The plaintiffs, Charles Nailor and James Nailor, appeal from a jury verdict in favor of the defendants, International Harvester Company (IH) and its insurer. The Nailors had sued IH under products liability for injuries suffered by Charles Nailor when an IH tractor, on which he was riding, broke in half while being towed.
The facts are as follows: Charles Nailor’s father, James Nailor, had purchased the 31-year-old International Harvester AV tractor about one year before the accident here occurred. On the morning of the accident, October 16, 1977, Charles Nailor was having trouble getting the tractor started. He changed the plugs but found the battery had run down from his previous attempts to start it. A friend who was helping him, Junius Dumas, suggested they tow the tractor to get it started. Accordingly, they hitched it up to James Nailor’s pickup truck, hooking the tow chain on the tractor’s front axle.
They pulled out of the Nailor yard onto the highway, stopping at an intersection and then moving forward again when they saw the way was clear. Charles had the tractor in fourth gear, having been told this was the best gear for towing. He had the clutch pedal in while they were at the intersection; as they pulled away, he let up on the clutch pedal when they picked up speed, moving about 7 to 10 miles per hour. Suddenly, about 85 to 90 feet from the intersection, the tractor broke in half. Charles Nailor fell off, and the rear part of the tractor flipped over onto him. He suffered an open comminuted fracture of his left arm, requiring extensive surgery, as well as two broken teeth, which required a fixed bridge replacement.
It was later determined the tractor fractured and broke at the clutch housing. Charles Nailor testified that he had never towed the tractor before, that he kept the tractor clean and in good running condition, and that he had never noticed any cracks on the chassis.
James Nailor, Charles’ father, testified he had purchased the tractor about a year before the accident, and the dealer told him nothing about the tractor’s history. He said he did not receive an owner’s manual with the tractor, but he had read other tractor manuals and felt most of them were probably the same. As far as he knew, the only mechanical work either he or his son ever did on the tractor was to change the plugs. He himself had towed the tractor once before to start it.
At trial, neither side established the precise cause of the clutch housing’s fracture. Both sides offered extensive expert testimo*786ny that conflicted in several important respects.
Plaintiffs’ metallurgical expert, Dr. Courtney Busch, testified he examined the fractured clutch housing in July 1978, conducting visual and metallurgical examinations and some mechanical testing. He determined that the fracture initiated in a notch in the clutch housing and that the fracture occurred all at once as a result of an overload. He found no indication of fatigue or a pre-existing crack. He stated the clutch housing was made of cast iron, an inherently brittle metal.
Dr. Busch also conducted Brinell hardness tests on cross-sections of the clutch housing. These tests are performed by pressing a hardened 10-millimeter ball into the surface of the metal under a weight load. The diameter of the indentation is then measured, and a hardness reading obtained by referring to the Brinell tables. Dr. Busch found a range of Brinell hardness numbers from 148 to 173 in his sampling of the clutch housing, with an average of 163. Referring to the American Society of Testing Materials (ASTM) standards, he determined the housing material should have a minimum tensile strength of 18,000 pounds per square inch (PSI).
He testified the part failed primarily due to a bending load, consistent with the type of failure break to be expected when a tractor is towed. When the clutch is let out while the tractor is moving, it develops a force or skiing action of the back wheels that has a tendency to bend the tractor. These parts break mostly due to bending, he said; it is very difficult to break them with tension. He stated the fact that the crack began at the bottom indicates the bottom was in tension while the top of the housing was in compression.
Dr. Busch further testified that the IH specifications called for the clutch housing to be made of gray cast iron Number 20 with Brinell hardness values ranging from 166 to 212, and a tensile strength of 25,000 PSI. Based on his testing, he concluded the clutch housing here did not meet minimum specifications because the Brinell hardness numbers fall out of the specified range on the low side, meaning the tensile strength would be somewhat less than the minimum value stated in the IH manual. He said he had done a microstructure test, but did no chemical analysis because he did not feel that was important.
Dr. Busch concluded this metal broke where it did because of the notch built into it, which caused the stresses to concentrate at the root of the notch. He said the notch may have been put there for a reason, but it considerably weakens the structure. He admitted he had discovered another crack in thé housing, on the right side across the flanged area. In his opinion, however, this crack occurred after the main fracture. He said if this had been a pre-existing crack, the housing would have broken through it first. In his view, the age of the metal had no connection with the cracking.
