On August 7, 1979, the Secretary of Labor brought this suit against TIC International Corporation, which he charged with having violated the "prudent man" rule in section 404(a)(1)(B) of the Employee Retirement Income Security Act, 29 U.S.C. Sec. 1104(a)(1)(B), by advising a teamsters union health and welfare plan to make a "hold harmless" agreement with an insurance company as a result of which the plan lost $750,000. The district court granted TIC's motion for summary judgment and dismissed the complaint, on the ground that the suit was barred by ERISA's statute of limitations. The Secretary has appealed. Ordinarily the issue on an appeal from a judgment entered on a motion for summary judgment is whether there was a genuine issue of material fact, in which event the grant of summary judgment would be inappropriate. In this case, however, the appellant does not want an opportunity to introduce evidence relating to the statute of limitations. He is content to argue that the record of the summary judgment proceeding establishes the inapplicability of the statute of limitations. In effect he asks us to review the decision of the district court as if the court had entered judgment following a bench trial. Compare Schlytter v. Baker, 580 F.2d 848 (5th Cir.1978); 10A Wright, Miller & Kane, Federal Practice and Procedure Sec. 2720, at pp. 26-27 (2d ed. 1983); 6 Moore's Federal Practice p 56.13, at p. 56-347 (2d ed. 1985).

Section 413(a), 29 U.S.C. Sec. 1113(a), provides that no suit with respect to a fiduciary's breach of duty under ERISA shall be brought after the earlier of (1) six years after the date of breach or

(2) three years after the earliest date (A) on which the plaintiff had actual knowledge of the breach or violation, or (B) on which a report from which he could reasonably be expected to have obtained knowledge of such breach or violation was filed with the Secretary under this title; except that in the case of fraud or concealment, such action may be commenced not later than six years after the date of discovery of such breach or violation.

The district court held that the Secretary "could reasonably be expected to have obtained knowledge" of TIC's alleged breach of trust from the plan's annual report, which had been filed with the Secretary on January 15, 1976--more than three years before the suit was filed. There is no suggestion of fraud or concealment, which would extend the deadline.

The report in question is Form D-2, the annual report that the plan was required to file with the Secretary for 1975. (Form D-2 was later superseded by Form 5500, but without any consequences for the issue in this case.) It is 26 pages long, the last two being the auditor's report. Note 3 to the auditor's report states that "The group insurance contract with the insurance carrier, which was terminated on February 1, 1975 (see Note 5), provided that to the extent premiums paid exceed the sum of the claims paid and provided for ..., a refund of premiums is to be made to the Fund [plan].... As of February 28, 1975, no significant adjustment of premium liability or refund is anticipated." Note 5 states:

The Board of Trustees terminated the group insurance contract and made the Fund self-insured as of February 1, 1975. The Board also signed a hold harmless agreement which released the insurance carrier from claims incurred prior to February 1, 1975 but not yet paid as of that date. Therefore, all future claims will be paid directly by the Fund. The liability for claims incurred but not paid at February 28, 1975 is estimated to total approximately $850,000. This estimate is based upon a review of direct claims paid since February 28, 1975.

In other words, before February 1 the employers' contributions to the plan had been used to pay premiums for health insurance and the insurance company had paid the employees' claims for benefits under the plan, but on February 1 the plan had become self-insured, meaning that it saved the premium expense but now would have to pay the claims directly. This switch in itself need not have caused any loss to the plan. But in connection with the switch the plan had made an agreement--the "hold harmless" agreement--excusing the insurance company from having to pay any claims accrued but unpaid before February 1, claims estimated at $850,000 as of February 28, the last day of the plan's fiscal year. One might have thought that in exchange for conferring this benefit on the insurance company the plan would have gotten a refund of premiums or some other consideration, but Note 3 to the auditor's report suggests that there was no other consideration. Read together, Notes 3 and 5 suggest that the plan made an utterly improvident, one-sided contract with the insurance company.

