OPINION OF THE COURT

WEIS, Circuit Judge.

In this Safety Appliance case we conclude that the district court properly charged that, if its equipment is not defective, a railroad may invoke the defense that railcars did not couple automatically because the couplers were out of alignment. We also determine that although plaintiff's counsel's closing argument went beyond the bounds of legitimate advocacy, that dereliction does not permit defense counsel to respond by referring to objectionable evidence not in the record. Because of the effect of those comments, we will reverse the judgment in favor of defendant and remand for a new trial.

Plaintiff, Charles Reed, was a freight conductor who was injured on May 7, 1986 in the course of his employment with defendant Philadelphia, Bethlehem New England Railroad. He brought suit under the Federal Safety Appliance Act and the Federal Employers' Liability Act, but withdrew the latter count shortly before trial.

On the day of the accident, Reed and his crew were assigned the task of weighing the train cars. To accomplish this, the workers were to "cut," or uncouple, each car and allow it to pass across an automatic scale on the tracks. The cars were then to reconnect at a point beyond the scale.

Reed uncoupled a boxcar from the rest of the train. He climbed aboard, and rode it down beyond the scale. After he signaled, the crew uncoupled the next car, a scrapcar. The scrapcar passed over the scale, and contacted the boxcar but failed to couple with it on impact.

Reed then went between the two cars and found the knuckles on both couplers closed. He first opened the knuckle on the boxcar, using a lever located on the outer-edge of the car. He attempted the same maneuver on the scrap car, but testified that the lever did not function. He stepped between the two cars to open the knuckle. As he was so engaged, the scrap car lurched forward, amputating part of his foot.

The case was tried to a jury. During closing arguments, plaintiff's counsel made emotional and misleading comments. To counter this, the trial judge permitted defense counsel to tell the jury that plaintiff received disability benefits.

The trial judge submitted the case to the jury on the plaintiff's theory that the railroad was liable on proof that the cars failed to couple automatically on impact and that the cutting lever on the scrap car was defective. The court also instructed the jury to consider the railroad's defense that the cars failed to couple because the couplers were out of alignment and the cutting lever was, in fact, operable.

In response to an interrogatory, the jury found that the railroad had not violated the Safety Appliance Act. Accordingly, the court entered judgment for the railroad. Plaintiff's post-trial motions were denied, and this appeal followed.

This appeal presents two unrelated issues: the application of the Safety Appliance Act and defense counsel's comments to the jury in the closing argument.

I. [11] FEDERAL SAFETY APPLIANCE ACT [12] A. THE EQUIPMENT

Individual railroad cars are attached by couplers, which are devices located at both ends of all train cars providing a means for connecting one car to another. The standard coupler consists of a knuckle attached to a drawbar. In simplistic terms, the drawbar is connected to the car, the knuckle is connected to the other end of the drawbar, and on coupling the knuckle connects to the knuckle of another car.

The knuckle may be loosely described as a sort of clamp with moveable jaws that may be opened and closed as required. When the knuckle of one car is open and is brought in forcible contact with the knuckle of another car, the knuckles will close automatically and lock into place, making a firm connection between the two cars. The jaws of the knuckle are opened by the use of a lever at the side of the car. When the equipment is functioning properly the crew need not step between the ends of the cars to open the knuckles.

The drawbar can be thought of as a pivot moving in a horizontal plane. It has dual functions — first to anchor the knuckle to the car and, second, to move somewhat from side to side so that the knuckle can move laterally as the railroad car travels around a curve. If the drawbar were rigid and had no lateral movement, cars would be derailed as they move through curves. Although drawbars must have some play, too much lateral movement may result in misalignment and failure of the coupling operation.

To function automatically, couplers must be in a "mating" position. That is, each drawbar must be aligned with its mate and one of the knuckles must be opened. If not aligned properly, the knuckle of one car will not contact the other and the cars will not couple.