Under cross-examination, Dr. Busch testified he had examined only the clutch housing and the area of the fracture. He found the tractor lying in a field with dirt and grass on it. The clutch housing had been exposed to the elements, and there was rust around the whole fracture surface. He denied that the rusted surface made it difficult to determine the type of fracture. He said the fracture surface is flat, sharp and brittle, with no plastic deformation.
If there had been a pre-existing crack, he said, the pre-existing crack would have been rusting from the time it opened, and the rust on it would have been magnetite rather than ferrous oxide. He admitted, however, that any rust could have been rubbed off when the clutch housing dragged on the ground after the accident. He also acknowledged he could have overlooked a pre-existing crack as small as Vw inch long, but felt such a small crack would not make much difference.
He further admitted that if there had been a crack in the area that was rubbed, it would have affected the stress in the clutch housing if it was at the point of the fracture’s tip. He conceded such a crack would have affected it almost as much as the notch built into the clutch housing.
*787On cross-examination of his testing methods, he said the ASTM manual lists the procedure for testing gray iron hardness as a 10-millimeter ball under 3,000 kilograms of pressure. He stated he had reduced the pressure to 1,500 kilograms, however, because the pieces cracked under the higher pressure. He said he retained the same size ball because it is in the specifications. The ASTM manual he used was the 1975 edition, with standards for this specification approved in 1972. He admitted he did not know if these ASTM standards were in effect in 1946 when this tractor was manufactured. He also admitted he had never run any tests on an AV tractor to determine the stress placed on the clutch housing.
The plaintiffs’ other expert witness, Dr. Andrew McPhate, was accepted by the court as an expert in the fields of mechanical engineering and machine design. He testified he determined the cause of the fracture was an overload; the housing was asked to support a load it was not capable of supporting, and it failed at the point where the loading in combination with the strength of the section was worst.
He attributed the failure to the placement of the hole (which Busch called the notch) in the housing. The levels of stress distribution around the edges of the hole are approximately twice what they would be in that section of the housing if the hole were not there. In his opinion, by placing the hole in that section, the designers of the machine interrupted the smooth flow from one section to another. He said the purpose of the hole was to carry a shaft that is connected to the clutch pedal.
Dr. McPhate said the crack traveled up both sides of the housing, indicating the bottom of the clutch housing was in tension at the time of the failure. He stated this would be consistent with where the tension would be located in a clutch housing of a tractor in tow. He said the part could have been made stronger either by using a higher grade material, or by making it larger and its walls thicker, or by redesigning it so the hole would not be next to the transition section.
On cross-examination, Dr. McPhate testified he never saw the tractor itself, only the clutch housing. He first examined the housing in January 1981, at which time, he said, the fracture face was obviously old. He admitted he would have preferred to have seen the fractures on the day of the failure. He also admitted he had not done any studies of towing situations, and he did not know what forces would be applied to the clutch housing when the tractor is towed.
He acknowledged it was possible the clutch housing could have been damaged in the tractor’s 30-year history; it could have run over fence posts, scraped the sides of barns, been run into ditches. He conceded there could have been a pre-existing crack in the housing, “ * * * just like a mousetrap ready to go off at any time.”
He testified further that he was aware this tractor was designed in the 1930s and that the analytical tools available for design back then were not the same as today’s. He said that modern farm equipment has a lot more welding done on it in fabrication than was done in 1946. He stated that welding as an art has progressed much more than casting. Casting is an old technique, the earliest technique for fabrication. In his opinion, this clutch housing and the basic design were the state of the art back in the 1930s, but there were better grades of cast iron available than evidenced in this one.
For the defense, Mr. Gordon Walter, an International Harvester employee, testified as an expert in metallurgical engineering. His visual examination of the clutch housing revealed that the fracture was two separate cracks which started on the bottom, where the hole is located, and progressed around to the top. He said he could not determine the type of fracture because of the amount of rust, which had destroyed most of the microscopic features. He could not tell from looking at it if there was a pre-existing crack or a fatigue failure. He found another crack on the flange on the top side, but did not think the second crack *788had any effect on the crack that occurred at the notch.
He stated they also noticed that the part was bolted on and severely loaded, and had been pulled away from the casting. He said this indicates some kind of loading, other than that caused by towing. In a towing mode, the top side would be in compression. There would have had to be some previous loads in tension to cause the crack on the topside. He could not say what caused the part to be pulled away with so much deformation, except that “there was some kind of other event that did that.”