When the report was filed with the Department of Labor, it was read by a compliance specialist who was made suspicious by the auditor's report and began an investigation which revealed that the plan's advisor, TIC, had recommended the hold-harmless agreement to the plan's trustees. (The report did not mention TIC but another report on file with the Department, the plan description, listed TIC as a consultant.) This investigation eventuated in the present lawsuit.

The question whether the Department of Labor "could reasonably be expected to have obtained knowledge" of TIC's alleged breach of trust from the Form D-2 filed in 1976 may seem artificial. Since we know that the report kicked off the investigation that led to the filing of this lawsuit, it might seem that whatever might reasonably have been expected, the Department in fact obtained knowledge of the violation from the report. But the purpose of the three-year statute of limitations is to penalize the Department not for extraordinary diligence and imagination in scrutinizing reports but for negligent dawdling in the face of unmistakable evidence of a probable violation. If a reasonable person would not have been alerted to a probable violation by reading the report that the plan filed for 1975, the suit was filed within the statutory period even if an extraordinary person would have been, and was, alerted by the report and set the investigatory machinery in motion.

The question when a reasonable person would have known that his legal rights had been invaded, so that the statute of limitations began to run, is a question of fact, meaning that we can reverse the district court's determination only if it was clearly erroneous. See, e.g., Glass v. Petro-Tex Chemical Corp., 757 F.2d 1554, 1561-62 (5th Cir.1985); Castorina v. Lykes Bros. S.S. Co., 758 F.2d 1025, 1034 (5th Cir.1985); Miles v. New York State Teamsters Conference Pension & Retirement Fund Employee Pension Benefit Plan, 698 F.2d 593, 598-99 (2d Cir.1983). This is the standard of review, the Supreme Court has reminded us recently, "even when the district court's findings do not rest on credibility determinations, but are based instead on physical or documentary evidence or inferences from other facts." Anderson v. City of Bessemer City, --- U.S. ----, 105 S. Ct. 1504, 1512, 84 L. Ed. 2d 518 (1985). It is not that the district court is in a better position than this court, in any realistic sense, to determine whether the plaintiff should have moved faster than it did to bring the lawsuit; it is that district judges are specialists in finding facts, as we are not, and that our primary function, which is to maintain the uniformity and coherence of the law, is not engaged by a judgment so dependent on the specific circumstances of each case. There will never be an auditor's report quite like this, so there is no pressing need to have a uniform rule prescribing its legal significance.

The district court's determination was not clearly erroneous. The auditor's report is the first place to which an analyst whose duty it is to read the annual reports of employee benefit plans for compliance with the federal standards, centrally including the prudent man rule, would turn; and when he did that here he would have seen (he did see) that the plan had made a contract under which it apparently gave up $850,000 in exchange for nothing. The auditor did not say, "The trustees or their advisors goofed, and blew $850,000 of plan assets"; that is not the customary diction of auditors; if the auditor had said that, the Department would have had actual knowledge of the breach of trust and the three-year period would have started to run anyway. The "could reasonably be expected" test, the alternative basis for starting the three-year period running, is intended to pick up the case where a probable violation is evident to anyone who reads the report with reasonable care. Here it was not even a matter of reading between the lines; all that was necessary was to read Notes 3 and 5 together--as Note 3 invited the reader to do. The Department says that "hold harmless agreement" is unclear, but Note 5 defines it--as excusing the insurance company from liability for $850,000 in accrued claims. There is no mention of any consideration for the fund's assuming this liability. Note 3 suggests that the consideration was to have been a refund or other adjustment in premiums paid by the plan to the insurance company; for the insurance contract had provided for a refund if premiums exceeded claims--and the insurance company had just been allowed to shuck off $850,000 in claims. But there had been no refund; none was anticipated; in short the consideration had in fact turned out to be zero. The Department's conjectures that there may have been consideration in the form of forgiveness of unpaid premiums or some adjustment between February 1 and 28 require a most strained reading of Note 3.

