Morrison-Knudsen Corporation (MK), a federal contractor, terminated its subcontractor, Ground Improvement Techniques (GIT), for an alleged default, and sued GIT for damages.1 GIT counterclaimed for wrongful termination. GIT's claimed damages included payment for completed work under the subcontract, equitable adjustments to the subcontract price for increased costs caused by MK, damages claimed by lower-tier subcontractors, and attorney's fees. MK unsuccessfully moved for judgment as a matter of law, claiming insufficiencies in GIT's evidence of damages and of MK's liability. A jury found the termination wrongful and awarded GIT roughly half the damages claimed.

On appeal, this court rejects MK's challenge to the liability verdict but reverses the damage award. GIT's evidence of several of its categories of damages was insufficient, and its claims on behalf of its lower-tier subcontractors were premature, as GIT had not yet itself settled with all of its subcontractors. Because the jury returned a general verdict, this court cannot determine whether any parts of the jury's award were for allowable categories of damages supported by sufficient evidence. We thus vacate the judgment and remand for a new trial limited to the issue of damages.I. FACTUAL AND PROCEDURAL BACKGROUND

The United States Department of Energy hired MK in 1983 to manage its Uranium Mill Tailing Remedial Action (UMTRA) project, a cleanup of radioactive mill tailings at sites around the country. MK subcontracted with GIT in March 1995 to clean up the Slick Rock, Colorado, site. GIT hired several lower-tier subcontractors ("subs"), including R.N. Robinson & Son, for excavation; Bogue Construction, for trucks; Keers Environmental, for asbestos abatement; and GA Western, for bridge work.

The MK-GIT Subcontract ("the contract") obligated GIT to complete the project by December 1996. The contract price was roughly $9.3 million. The contract incorporated almost verbatim several standard federal clauses for fixed-price construction contracts, including clauses governing terminations for default and convenience. It also incorporated by reference the federal regulations governing the compensability of contractors' costs in the event of a termination for convenience. The contract provided that the law applicable to government procurement would govern all substantive issues in any litigation.

The project did not go well. GIT and its subs encountered delays, difficulties, and increased costs. GIT attributed these to MK's defective specifications, failure to timely secure permits, rigid interpretation of specifications and safety requirements, and propensity to reject proposed work plans. During the contract's performance, GIT requested extra compensation and extensions of time because of delays to, changes in, and increased costs of the work which GIT attributed to MK. GIT's central theory is that its plan to complete the project before the deadline displeased MK, who could not then earn the maximum possible fees from DOE. MK, in GIT's view, thus sought to hinder and delay the work. MK, on the other hand, attributed the delays and increased costs to errors, omissions, and delinquencies by GIT and its subs.

In September 1995 MK terminated GIT for default. The contract allowed MK to do so if GIT was not prosecuting the work with a diligence that would ensure its timely completion. MK simultaneously sued GIT for damages caused by its alleged default. While requiring GIT to cease work and vacate the site, MK also directed it to perform certain cleanup work and leave certain equipment behind. MK allegedly retained and used that equipment during the ensuing litigation. After the termination, MK denied almost all of GIT's requests for change orders to increase its compensation under the contract. MK also failed to pay GIT for various parts of the completed work and for the post-termination work and retention of equipment.

GIT protested the termination and urged MK to let it complete the project or bid on the reprocurement of the work. MK ignored these requests. In February 1996 GIT counterclaimed for wrongful termination, seeking damages in the form of payment for completed work under the contract and compensation for additional costs occasioned by MK and not contemplated by the contract.

Contemporaneously, the subs were demanding payment from GIT. After Keers filed suit, GIT settled its claims. At the time of the GIT-MK trial, however, GIT was still involved in litigation with Robinson and had not settled with or paid Bogue or GA Western. In its counterclaim, GIT also sought damages on behalf of the subs.

The district court eventually set trial for November 1996. In October 1996 GIT supplemented its pretrial damages disclosure, increasing the amount claimed from roughly $8.4 to $11.4 million and increasing the number of categories of damages. MK repeatedly but unsuccessfully challenged the supplementation, arguing that GIT had changed its damage theory just weeks before trial and was using previously undisclosed documentation.