Dr. Walter also criticized the results of Dr. Busch’s Brinell hardness tests. He said the cross-sections Busch tested were too thin to give a proper reading under a 3,000 kilogram load. He said using the 1,500-kilo-gram load, as Busch did, gives a lower reading. The ASTM book on hardness testing states you will get different values when you use different loads for the same size ball in Brinell tests. In his opinion, considering the experimental error involved in reading Brinells, the 163 average obtained by Busch meets the 166 required by IH’s specifications. Using Busch’s Brinell hard-nesses, Dr. Walter estimated the strength of the casting to be 23,000 to 24,000 PSI.
(The trial judge properly prohibited Dr. Walter from testifying as to the results of IH’s own Brinell hardness tests and chemical analysis, for failure to lay a proper foundation. Dr. Walter admitted he had not personally conducted the tests, and the individuals who had were not present to testify. Although IH presented this evidence by proffer, we do not consider it on appeal as it would be hearsay.)
According to Dr. Walter, this was a suitable grade of cast iron for a clutch housing. He said it was the state of the art as far as metallurgical design for this tractor’s era.
Ms. Patricia Crowley, also an IH employee, qualified as an expert in mechanical engineering for the defense. At the time of trial, she was working as a product integrity coordinator for IH, investigating accidents to determine how IH products perform. She testified she had conducted a visual examination of the clutch housing, but there was so much rust on it she could not tell much from a visual inspection.
She did observe that there were two different cracks which met up at about a two o’clock position on the fracture surface. She also noted that the steering support housing was pulled away from the rest of the housing. She said this would require severe loading, actually overloading, of the part and the way it was connected in order to deform it as it was.
Ms. Crowley also noticed a crack on the flange, on the other side from where the fracture began. This indicated to her that the second crack was caused by a different incident. She reached this conclusion because the top of the housing would have been in compression during the towing operation, and such a crack could not have been created during compression. From this, she concluded this crack was in existence before the Nailor accident. She stated that in order to have such a crack, there would have had to be some kind of accident, maybe a rollover or running into a stump or a building as Mr. McPhate had suggested.
She testified the clutch housing was designed to withstand the loads associated with using different farm implements, whether forward-mounted, middle-mounted or rear-mounted. Some of these implements would put the bottom of the housing in more tension than a towing operation would. In her opinion, a clutch housing of this design will not break when it is being towed unless there is some sort of pre-exist-ing damage. The part was designed to withstand the stresses of using different kinds of implements, and was designed, tested and analyzed to make sure it would withstand such uses. Towing would not put the bottom part in as much tension as using a forward-mounted implement would.
Ms. Crowley also computed the force required to cause an AV clutch housing without any pre-existing damage to fail. To do so, she used Dr. Busch’s figure of 18,000 PSI tensile strength because that is a very conservative figure. Figuring in the geom*789etry or configuration of the clutch housing, she calculated it would require in excess of 9,000 pounds, or over 4 tons, of pulling force on the tractor to break the clutch housing in this way. Since the weight of the AV tractor is approximately 2,900 pounds, she said, the maximum force required to pull it would be no more than 2,900 pounds, unless the tractor was tied down to a tree or a building.
Since the testimony was that the tractor was not tied down, that it was free to move and was moving, this indicated to Ms. Crowley that there is no way a clutch housing without some kind of pre-existing damage could be pulled apart. She said the type of damage which could weaken the housing could start on the inside, where it would not necessarily be visible from the outside. In her opinion, there had to be some kind of pre-existing crack or other damage that weakened the housing to the extent that the amount of force that you could get pulling it with a pickup truck was enough to break it.
Ms. Crowley opined that the design of the clutch housing was and still is good. She said the engineers who made it would have tested it extensively and would have made it stronger if they had a problem. Since the part had been in use for many years, this indicated to her that it was a satisfactory design.
She stated she had checked the available records of IH on the AV tractor, and had consulted with IH personnel who had been involved over the years with the AV tractor. She could find no indication that there was a major problem with clutch housings breaking. In her opinion, if there was a manufacturing or design problem with this clutch housing, it would have shown up long before 31 years.