The Department makes much of the fact that the Form D-2 nowhere mentions TIC. That is irrelevant. If the form reveals a probable breach of trust, someone must be liable, quite possibly the trustees, see 29 U.S.C. Sec. 1109(a); if they think an advisor or consultant is liable instead, see 29 U.S.C. Secs. 1002(21)(A), 1105, they will not be slow to point that out. There is no suggestion that anyone was deceitful or even coy with regard to the contract between the plan and TIC. The Form D-2 indicated that whoever had advised the plan to make the "hold harmless" agreement had acted imprudently, in violation of ERISA; and it took almost no time to find out who that was.

The omission of TIC's name could be significant only if we accepted the Department's argument that the three-year statute of limitations applies only if every element of the violation, including the name of the violator, is evident on the face of the report, with no need for inference or further inquiry. The language of section 413(a)(2)(B) cannot sustain such an interpretation; "a report from which he could reasonably be expected to have obtained knowledge of such breach or violation" cannot mean, "a report that demonstrates on its face that a breach or violation has occurred." If it did mean that, subsection B would have no application to the Department of Labor, for the Department has actual knowledge of the violation (subsection A) in any case where a report actually confessing a violation is filed. The Department states in its main brief in this court that it wants us to require "actual knowledge" for B as well as for A; that would make B superfluous.

The Department makes a number of arguments in support of this position. One is that the principle that a statute of limitations begins to run when the plaintiff should have known that his rights were being violated is normally used to extend, rather than as here to contract, statutes of limitations; a discovery rule is a tolling rule. But we cannot see what difference that makes. Congress evidently believed that three years was enough time to bring an ERISA suit once the plaintiff either knew or, from a filed report, should have known of the violation. We attach no importance to the fact that Congress did not enact a version of ERISA which would have applied the three-year statute of limitations to any case where the plaintiff "knows or has reason to know of such violation." S. 1179, 93d Cong., 1st Sess. Sec. 501(d)(15), at 176 (1973). In the bill ultimately passed by the Senate this provision was deleted--but why, no one knows. Section 413(a)(2)(B) was part of the House bill. At conference--again for reasons nowhere explained--the Senate conferees agreed to adopt this part of the House bill. See Summary of Differences Between the Senate Version and the House Version of H.R. 2 to Provide for Pension Reform, pt. 3, at p. 11 (Comm. Print 1974), reprinted in 3 Legis. History of ERISA, 94th Cong., 2d Sess. 5249, 5261 (Comm. Print 1976). To interpret this history as a congressional determination to require actual knowledge under B as well as A--the Department's position in this case--is imaginative in the extreme.

The Department points to what it considers the following anomaly if the statute of limitations is read as written: if the Department has actual knowledge of a violation, it has three years to sue; if it has only suspicion prompted by a report, it still has only three years to sue. Yet whenever a statute of limitations is subject to a discovery rule, the interval for preparing and filing a suit is the same whether the plaintiff knows that he has a cause of action or should know, even though in the latter case he might have to do some more investigating before he could be certain that he had a cause of action. This point is so elementary that the cases assume rather than discuss it. See, e.g., United States v. Kubrick, 444 U.S. 111, 121, 123, 100 S. Ct. 352, 359, 360, 62 L. Ed. 2d 259 (1979); Dozier v. Trans World Airlines, Inc., 760 F.2d 849, 851 (7th Cir.1985) (per curiam); Metz v. Tootsie Roll Industries, Inc., 715 F.2d 299, 304 (7th Cir.1983); United States v. Beck, 758 F.2d 1553, 1558 (11th Cir.1985); Samples v. Ryder Truck Lines, Inc., 755 F.2d 881, 887 (11th Cir.1985); Suthoff v. Yazoo County Industrial Development Corp., 722 F.2d 133, 138 (5th Cir.1983). The Department argues that the approach taken by the district court will encourage plan trustees to file unilluminating reports. But it will not, because if they do the statute of limitations will not begin to run so soon; and of course if they actually conceal the breach of trust the Department will have six years from the date of discovery to sue.