Twelve days before trial, the court assigned the case to a new judge. That judge presided over a three-week trial, which focused on whether MK's termination of GIT had been wrongful. GIT presented one witness, its secretary/treasurer, Kip Cooper, to explain its damage exhibits and claims.

Before the court submitted the case to the jury, MK filed several motions for judgment as a matter of law under Federal Rule of Civil Procedure 50(a). MK made several challenges to GIT's damage evidence. It argued, inter alia, that GIT had not presented evidence of the types of damages allowed by the contract, or of causation, or of its attorney's fees. The court denied MK's motions. MK also argued in a jury-instruction conference that the contract barred GIT from recovering on behalf of subs whose claims GIT had not settled and paid. The court rejected MK's proposed jury instruction to that effect.

The jury found the termination wrongful and awarded GIT $5.6 million. MK then renewed most of its motions for judgment as a matter of law under Rule 50(b), and the court denied them again.

MK argues that GIT's supplemental damage disclosure was subject to mandatory exclusion under Federal Rule of Civil Procedure 37(c)(1).2 On appeal, MK makes three arguments based on this alleged discovery violation. Most broadly, MK asserts that GIT's untimely change in damage theory so prejudiced MK's ability to conduct the trial as to require a new trial on all issues. MK also argues that the trial court abused its discretion in denying MK's motion to continue the trial to allow it time to respond to GIT's disclosure. Finally, MK argues that the court abused its discretion in not striking GIT's revised damages claim, thereby suggesting a need to remand for a new trial on damages.

MK has not adequately developed its first two arguments, which seek a new trial on all issues based on the prejudice of GIT's disclosure3 or on the trial court's denial of a continuance.4 As for its third argument, MK has consistently emphasized the prejudice that it suffered in having to assimilate GIT's allegedly new theory and documentation of damages mere weeks before trial. As discussed below, however, this court has determined that it is necessary to vacate the award of damages and remand for a new trial on that issue. This appeal has thus eliminated any time-based prejudice and has essentially mooted any abuse in the court's ruling.5 Cf. Wylie v. Ford Motor Co., 502 F.2d 1292, 1295 (10th Cir.1974) (holding that remand for new trial obviated need to decide whether court had abused discretion in failing to impose discovery sanction). Even if GIT did untimely change its theory or disclose documents, this appeal and remand will have afforded MK ample time to investigate and rebut GIT's damage claims before the retrial.

MK makes but one argument which, if accepted, would require reversal of the liability portion of the judgment. It asserts that the court erred in denying a proposed jury instruction. The proposed instruction specified that, in order to prove a delay "excusable" under the contract, GIT had to show both that something beyond its control delayed part of its work, and that the problem would have delayed completion of the entire project.

1. The Applicable Law and the Instruction Given

The contract's default-termination clause, GP 56, incorporated essentially verbatim the standard federal Default clause for fixed-price construction contracts. See 48 C.F.R. § 52.249-10 (1997). That clause is part of the Federal Acquisition Regulation System (FAR). See 48 C.F.R. pts. 1-99. As incorporated in the contract, the clause allowed MK to terminate GIT if the work had been delayed, not for excusable reasons, but by GIT's lack of diligence:

A. If [GIT] refuses or fails to prosecute the work ... with the diligence that will ensure its completion within the time specified in this Subcontract, including any extension ... [MK] may ... terminate [GIT's] right to proceed....

B. [GIT]'s right to proceed shall not be terminated nor [GIT] charged with damages under this article, if:

(1) The delay in completing the work arises from causes ... beyond the control and without the fault or negligence of [GIT]. [A list of examples follows.]

See 48 C.F.R. § 52.249-10. The clause further provided that a wrongful default-termination would be treated as a termination for convenience:

C. If, after termination ..., it is determined that [GIT] was not in default or that the delay was excusable, the rights and obligations of the parties will be the same as if the termination had been issued for the convenience of the Government.

The contract's Disputes clause, GP 41, provides that "[a]ny substantive issue of law in [litigation concerning the contract] shall be determined in accordance with the body of law applicable to procurement of goods and services by the Government." Accordingly, this court must interpret the above default-termination clause, and the rest of the contract, in light of the regulations and case law governing the interpretation of federal contracts. The controlling regulations have since 1984 been the FAR. See 48 C.F.R. pts. 1-99; see generally John Cibinic, Jr., & Ralph C. Nash, Jr., Administration of Government Contracts 14-22 (3d ed.1995). The relevant caselaw has been developed mainly by the boards of contract appeals found in various federal departments; the Court of Federal Claims and its predecessor trial courts; and the Federal Circuit and its predecessor appellate courts. See generally id. at 22-23, 1239-41, 1311-16.