Ms. Crowley testified the hole in the clutch housing had been put there to allow for adjustment of the amount of free travel on the clutch pedal, so the tractor would be simpler to use and maintain. She stated the hole was a perfectly good design because there is more material around the hole, strengthening it to withstand the kinds of stresses that will be put on it.
She further testified the part was designed to withstand things much more difficult and harder on the tractor than being towed. There were other means of starting the tractor and not as many means as today of towing the tractor, so it would not be an everyday experience to tow it. Accordingly, she did not feel IH should have warned people not to tow-start the AV tractor in 1946.
Based on what she had looked at and what she knew of the records, Ms. Crowley gave it as her opinion there was no defect in the design or manufacture of the clutch housing when it left IH’s control in 1946.
Under cross-examination, Ms. Crowley agreed that the manufacturer and designer of the AV tractor should have been able to foresee, in 1946, that the tractor might be towed in an attempt to start it. She admitted the AV tractor owner’s manual contained no specific references to towing. She acknowledged that increasing the strength of the cast iron would make the housing stronger and more able to withstand tension and compression.
Plaintiffs’ counsel also questioned Ms. Crowley at length about the formula she used to arrive at the 9,000 pounds of force required to break the clutch housing. Ms. Crowley explained her calculations and also how she arrived at her conclusion that there was at least one pre-existing crack in the clutch housing. She stated that the area in which this second crack was located was an area that would be in compression during towing and thus would not crack at the same time. She also testified that the second crack could not have occurred after the accident, because that part of the tractor would not have struck the ground following the accident. She said the drive shaft would have taken most of the force of impact with the ground. Even if the drive shaft hit the inside of the housing, it would not have caused the second crack because the area would have been in compression. Further, if there was going to be a crack, the crack would form where the drive shaft actually hit, not farther away.
*790Ms. Crowley also stated the second crack would not have a direct relationship with the crack that caused the housing to fail. She said, however, the fact there is a crack indicated to her that the tractor had “gone through something.” She admitted, on the other hand, that she had no way of knowing what the tractor had been through in its lifetime, and that it could have spent a great deal of time just sitting in a shed.
Elliott Bourgeois also testified for the defense. He had been employed by IH for more than 20 years in numerous positions dealing with farm equipment, in which he had extensive contact with sales representatives, service departments, dealers and customers of IH. He testified he had never heard of any other failure of a clutch housing on an AV tractor being towed.
Dr. Courtney Busch was recalled to the stand as a rebuttal witness. He testified he felt the second crack occurred after the accident because he concluded the drive shaft pushed against the side of the housing after the fracture, and the housing reacted by trying to hold down the cracked part. Because the fracture surface had been pushed over, he concluded the crack occurred after the fracture.
Dr. Busch also criticized Ms. Crowley’s calculations of the force required to break the housing. He testified her calculations omitted the force that would be created by putting the tractor in gear during towing.
He also defended the validity of his Bri-nell hardness tests, stating that it made no difference that he had reduced the pressure load from the specified 3,000 kilograms to 1,500 kilograms. Dr. Busch said the values are calculated from a formula and he had used standard practice.
On appeal the plaintiffs contend the tractor had a three-fold defect: (1) that the clutch housing was defectively designed because of the placement of the hole at the bottom of the housing, where the fracture -began; (2) that the clutch housing was defectively manufactured because the cast iron had a tensile strength below the manufacturer’s specifications; and (3) that the manufacturer was negligent in failing to warn users of the potential danger involved in towing the tractor.
Under Louisiana law, a plaintiff suing a manufacturer for an injury resulting from an alleged defect in a product has the burden of proving that the product was defective, that is, unreasonably dangerous to normal use, and that the plaintiff’s injuries were caused by reason of the defect. Weber v. Fidelity & Casualty Insurance Co. of N.Y., 259 La. 599, 250 So.2d 754 (La. 1971); Thornhill v. Black, Sivalls & Bryson, Inc., 394 So.2d 1189 (La.1981).
The issue of defective design is factual and depends on the circumstances of each case. Cobb v. Insured Lloyds, 387 So.2d 13 (La.App. 3d Cir.1980).
