that the vitality of Saylor is in doubt, as defendants suggest. See Ferguson v. Cunningham, 556 S.W.2d 164 (Ky.App.1977). We note that the Kentucky Supreme Court more recently has reiterated the proposition in Saylor that a “cause of action does not exist until the conduct causes injury that produces loss or damage.” Louisville Trust Co. v. JohnsManville Products, 580 S.W.2d 497 (Ky.1979) (Reed, J.), quoting Saylor v. Hall, 497 S.W.2d 218, 225 (Ky.1973) (applying “discovery rule” to actions for injury from latent disease).