The sole issue of liability at trial was whether the default-termination was wrongful. By terminating GIT, MK indicated a belief that GIT was so far behind schedule, without adequate excuse, as to show a lack of diligence that made the project's timely completion uncertain. The termination was proper if such a belief was reasonable.6 In challenging the termination, GIT claimed that it had been entitled to extensions of time for various "excusable delays" under GP 56B. The jury's general verdict simply indicated that it found the termination wrongful. That conclusion may have depended on a finding that GIT was entitled to one or more excusable delays. Any error in the instruction defining excusable delays would thus prevent this court from affirming the verdict on liability. See City of Wichita, Kan. v. United States Gypsum Co., 72 F.3d 1491, 1495 (10th Cir.1996) (requiring reversal if jury may have relied on erroneous instruction in reaching verdict, even if such reliance is unlikely).

MK has appealed the court's refusal to give an instruction specifying that excusable delays had to threaten to delay completion of the project as a whole. This court reviews a trial court's refusal to give a particular instruction for abuse of discretion. See, e.g., Webb v. ABF Freight Sys., Inc., 155 F.3d 1230, 1248 (10th Cir.1998). That deferential review is superseded, however, by this court's de novo review of the instructions given to determine whether, in the absence of the refused instruction, they misstated the applicable law. See, e.g., United States v. Wolny, 133 F.3d 758, 765 (10th Cir.1998); Wolfgang v. Mid-America Motorsports, Inc., 111 F.3d 1515, 1526 (10th Cir.1997). This court reviews the instructions as a whole to determine whether they "adequately apprised the jury of the issues and the governing law." Wolny, 133 F.3d at 765.

In this case, the court's instruction # 18, "Contractual Excuse of Delays," told the jury that "[f]or each claim of excusable delay, GIT has the burden to prove that it met the subcontract requirements for excusable delay." The instruction noted that "[t]he subcontract requirements for establishing an excusable delay are explained in instruction 22." Instruction # 22, in turn, concerned "GIT's Counterclaim for Extra Work, Delays and Acceleration--Essential Elements." That instruction told the jury what GIT had to prove in order to recover damages on its delay claims. It required GIT to show that something beyond its control or fault, and unauthorized by the contract, had delayed its work; that it had notified MK; and that MK had "refused to allow GIT the additional time necessary to complete the work."7

MK proposed an instruction listing similar elements of excusable delay. MK's instruction did not require proof that MK had "refused to allow GIT the additional time necessary to complete the work." It did, however, require proof that the delay was a "critical delay." It defined a critical delay as one that "would delay not just the particular activity at issue, but the overall completion date of the Work." The court declined over MK's objection to give this instruction.

MK argues that a contractor is not entitled to an excusable delay unless it proves that the delay was on the project's "critical path," i.e., that it would delay completion of the entire project. MK thus asserts that the court erred "in refusing to instruct the jury on the element of 'criticality.' " GIT responds that MK has not identified any part of the contract or any cases using the terms "critical path" or "criticality" or mandating jury instructions on those terms. Alternatively, GIT claims that the instructions as a whole were sufficient. MK does not claim that the contract expressly refers to the "critical path" or requires that excusable delays affect the project's overall completion. It argues instead that the latter requirement is a well-established rule of federal-contracting law.

MK's description of federal-contracting law is correct. The leading treatise explains that "[a] contractor is not entitled to relief upon the mere occurrence of an event that qualifies as an excusable delay. The contractor must show that the event caused delay to the overall completion of the contract." Cibinic & Nash, supra, at 577 (emphasis omitted); see, e.g., Essential Constr. Co., ASBCA No. 18491, 78-2 B.C.A. (CCH) p 13,314, at 65,122, 1978 WL 2282 (1978); AB-Tech Constr., Inc., VABCA No. 1531, 82-2 B.C.A. (CCH) p 15,897, at 78,823-11, 1982 WL 7201 (1982). A subsection of that treatise entitled "Delay of Overall Completion Required" discusses the case law establishing that a contractor "is not entitled to an excusable delay unless it can prove that the time lost delayed the completion of the job. It is not sufficient to establish that some work was prevented; the work prevented must be work that will delay the overall completion of the job." Cibinic & Nash, supra, at 579 (citations omitted); see generally id. at 579-80, 583-84.