Drawbars are aligned manually, but they sometimes move out of position because of the vibration and movement of the railroad car. Therefore even if a crewman corrects the alignment before a car begins its movement toward the point where coupling is to occur, it is possible that the car's motion may jar the drawbar out of position. Aligning the drawbar requires the crewman to go between the cars.

Failure to couple because the couplers are not in the proper position may be described as misalignment of the drawbar. The legal significance of misaligned drawbars is the issue on which we focus.

B. THE LAW

In the early days of railroading, crewmen suffered an alarming number of injuries when forced to step between cars to operate the coupling devices then in use. Taking notice of the harm caused by defective and inadequate equipment, Congress in 1893 enacted the Safety Appliance Act. Section 2 of that legislation requires railroads to use automatic coupling devices. In its present form, the Act prohibits railroads from using cars "not equipped with couplers coupling automatically by impact, and which can be uncoupled without the necessity of men going between the ends of cars." 45 U.S.C. § 2.

In response to the Act, railroads adopted automatic couplers whose knuckles are opened by levers at the outside edge of the cars. No device, however, has yet been put into wide-spread use that would automatically align the drawbar.

In construing the Safety Appliance Act, the Supreme Court has imposed absolute liability for "equipment defects," making it clear that the failure of coupling equipment to perform "as required" by the statute was "in itself an actionable wrong, in no way dependent upon negligence and for the proximate results of which there is liability — a liability that cannot be escaped by proof of care or diligence." O'Donnell v. Elgin, J. E. Ry. Co., 338 U.S. 384, 390, 70 S.Ct. 200, 204, 94 L.Ed. 187 (1949); see also United Transportation Union v. Lewis, 711 F.2d 233, 248-51 (D.C.Cir. 1983) ("predicate for liability under section 2 is the failure to provide equipment that functions as the statute commands").

The Supreme Court has emphasized that "the duty under the Acts is not based on the negligence of the carrier but is an absolute one requiring performance `on the occasion in question.'" Affolder v. New York, C. S.L. R.R. Co., 339 U.S. 96, 98, 70 S.Ct. 509, 510, 94 L.Ed. 683 (1950). Because equipment failure itself suffices to fasten liability upon the railroad, an injured railworker need not show a defect in the coupler to recover for his injuries.

Nevertheless, liability for equipment failure under the Safety Act does not preclude some defenses arising out of operation of the equipment. The Affolder Court pointedly noted that, "Of course this assumes that the coupler was placed in a position to operate on impact. Thus, if `the failure of these two cars to couple on impact was because the coupler [knuckle] on the Pennsylvania car had not been properly opened,' the railroad had a good defense." Id. at 99, 70 S.Ct. at 511. A majority of the Court concluded that the trial judge's charge to the jury had adequately explained this proposition. In dissent, Justice Jackson opined that the jury charge had failed to clarify the issue, but he agreed with the majority on the rule of law: "Before a failure to couple establishes a defective coupler, it must be found that it was properly set so that it could couple. If it was not adjusted as such automatic couplers must be, of course the failure is not that of the device." Id. at 101, 70 S.Ct. at 512 (Jackson, J., dissenting).

Affolder stands for the proposition that if the jaws of both knuckles are closed at the time of impact, then the railroad may offer evidence to explain the mishap and the failure to couple does not, in and of itself, establish a per se violation of the Act. The Affolder case, however, did not present facts dealing with another frequent cause of failed coupling which we discussed earlier, that is, misalignment of the drawbar.

The Supreme Court cryptically referred to the misalignment problem in Atlantic City R.R. Co. v. Parker, 242 U.S. 56, 37 S.Ct. 69, 61 L.Ed. 150 (1916), where plaintiff was injured while straightening the drawbar. After acknowledging that "[s]ome lateral play must be allowed to drawheads," the Court said, "[i]f couplers failed to couple automatically upon a straight track, it at least may be said that a jury would be warranted in finding that a lateral play so great as to prevent coupling was not needed, and that in the absence of any explanation believed by them, the failure indicated that the railroad had not fully complied with the law." Id. at 59, 37 S.Ct. at 70. The critical phrase is "in the absence of any explanation believed by them." The Court apparently determined that the jury could draw an inference that the railroad was liable because of the failure to couple, but the railroad had an opportunity to explain the incident.

