concurring and dissenting.
I agree with the majority’s conclusion that the district court’s judgment should be affirmed. I write separately, however, because I disagree that ERISA’s statute of limitations bars the Appellants’ claim.
My disagreement lies with the majority’s interpretation and application of Gluck v. Unisys Corporation, 960 F.2d 1168 (3d Cir. 1992). In Gluck, we held that “actual knowledge of a breach or violation requires knowledge of all relevant facts at least sufficient to give the plaintiff knowledge that a fiduciary duty has been breached or ERISA provision violated.” Gluck, 960 F.2d at 1178. This requires “the district court [to] determine, as a factual matter, the date on which each employee had actual knowledge of the breach or violation.” Id. at 1176. Knowledge of the breach is distinguished from knowledge of the actions constituting the breach or violation. See id.; International Union of Electronic Workers v. Murata Erie North America, Inc., 980 F.2d 889, 900 (3d Cir.1992) (requiring a showing “that plaintiffs actually knew not only of the events that occurred which constitute the breach or violation but also that those events supported a claim of breach of fiduciary duty or violation under ERISA”).
Noting Gluck’s “stringent” actual knowledge requirement, the district court found that although the pension fund “demonstrated that plaintiffs became aware of their loss of pension credits during 1981,” the fund had “failed to demonstrate that [Connell and Nelson] actually knew of their potential ERISA cause of action.” Connell v. Pension Fund, Civ. No. 92-1655, slip op. at 16 (D.N.J. Jan. 2, 1996). To support this conclusion, the district court drew certain factual inferences from the testimony presented at the non-jury trial. Specifically, the court found that Connell’s failure to take any action after learning of his loss of pension credits other than *159contacting the plan administrator and Nelson’s failure to express any concerns to the plan administrator about the break-in-service rule “clearly indicate[d] that [they] were not aware of their potential cause of action.” Id. at 6,11 & 16.
The majority seems to overlook the district court’s factual inferences, concluding:
[I]t is undisputed in this case that Connell and Nelson knew as early as 1981 or 1982 that the Fund had canceled their pension credits by applying the break-in-service rule and that the Fund’s actions would diminish their future benefits. Their actual knowledge of both the alleged breach and the consequential injury they would thereby suffer gave them “knowledge that a fiduciary duty [may have] been breached or ERISA provision violated.”
Maj. Op. at 158 (quoting Gluck, 960 F.2d at 1178) (alteration of Gluck in original). The majority’s conclusion does not follow from its premise. Although it was established that Connell and Nelson knew in 1981 or 1982 of the pension fund’s action which constituted the alleged breach—that is, its enforcement of the break-in-service rule—it does not follow that they had actual knowledge of the breach itself. The majority further notes, again without reference to the district court’s findings, that Connell and Nelson must have had “ ‘actual knowledge of all of the [material] elements of the violation’ ” because they “felt sufficiently aggrieved by the Fund’s actions to question their union representatives.” Maj. Op. at 158 (quoting Gluck, 960 F.2d at 1178).
As noted earlier, however, the district court reached the exact opposite conclusion from the same facts and testimony cited by the majority. Indeed, the district court found that Connell’s and Nelson’s actions after receiving notice of their lost pension credits “clearly indicate[d] that plaintiffs were not aware of their potential ERISA cause of action.” Connell, Civ. No. 92-1655, slip op. at 16. I would not disregard, as the majority does, the district court’s factual determination that Connell’s and Nelson’s actions, or lack thereof, indicated that they had no knowledge of a breach.
Moreover, contrary to the majority’s indication, it is not enough that the plaintiffs have knowledge that a fiduciary duty “may have” been breached or ERISA provision violated. Rather, the plaintiff must have actual knowledge that a fiduciary duty “has” been breached or ERISA provision violated. See Gluck, 960 F.2d at 1178.
While this distinction may seem merely semantic, it is important, as perhaps the following hypothetical will explain. Assume that Smith and Jones were non-union iron workers who occasionally worked in the iron trade and participated in a multi-employer pension fund that supported both union and non-union workers. In 1981, Smith and Jones received notice that some of their pension credits would be lost because they had incurred breaks in service. Assume further that they questioned a representative of the Fund but were told that the Fund was merely enforcing its break-in-service rule, of which they were on notice. Jones and Smith were upset about their lost credits and may have even assumed that the Fund was treating them unfairly. But, without any other information, Jones and Smith thought that they were just out of luck. Now assume that in 1990, Jones and Smith discovered that the Fund was in practice only applying the break-in-service rule to non-union workers.
Under Gluck, the statute of limitations would start to run in 1990, when Smith and Jones learned that the Fund was discriminating in its enforcement of the break-in-service rule. In other words, the statute would start to run when they possessed knowledge that the enforcement of the break-in-service rule violated ERISA. Under the majority’s analysis, which, in my view, relaxes Gluck’s “stringent” requirement, Jones and Smith would be barred from bringing their claim because the three-year statute would have started running in 1981 when they first learned that they had lost credits due to their break in service.
The hypothetical discussed above is not that different from the ease at hand. Here, Connell and Nelson admit that they knew in 1981 or 1982 that they had lost credits. But they allege that they incurred the breaks in service because the union had discriminated *160against them in allocating iron work. Thus, if Connell and Nelson could show that they had left the ironworking trade involuntarily, enforcement of the break-in-service rule against them might have constituted a breach of ERISA.1 Had the union’s discrimination forced Connell and Nelson out of the trade involuntarily, and had they not had actual knowledge of the discrimination (say, for example, they thought that they just were unlucky in obtaining work), they would not have possessed the requisite “actual knowledge” of a breach of fiduciary duty or ERISA violation merely when they had been notified of their lost credits. Indeed, they would be lacking “actual knowledge of all material facts constituting [the] breach of fiduciary duty or violation of ERISA [which] is the sine qua non for application of [ERISA’s] three-year limitation.” Gluck, 960 F.2d at 1177.
Put simply, then, the actual knowledge requirement is necessarily intertwined with the cause of action or the theory of the breach. See Martin v. Consultants & Administrators, Inc., 966 F.2d 1078 (7th Cir. 1992) (courts must take into account “the complexity of the underlying factual transaction, the complexity of the legal claim and the egregiousness of the alleged violation”).
At bottom, a determination of what the plaintiffs knew and when is a very fact-intensive inquiry. See Gluck, 960 F.2d at 1176 (requiring the district court to determine “as a factual matter” when the statute began to run). I cannot agree with the majority’s disregard of the inferences that the district court drew from the facts here.
. We recognized in Knauss v. Gorman, 583 F.2d 82 (3d Cir.1978), that if a plan beneficiary incurs a break in service involuntarily, the pension fund must come forward with a justification for why the rule is enforceable. See also Van Fossan v. International Brotherhood of Teamsters, 649 F.2d 1243, 1248 (7th Cir. 1981) ("We believe the distinction between voluntary and involuntary breaks in service is crucial to determining the arbitrariness of the operation of a given break in service rule.”).