The two cases MK cited below and in its opening brief plainly state and apply the requirement MK sought to include in the jury instructions, albeit without reference to the "critical path." See Essential, 78-2 B.C.A. (CCH), at 65,122; AB-Tech, 82-2 B.C.A. (CCH), at 78,823-11. One of those opinions specifies that "persuasive evidence showing that [the delay] extended the overall completion of the contract work ... is an essential element of the proof required to support a [claim for a] time extension." Essential, 78-2 B.C.A. (CCH), at 65,124 (emphasis added). A third case that MK cites in its reply brief plainly applies that requirement while using "critical path" terminology. See Harrison Western/Franki-Denys (JV), ENG BCA No. 5506, 93-1 B.C.A. (CCH) p 25,406, at 126,586-87, 1992 WL 230234 (1992). In a recent opinion interpreting the default-termination clause at issue in this case, the Armed Services Board of Contract Appeals acknowledged the potentially dispositive importance of the requirement that an excusable delay affect completion of the contract as a whole:

The outcome of this appeal is controlled by the principle that, where a contractor attributes its delayed performance to excusable causes, the contractor bears the burden of proving that time lost due to an excusable delay actually delayed completion of the entire job. Appellant has failed to satisfy that burden....

D.J. Simons Constr. Co., ASBCA No. 41336, 93-1 B.C.A. (CCH) p 25,306, at 126,066, 1992 WL 187585 (1992) (citing Arctic Corner, Inc., ASBCA No. 29405, 88-1 B.C.A. (CCH) p 20,396, at 103,188, 1987 WL 46155 (1987)).

The parties' dispute on this issue has been exacerbated by MK's use of "critical path" terminology in its proposed instruction. "Critical Path Methodology" (CPM) is a term of art for a method of scheduling and administering construction contracts. The Court of Claims has explained that CPM enables contractors performing complex projects to identify a critical path of tasks that must each be completed before work on other tasks can proceed. A delay on the critical path will thus delay the entire project:

Essentially, the critical path method is an efficient way of organizing and scheduling a complex project which consists of numerous interrelated separate small projects. Each subproject is identified and classified as to the duration and precedence of the work. (E.g., one could not carpet an area until the flooring is down and the flooring cannot be completed until the underlying electrical and telephone conduits are installed.) The data is then analyzed, usually by computer, to determine the most efficient schedule for the entire project. Many subprojects may be performed at any time within a given period without any effect on the completion of the entire project. However, some items of work are given no leeway and must be performed on schedule; otherwise, the entire project will be delayed. These latter items of work are on the "critical path." A delay, or acceleration, of work along the critical path will affect the entire project.

Haney v. United States, 230 Ct.Cl. 148, 676 F.2d 584, 595 (1982); see also, e.g., Wilner v. United States, 24 F.3d 1397, 1398 n. 5 (Fed.Cir.1994) (en banc) (" '[O]nly construction work on the critical path had an impact upon the time in which the project was completed.' " (quoting G.M. Shupe, Inc. v. United States, 5 Cl.Ct. 662, 728 (1984))).

Courts often use CPM to resolve disputes over excusable-delay claims. See Cibinic & Nash, supra, at 584. CPM provides a useful, well-developed nomenclature and analytic framework for expert testimony. While CPM has generated a technical terminology, the legal requirement that it is used to analyze is general and commonsensical: a contractor must prove that a delay affected not just an isolated part of a project, but its overall completion. Courts often do not use formal CPM terminology, but simply an informal, CPM-like analysis to determine whether a contractor has met its burden of proof on that general requirement. See id.

MK's proposed instruction comprehensibly conveyed that requirement by explaining that GIT had to prove that

[t]he delay was a "critical delay." A critical delay is one which would delay not just the particular activity at issue, but the overall completion date of the Work. Many activities may be performed on a project at any time without any effect on the completion of the project. A delay in such non-critical activities will not delay the project overall and cannot constitute an excusable delay. Only delays to activities on the critical path--activities with no leeway in the schedule--may give rise to excusable delay.