A plaintiff must prove his case by a clear preponderance of the evidence, which means the plaintiff need only show it is more probable than not that defendant’s negligence was the cause of the harm, damage or injury inflicted. Causation may be proved by circumstantial evidence which need not necessarily exclude all other possible causes. Mere proof that something is possible is of little probative value as to an ultimate issue of fact, unless it be established with reasonable certainty that all other alternatives are impossible. Where several possible causes in fact are established with equal probability, but no one is shown by that preponderance which the law demands, the plaintiff has not discharged the burden of proof. Lutheran Church of Good Shepherd v. Canfield, 233 So.2d 331 (La.App. 1st Cir.1970), writ denied 236 So.2d 497.
Where the testimony of expert witnesses differs, it is largely a matter of fact for the jury to determine the most credible evidence, and a finding of fact in this regard will not be overturned unless manifest error appears in the record. A jury verdict should be maintained unless the record reflects that its conclusions of fact are not supported by the evidence, or its application of the law is clearly erroneous. St. Pierre v. Gabel, 351 So.2d 821 (La.App. 1st Cir.1977).
*791In a products liability case, whether the product is defective at the time of sale is a question of fact and determination of that fact by the trier will not be disturbed on appeal in the absence of manifest error. McClinton v. Reid, 417 So.2d 128 (La.App. 3d Cir.1982).
Where there is a conflict in the testimony, reasonable evaluations of credibility and reasonable inferences of fact should not be disturbed on review, even though the appellate court may feel that its own evaluations and inferences are as reasonable. Dieudonne v. Guidry, 336 So.2d 990 (La.App. 3d Cir.1976), writ denied 339 So.2d 853.
In Ashley v. Nissan Motor Corp. in U.S.A., 321 So.2d 868 (La.App. 1st Cir.1975), writ denied 323 So.2d 478, where there was conflicting evidence as to whether there was any defect in the right rear wheel cylinder of an automobile involved in an accident, the court found the jury verdict absolving the manufacturer from any liability was not manifestly erroneous.
Applying these principles to the case before us, we find this appeal turns on the testimony of the expert witnesses. Plaintiffs’ experts attempted to prove the cast iron of the clutch housing was not strong enough and that the hole in the housing made it weaker. Defendants’ experts attempted to rebut this by casting doubt on the hardness tests of the plaintiffs’ witness Dr. Busch and by proving the design took into consideration the additional stresses created by the presence of the hole in the housing. Defendants’ witnesses further tried to prove the clutch housing had preexisting cracks indicating some prior event extreme enough to damage and thus weaken the clutch housing.
It is obvious, from the verdict in defendants’ favor, the jury found the defense witnesses more credible than the plaintiffs’ witnesses. We can find no manifest error in their conclusion. Our examination of the record does not establish that the plaintiffs proved by a preponderance of the evidence that the clutch housing was either defectively designed or defectively manufactured, or, further, that any such defect was the proximate cause of the fracture which broke the housing apart.
Since the evidence does not establish that the jury was clearly wrong, the verdict for defendants must be affirmed.
With respect to procedural questions raised by plaintiffs, they cannot complain now of the trial judge’s failure to give the jury special interrogatories to consider, because they failed to object at trial when the court omitted to give them.
Further, in light of our findings here, it is unnecessary to discuss the trial court’s refusal to grant a new trial.
Finally, the plaintiffs’ complaint that the jury was “too confused” to return a proper verdict is totally without merit. Confusion of the jury is one of the hazards one must foresee when trying a case to a jury; it is the obligation of the parties—or, more properly,. their attorneys—to present their side of the evidence in such a way that the jury will understand it.
In connection with this last complaint, plaintiffs have appended to their brief the affidavits of two jurors relative to the jury’s deliberations. Defendants filed a motion to strike this portion of plaintiffs’ brief; we referred the motion to the merits.
We hereby grant the motion to strike. Documents not offered in evidence in the trial court may not be considered by the Court of Appeal. Vinson v. Levy, 372 So.2d 694 (La.App. 1st Cir.1979), writ denied 375 So.2d 942; see also Jackson v. Gordon, 381 So.2d 520 (La.App. 1st Cir. 1980); Leger v. Delano Plantation, Inc., 350 So.2d 377 (La.App. 3d Cir.1977). Further, except for polling and certain types of extreme irregularities such as tampering, the law does not allow any inquiry into the method or thought process by which a jury reaches its verdict. Morris v. Guidry, 335 So.2d 75 (La.App. 3d Cir.1976), writ denied 338 So.2d 702.
For the foregoing reasons, therefore, the judgment of the district court is affirmed. *792Costs of this appeal are assessed against appellants.
AFFIRMED.