The other Supreme Court case discussing drawbars relied on by plaintiff is San Antonio A.P. Ry. Co. v. Wagner, 241 U.S. 476, 36 S.Ct. 626, 60 L.Ed. 1110 (1916). That case, however, is inapposite because there was evidence of defective equipment and the Court stated, "[w]e need not in this case determine . . . that the failure of a coupler to work at any time sustains a charge that the Act has been violated." Id. at 484, 36 S.Ct. at 629.

In short, Affolder put the closed knuckles controversy to rest, but the other common cause of failed coupling, drawbar misalignment, has not been definitively resolved by the Supreme Court. Perhaps not surprisingly, the Courts of Appeals and various state courts have differed in their approach to this question.

Some courts have given Affolder a narrow reading. In Hallada v. Great Northern Railway, 244 Minn. 81, 69 N.W.2d 673, cert. denied, 350 U.S. 874, 76 S.Ct. 119, 100 L.Ed. 773 (1955), the Minnesota Supreme Court decided that misalignment created absolute liability and construed Affolder to apply only to situations in which the jaws of knuckles were not opened. In Kansas City S. Ry. Co. v. Cagle, 229 F.2d 12 (10th Cir. 1955), cert. denied, 351 U.S. 908, 76 S.Ct. 697, 100 L.Ed. 1443 (1956), the Court of Appeals for the Tenth Circuit took a similar stance. See also Clark v. Kentucky I. Terminal R.R., 728 F.2d 307, 312 (6th Cir. 1984) and cases cited within. Dictum in the latter case is consistent with Hallada, but the opinion alludes to evidence showing misalignment caused by defective equipment.

Other courts have acknowledged a broader application of Affolder. In Maldonado v. Missouri P. Ry. Co., 798 F.2d 764 (5th Cir. 1986), cert. denied, 480 U.S. 932, 107 S.Ct. 1571, 94 L.Ed.2d 762 (1987), the Court determined that the railroad failed to produce sufficient evidence to justify submitting a defense of misalignment to the jury. The Court, however, reasoned that the Safety Appliance Act was essentially an equipment safety statute and said,

"[D]rawbar misalignment prior to impact may not support a defense unless the misalignment is due to some factor other than equipment failure or defect. . . . Evidence that the drawbars were misaligned so as to prevent coupling places a burden on the railroad to produce evidence tending to show a separate cause for the failure to couple."

Id. at 768. The Court did caution that a drawbar with "more lateral play than is appropriate for normal operation" would violate the Act.

The Court in United Transportation Union v. Lewis, 711 F.2d at 243, after an extensive review of the statutory language and legislative history, concluded that the Act prescribed the equipment that railroads must provide. "Nothing in the plain language suggests an intent to impose an independent prohibition against procedures that require men to go between cars." The Court also noted, "that the drawbar goes out of alignment with the next car due to vibration or jarring does not indicate that the coupler is defective." Id. at 235 n. 5.

We think the rationale of Affolder, which allows a defense of closed knuckles, applies equally to a situation where non-defective equipment is misaligned. In both instances, the causative factor which prevents the coupling may result from jarring or vibration of the car while moving, or from human error in failing to open the knuckle or align the drawbar. In these situations, the failure to couple might not be caused by defective equipment and consequently is not a violation of the Safety Appliance Act.

In light of the reasoning in Maldonado and United Transportation and after thorough consideration of the issue, we are persuaded that the district court correctly analyzed the statute and appropriately charged the jury: "The defendant must present evidence to show that the misalignment did not occur because of equipment failure. In other words, the defendant has to establish a separate cause for failure to couple, other than equipment failure." We therefore reject plaintiff's contention that the mere failure to couple establishes a per se violation of the Safety Appliance Act and forecloses a defense by the railroad.