Contrary to GIT's suggestions, MK has not argued that GIT had to use, or that the court had to instruct the jury to use, formal CPM analysis. The term "critical path" was simply appropriate, though not necessary, jargon in a proposed instruction that, as a whole, comprehensibly stated the general requirement that an excusable delay must "delay the project overall."

MK's proposed instruction thus accurately stated an applicable rule of law which the court's instructions did not expressly state. As a consequence, this court must determine whether the district court's instructions, taken as a whole, nonetheless "correctly stated the governing law and provided the jury with an ample understanding of the issues and applicable standards." Allen v. Minnstar, Inc., 97 F.3d 1365, 1368 (10th Cir.1996). We have stressed that the instructions "must be read and evaluated in their entirety." United States v. Denny, 939 F.2d 1449, 1454 (10th Cir.1991). " '[N]o particular form of words is essential if the instruction as a whole conveys the correct statement of the applicable law.' " Faulkner v. Super Valu Stores, Inc., 3 F.3d 1419, 1424 (10th Cir.1993) (quoting Perrell v. FinanceAmerica Corp., 726 F.2d 654, 656 (10th Cir.1984)).

The court's instruction # 22 enumerated six elements of a compensable delay, and then required GIT to prove that MK refused it the additional time to complete the project that a delay made necessary:

GIT ... must prove each of the following elements ...:

1. GIT was delayed or accelerated in its performance;

2. The delay was beyond the control of GIT;

3. The delay was without the fault or negligence of GIT;

4. [MK's] conduct in causing the delay was not authorized by the subcontract;

5. GIT incurred additional costs as a direct result of the delay;8 and

6. GIT complied with the written notice provisions of the subcontract [or proved an excuse for having failed to do so].

Before you may find that [MK] is liable to GIT under these elements for accelerating or delaying GIT's performance, GIT must prove that [MK] accelerated or delayed GIT's performance or required extra work of GIT, but refused to allow GIT the additional time necessary to complete the work.

As noted above, the court's instruction # 18, defining "excusable delays," incorporated by reference the above part of instruction # 22.

MK's proposed instruction began with a similar list of elements. It differed, though, in that it listed the critical-path requirement as its final element, while it did not refer to "additional time necessary to complete the work":

In order to prove that a given delay ... was excusable under GP 56B, GIT must prove ... that:

1. The delay was unforeseeable.

2. The cause of the delay was beyond the control of GIT.

3. The cause of the delay was without the fault of GIT.

4. The cause of the delay was without the negligence of GIT.

5. GIT gave notice to [MK] of the cause of the delay within 10 days....

6. The delay was a "critical delay." A critical delay is one which would delay not just the particular activity at issue, but the overall completion date of the Work. Many activities may be performed on a project at any time without any effect on the completion of the project. A delay in such non-critical activities will not delay the project overall and cannot constitute an excusable delay. Only delays to activities on the critical path--activities with no leeway in the schedule--may give rise to excusable delay.

After carefully comparing the court's instructions, taken as a whole, with MK's proposed instruction, taken as a whole, this court concludes that the district court's description of the law, while not faultless, was not reversibly erroneous. We rely primarily on the final paragraph of the court's instruction, which required GIT to prove that MK had delayed its performance, yet "refused to allow GIT the additional time necessary to complete the work." If an otherwise-excusable delay did not affect the project's overall completion, the jury could have applied that paragraph, noted that no "additional time" was "necessary to complete the work," and thus concluded that GIT had not established an "excusable delay."

The court's and MK's definitions of excusable delay had similar structures. The court's "additional time" requirement and MK's criticality requirement occupied the same position at the end of that structure. Each followed a list of the basic elements of an excusable delay. In each instruction, those elements defined whether a delay was "excusable" in the common sense of the term, i.e., whether it was beyond GIT's control and not due to GIT's fault or negligence, and whether GIT promptly notified MK, so that the delay would not extend unnecessarily.9 Each instruction proceeded to focus the jury on the conceptually distinct question whether the delay was relevant, i.e., whether it affected GIT's ability to timely complete the project as a whole.