The Department's real quarrel is not with the standard used by the district court but with Congress's determination to allow the Department only three years to bring a suit when the Department could reasonably be expected to learn of a violation from a report. The Department says it lacks the resources to investigate the 800,000 annual reports that it receives. But if so it should ask Congress for a higher appropriation or (more realistically in this era of budgetary stringency) a longer statute of limitations. As a matter of fact its analyst did read the plan's annual report. It was filed in January and by October the analyst was phoning the trustees to find out more about the hold-harmless agreement. At this point the statute of limitations still had more than two years to run. The delay in filing the suit has nothing to do with any deficiency in the Department's resources for screening annual reports.

The Department points out that the plan's beneficiaries will have even more trouble complying with the three-year deadline than the Department. This is because the beneficiary receives, not the report itself, but a summary of the report, and the summary may not contain any hint of the violation that could reasonably be obtained from knowledge of the full report. It may be anomalous to give the beneficiaries only three years to sue, when they must first send for the report in order to get it, but this anomaly is built into the statute, reflects we imagine a realistic awareness that only the Department is going to spot a violation from a report, and will not be cured by our rewriting the statute as the Department asks us to do.

Notes 3 and 5 in the auditor's report would not have tipped off beneficiaries to the possibility of a breach of trust even if they had sent for the report, as they are entitled to do. Realistically, the responsibility for reading the fund's annual report with an eye to possible violations of ERISA is the Department's rather than the plan beneficiaries'; and the latter may actually benefit from our holding that the three-year statute of limitations for suing on violations revealed by a report means what it says. The Department will have a greater incentive to be prompt in investigating and bringing suit for breaches of trust, and the gain to the beneficiaries of employee benefit plans from its promptness may outweigh the occasional losses to them from the Department's failing to sue within the statutory period.

We are given brief pause by the statement in Fink v. National Savings & Trust Co., 772 F.2d 951, 957 (D.C.Cir.1985), that "the disclosure of a transaction that is not inherently a statutory breach of fiduciary duty (in contrast to, e.g., a loan to a party-in-interest which is explicitly prohibited under 29 U.S.C. Sec. 1106(a)) cannot communicate the existence of an underlying breach." The "hold harmless" agreement in this case was not "inherently a statutory breach of fiduciary duty," in the sense of a type of agreement explicitly prohibited by ERISA; the problem was the apparent lack of any consideration for the agreement. But the statement in Fink is unnecessarily broad. The transaction at issue there was the plan's continued investment of the bulk of its assets in the employer's securities during a year in which the plan was losing money, without adequate investigation of whether this was a prudent course of action. The fact that the plan was buying securities of the employer was no evidence of a probable breach of trust; that type of "self-dealing" is (as we are about to see) allowed by ERISA. The fact that the plan lost money one year did not show a probable breach of trust either. And although the fact that it was investing most of its assets in just the one company (the employer) suggested a failure to diversify, the opinions in Fink make little of the point, because a plan of the sort involved in Fink is allowed to invest all of its assets in the employer's securities, if that is prudent in the circumstances. See 29 U.S.C. Secs. 1104(a)(2), 1108(e). The only apparent breach of trust was the plan's failure to look before it leaped--and of that failure the report gave at most a hint. Here the report clearly signaled imprudence by describing a transaction in which the plan had given up a major asset apparently for nothing.

The difference between the cases is not great and we might have decided Fink differently from the way the majority of the D.C. Circuit's panel decided it but there is enough factual difference to allow both decisions to co-exist. Brock v. Tricario, 768 F.2d 1351 (11th Cir.1985) (per curiam), a recent unpublished opinion of the Eleventh Circuit, merely upheld as not clearly erroneous the district court's finding that a plan's annual report had not furnished adequate notice of a probable violation. The court applied the same standard as we (clear error), to different facts.

The district court did not commit a clear error in holding that the Department of Labor could reasonably have been expected to discover the advisor's alleged breach of trust from the plan's annual report. The auditor's notes fairly shouted imprudence; no more was required to set the three-year statute of limitations running.

AFFIRMED.