After struggling to reconcile the goals of the Safety Appliance Act and FELA with the hazards and expense of litigation, one is led to conclude that Justice Douglas was on the right track when describing FELA as "[c]rude, archaic, and expensive as compared with the more modern systems of workmen's compensation." Bailey v. Central V. Ry. Inc., 319 U.S. 350, 354, 63 S.Ct. 1062, 1064, 87 L.Ed. 1444 (1943). Justice Frankfurter took the same route in describing FELA as "anachronistic" and lamenting the absence of workers' compensation programs for railworkers. Carter v. Atlanta S.A.B. Ry. Co., 338 U.S. 430, 437-38, 70 S.Ct. 226, 230-31, 94 L.Ed. 236 (1949) (dissenting opinion).

History provides some explanation for the lack of workers' compensation for railworkers. In 1908 when FELA was enacted, the constitutionality of workers compensation plans was in doubt and it is understandable that Congress chose not to provide a remedy that might not last. Report of the Federal Courts Study Committee, 62-63 (1990). That circumstance no longer exists, and "[w]ere Congress today to establish a program to provide compensation for injured workers, it would undoubtedly adopt a workers' compensation program." Id. at 63.

Under Workers' Compensation laws, reimbursement of medical expenses and, at least to some extent, lost wages do not rely on such subtleties as have been raised in this case. Recompense for industrial injuries such as those suffered by plaintiff here should not have to depend on the vagaries of a statute such as the Safety Appliance Act or FELA or on the whims of juries or the delays of litigation. See Schwartz Mahshigian, The Federal Employers' Liability Act, a Bane for Workers, a Bust for Railroads, a Boon for Lawyers, 23 San Diego L.Rev. 1 (1986) (arguing that FELA is counterproductive to its original purposes, and calling for more suitable workers' compensation mechanisms).

II. [38] THE CLOSING ARGUMENTS

During his closing argument, counsel for plaintiff spoke in terms that the trial judge described as "misleading and prejudicial," and "unfair in tone, nature, and content." The judge observed that, "the written record does not disclose the inflammatory and emotional performance by plaintiff's counsel."

Plaintiff's lawyer stated that the railroad was "callous" in producing evidence that most of its employees retired at the age of 62. He went on to say, "They treated Charlie like the junk that they carry in these things." At this point, defense counsel objected and the trial judge responded, "It is argument." Defense counsel then objected specifically to the "reference to treating people like junk." The judge said at that point, "Well, it's argument. The jury will have to determine what it wants to do with the argument. Go ahead, I will overrule it. It's argument."

Plaintiff's counsel then returned to his callousness comments and later appealed to the jury as the "collective conscience of this community." He challenged them to decide "whether it's appropriate for a railroad . . . to injure a man because their equipment is defective and in violation of the Safety Appliance Act and then to try to deprive him of every right that he has under the law."

Before defense counsel began his closing remarks, he renewed his objection to plaintiff's argument, pointing out that plaintiff had received disability payments — evidence the court had ruled inadmissible in an in limine order. The defense lawyer stated he wished to tell the jury that "we had taken care of him [plaintiff] and that he is on disability." After listening to plaintiff's counsel's objections, the trial judge replied, "You're entitled to make an emotional appeal, but you are not entitled to make an egregiously unfair emotional appeal, which is what you have done. And therefore I will permit [defense counsel] to meet that in the only way he can and that is to comment on the things he wishes to comment."

At the beginning of his argument, defense counsel told the jury, "We have not cast Mr. Reed aside. Mr. Reed has been taken care of. Mr. Reed is being provided disability through a federal system and by the company and we have not been callous or disregard of Mr. Reed's welfare or his years of service with the company."

Plaintiff's counsel interjected: "I object to [defendant] taking any credit for a federal system and that is a direct violation of the collateral source rule. There is no evidence whatsoever in the record." At a later point in his closing, defense counsel commented on the possibility that plaintiff would have retired early had the injury not occurred. "[Y]ou heard his description for his own proof in response to a question by his own attorney that with retirement he felt — or with being on disability, he felt like a free man, that he didn't have to punch a time clock."