MK's proposed instruction did so by specifying that a delay must "delay not just the particular activity at issue, but the overall completion date of the Work." The court's instruction did so by requiring the jury to find that MK had "refused to allow GIT the additional time necessary to complete the work." To find a delay excusable under MK's instruction, the jury had to find that it would delay "the overall completion date of the Work." Under the court's instruction, the jury had to conclude that a delay made "additional time necessary to complete the work." Each instruction thus required the jury to decide, first, whether GIT should be excused for a delay to a specific task and, second, whether that specific delay should excuse GIT from meeting the project's deadline.

Unlike the instructions given, MK's proposed instruction spelled out at some length, in the same terms as federal-contracting case law, the requirement that a delay affect overall completion. But this court's concern is not whether a district court's instructions are inferior to a proposed instruction. Our concern is to ensure that our review does not leave us with "substantial doubt whether the instructions, considered as a whole, properly guided the jury in its deliberations." Mason v. Oklahoma Turnpike Auth., 115 F.3d 1442, 1454 (10th Cir.1997). These instructions do create some doubt whether the jury was properly guided, but that doubt is not sufficiently substantial to constitute error.

Even if a review of the instructions, read in isolation from the rest of the trial, did leave this court with a substantial doubt, it would then be necessary to determine whether any error prejudiced MK. See Lusby v. T.G. & Y. Stores, Inc., 796 F.2d 1307, 1310 (10th Cir.1986) ("Harmless error analysis normally applies in civil cases ... and it specifically applies to faulty jury instructions.") (citing 28 U.S.C. § 2111; Fed.R.Civ.P. 61). See generally 11 Charles Alan Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice & Procedure § 2886 (2d ed.1995). Instructional error requires reversal " ' "only if the error is determined to have been prejudicial, based on a review of the record as a whole." ' " Shugart v. Central Rural Elec. Coop., 110 F.3d 1501, 1508 (10th Cir.1997) (quoting Wheeler v. John Deere Co., 862 F.2d 1404, 1411 (10th Cir.1988) (quoting Durflinger v. Artiles, 727 F.2d 888, 895 (10th Cir.1984))).

Unfortunately, the parties have offered no guidance to the over-3,000-page record of this complex, three-week trial. This court thus has no idea which parts of the record might bear on prejudice, and no sensible way to review "the record as a whole." That failure leaves us unable to determine whether the error, if any, prejudiced MK.

a. The standard for whether an instruction error was harmless

MK cites Farrell v. Klein Tools, Inc., for the proposition that, if instructions are erroneous, this court must reverse unless we can say with "absolute certainty" that the error did not influence the jury. 866 F.2d 1294, 1301 (10th Cir.1989). MK apparently reads Farrell to relieve it of any obligation to explain how an error may have prejudiced its substantial rights, or to identify any parts of the record that might show such prejudice.

Farrell, however, is inapposite. It concerned not an instruction that misstated the law, but an instruction on a theory unsupported by sufficient evidence. See id. at 1297-98. The improper theory was an affirmative defense, and the jury returned a general verdict for defendant. See id. at 1298. On appeal, this court noted the general rule in such cases: if one of several issues submitted to a jury should not have been, a general verdict cannot stand. See id. at 1299. After noting exceptions to the rule, we held that we were "bound, at least in this context, by our two most recent [general-verdict] cases, both of which strictly imposed the general rule." Id. at 1299-1300 (following McMurray v. Deere & Co., 858 F.2d 1436, 1444 (10th Cir.1988) (not applying harmless-error review in case involving general verdict and affirmative defense unwarranted by evidence); Smith v. FMC Corp., 754 F.2d 873, 876-77 (10th Cir.1985) (same)). We held that those precedents "leave[ ] no room for harmless error analysis." Id. at 1300.

In this case, the court allegedly erred not by instructing on an unwarranted theory, but by misstating the law on a viable theory. Harmless-error analysis does apply in such cases.10 When sufficient evidence supports each theory submitted to a jury, an error in stating the law that governs one of the theories may be harmless. See, e.g., Allen, 97 F.3d at 1372-73; Mehojah v. Drummond, 56 F.3d 1213, 1215-16 (10th Cir.1995).

MK also quotes an opinion saying that this court " 'must reverse if the jury might have based its verdict on the erroneously given instruction.' " Coleman v. B-G Maint. Mgmt. of Colo., Inc., 108 F.3d 1199, 1202 (10th Cir.1997) (quoting City of Wichita, Kan. v. United States Gypsum Co., 72 F.3d 1491, 1495 (10th Cir.1996)). Three prior opinions also state that rule;11 two specify that we must reverse "[e]ven if that possibility is 'very unlikely.' " City of Wichita, 72 F.3d at 1495 (quoting Adams-Arapahoe Jt. Sch. Dist. No. 28-J v. Continental Ins. Co., 891 F.2d 772, 780 (10th Cir.1989) (quoting Farrell, 866 F.2d at 1301)). An earlier opinion, however, sets a much less strict standard, requiring affirmance unless it is fifty percent likely that an error was prejudicial. See U.S. Indus., Inc. v. Touche Ross & Co., 854 F.2d 1223, 1253 n. 39 (10th Cir.1988) [USI ] (in instruction-error case, noting circuit split on "standard of review ... in gauging the effect of an error in a civil case" and adopting rule that error is harmless if appellant's "substantial rights ... were more probably than not unaffected").

This case does not require this court to reconcile those precedents. None says that harmless-error analysis does not apply.12 They conflict on a substantive question: how likely must it be that an instructional error was prejudicial for this court to reverse? They do not concern the procedural question posed by this case: can this court determine the chance that an error was prejudicial if the parties provide no relevant appendix or arguments? Even assuming that a "very unlikely" chance compels reversal, our precedent setting that standard did not obviate the rules of appellate procedure.13 Without a sufficient appendix, or any arguments citing the record, this court cannot determine whether the chance that an error prejudiced MK is fat, slim, or nil.

b. The parties' obligations to provide an appendix and arguments

As appellant, MK was bound to "file an appendix [of record excerpts] sufficient for consideration and determination of the issues on appeal." 10th Cir. R. 30.1.114; see also 10th Cir. R. 10.1.1; Fed.R.App.P. 10(b)(1) & 30(a). MK's appendix contains a few hundred pages of trial excerpts, but less than 10% of the entire 3,123-page transcript.15 GIT's supplemental appendix adds a few hundred more.

This court has held "on a number of occasions and in a variety of settings that the lack of a required transcript leaves us with no alternative but to affirm." McGinnis v. Gustafson, 978 F.2d 1199, 1201 (10th Cir.1992). An appellant's failure to provide a necessary transcript entails more than mere "noncompliance with some useful but nonessential procedural admonition"; it "raises an effective barrier to informed, substantive appellate review." Id. Such failures have kept this court from reviewing alleged instruction errors. See Dikeman v. National Educators, Inc., 81 F.3d 949, 955 (10th Cir.1996); King v. Unocal Corp., 58 F.3d 586, 587-88 (10th Cir.1995). King began its analysis by stressing this court's duty to " ' "consider all the jury heard." ' " 58 F.3d at 587 (quoting Considine v. Newspaper Agency Corp., 43 F.3d 1349, 1365 (10th Cir.1994) (quoting Resolution Trust Corp. v. Stone, 998 F.2d 1534, 1549 (1993))). As a result, this court explained, "we must have a proper record before us" in order to review an alleged error in admitting or excluding a jury instruction. Id. In Allen, this court held in the alternative that an instruction had not been prejudicial, and that appellant's failure to provide a trial transcript left no option but to affirm:

Although plaintiff speculates the verdict would have been different had the [challenged instruction] been different, we have no basis for reaching that conclusion without first reviewing the evidence that was admitted at trial. On the record presented, we are unable to conclude that plaintiff was prejudiced by the [assumed error in the instruction].