In matters of trial procedure such as that involved here, the trial judge is entrusted with wide discretion because he is in a far better position than we to appraise the effect of the improper argument of counsel. We accept the trial judge's evaluation that plaintiff's closing argument was improper, but we cannot agree that it was appropriate corrective action to permit defense counsel's reference to evidence not in the record and objectionable in any event.

The prohibition against use in argument of evidence not in the record is straightforward. Waldron v. Waldron, 156 U.S. 361, 380, 15 S.Ct. 383, 388, 39 L.Ed. 453 (1895) ("it is equally well established that the assertion by counsel, in argument, of facts, no evidence whereof is properly before the jury, in such a way as to seriously prejudice the opposing party, is, when duly excepted to, also ground" for reversal); Draper v. Airco. Inc., 580 F.2d 91, 96 (3d Cir. 1978) ("Reference to facts not in evidence is improper."); Ayoub v. Spencer, 550 F.2d 164, 170 (3d Cir.) ("The remarks of counsel were required to be confined to the evidence admitted in the case. . . . Reversible error is committed when counsel's closing argument to the jury introduces extraneous matter which has a reasonable probability of influencing the verdict."), cert. denied, 432 U.S. 907, 97 S.Ct. 2952, 53 L.Ed.2d 1079 (1977).

Moreover, the situation is exacerbated because the same facts that the defense presented to the jury had been explicitly ruled inadmissible at an in limine hearing, a decision in accord with existing law. We concede that the continuing vitality of the collateral source rule has been challenged on the ground that its economic justification is marginal at best, but we cannot disregard the teachings of the Supreme Court in such cases as Tipton v. Socony Mobil Oil Co., Inc., 375 U.S. 34, 84 S.Ct. 1, 11 L.Ed.2d 4 (1963) and Eichel v. New York Central R.R. Co., 375 U.S. 253, 84 S.Ct. 316, 11 L.Ed.2d 307 (1963).

In Eichel, a FELA suit, the Court said that evidence of a disability pension was properly excluded. "[T]he likelihood of misuse by the jury clearly outweighs the value of this evidence . . . [P]etitioner's receipt of collateral social insurance benefits involves a substantial likelihood of prejudicial impact." Id. at 255, 84 S.Ct. at 317. In Tipton, the Court rejected defense counsel's argument that because no damages were awarded, evidence of payments under the Longshoremen and Harbor Workers Compensation Act, 33 U.S.C. § 901 et seq., was harmless error. 375 U.S. at 35, 84 S.Ct. at 2. The Court did not limit the application of the collateral source rule to the damage phase of the trial and noted that the jury was led to place undue emphasis on the compensation benefits in determining a liability issue, namely whether the plaintiff was a seaman under the Jones Act.

The situation here is somewhat different, but under the current state of the law we cannot say the error was harmless. See McQueeney v. Wilmington Trust Co., 779 F.2d 916, 917, 924-928 (3d Cir. 1985) (holding that court can find errors harmless only if it is highly probable that the errors did not affect the outcome of the case); cf. Eichmann v. Dennis, 347 F.2d 978, 982 (3d Cir. 1965) (court's instructions to jury that plaintiff had right of recovery under workmen's compensation law was prejudicial and thus not harmless error); Snyder v. Lehigh Valley R.R. Co., 245 F.2d 112, 116 (3d Cir. 1957) (en banc) ("The possibility that [the jury] might have been influenced to reach a verdict of non-negligence on the part of defendant by reason of the circumstance that the plaintiff was not receiving workmen's compensation cannot be lightly discounted."). Plaintiff's counsel's behavior warranted strong action by the trial court, but the proverbial two wrongs do not make a right.

Accordingly, the judgment of the district court will be vacated and the matter will be remanded for a new trial. Each party to bear its own costs.