97 F.3d at 1372-73 (citing 10th Cir. R. 10.1.1).16

Precedent would thus justify affirmance based on MK's failure to provide a sufficient appendix. Even if this court ignored that failure and sua sponte obtained the 3,123-page transcript, we would then confront the parties' shared failure to make arguments citing that transcript or even providing general guidance as to which parts of it might bear on the issue of prejudice. Each party on appeal must make arguments including "contentions and the reasons for them, with citations to the ... parts of the record on which [it] relies." Fed. R.App. P. 28(a)(9) & (b). In this case, neither party has made any argument that might guide this court in reviewing the voluminous transcript, were we inclined to remedy MK's failure to provide it.17

Without citing the record, MK deems it "impossible to determine if the jury would have treated GIT's claims differently had the criticality requirement been imposed." Harmless-error analysis, however, requires this court to address not the unanswerable question whether the jury would have treated GIT's claims differently, but the eminently answerable question whether the jury might have treated GIT's claims differently. MK's failure to address that question is complete: it never argues that GIT's "excusable delays" did not in fact cause overall delay. Nor does it assert that it ever did so at trial, i.e., that it ever argued one of GIT's claimed delays was not "excusable" because the delay had not affected overall completion. It does claim that "issue[s] of delay and timely completion" were central to the trial, and argues that caselaw required GIT to show each individual delay caused an overall delay. But it never asserts a link between that legal argument and the outcome in this case: it says nothing inconsistent with an assumption that each delay claimed by GIT did in fact contribute to overall delay, and that any error in omitting the issue from the instructions was purely technical.

MK thus has not made a basic "contention" essential to any conclusion that the error prejudiced its substantial rights. Fed. R.App. P. 28(a)(9). Even were this court to infer from MK's brief a bare contention that some of GIT's claimed delays did not cause overall delay, the brief would still lack "reasons" for that implicit contention and "citations to the ... parts of the record on which [it] relies." Id. MK's failure to provide a sufficient appendix, and both parties' failure to make arguments citing the record, leave this court only two alternatives: to speculate baselessly about prejudice, or to affirm.

c. It is unnecessary to decide whether either party bears a burden of persuasion regarding prejudice

In rejecting the path of speculation, this court need not enter the bog surrounding the issue whether appellants in civil cases bear a burden of persuasion on the issue of prejudice. See Allen, 97 F.3d at 1374-75 (Lucero, J., concurring). We do observe that an appellant must provide a sufficient record on appeal, and that both parties must make arguments which enable us to meaningfully review the record. To note those duties, however, is not to say that, if the parties fulfill them, this court will then allocate to one party or the other an appellate burden of persuasion.18 This appeal turns not on any such rule, but on the answer to a simple, narrow question: if this court is unable, for whatever reasons, to determine whether an error was prejudicial or harmless, which party must lose? In ordinary civil cases, we conclude, that party is the appellant. To explain that holding's narrow compass, we must address the Supreme Court's recent discussion of burdens of persuasion in harmless-error review.

In habeas cases, the Court held, an appellate court must decide whether errors were harmless or prejudicial " 'without benefit of such aids as presumptions or allocated burdens of proof that expedite fact-finding at the trial.' " O'Neal v. McAninch, 513 U.S. 432, 436-37, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995) (quoting Roger J. Traynor, The Riddle of Harmless Error 26 (1970)). Part of O'Neal 's holding depends on concerns unique to habeas cases, but its rejection of the use of burdens in harmless-error review apparently does not. See id. at 442, 115 S.Ct. 992 ( "[E]ven if ... we were to assume that the civil standard for judging harmlessness applies to habeas proceedings ... it would make no difference.").

This court has not addressed whether O'Neal reversed our longstanding rule that appellants in ordinary civil cases bear the burden of showing that any errors prejudiced their substantial rights. See, e.g., K-B Trucking Co. v. Riss Int'l Corp., 763 F.2d 1148, 1156 (10th Cir.1985); Scritchfield v. Kennedy, 103 F.2d 467, 474 (10th Cir.1939). Nor does this case require that we do so. In Scritchfield, a 1939 instructional-error case, this court held that the original harmless-error statute imposed on appellants the burden of showing prejudice. See 103 F.2d at 474.19 Four years later, the Supreme Court stated the same rule: "[the party] who seeks to have a judgment set aside because of an erroneous ruling carries the burden of showing that prejudice resulted." Palmer v. Hoffman, 318 U.S. 109, 116, 63 S.Ct. 477, 87 L.Ed. 645 (1943). In seven opinions between 1958 and 1985, this court reiterated the rule of Scritchfield and Palmer.20

In 1986, however, a panel said, in dictum,21 that "[i]n harmless error analysis the beneficiary of the error (usually the appellee) has the burden to show that the error almost surely did not affect the outcome." Lusby, 796 F.2d at 1312 n. 4. The panel cited no authority for this proposition, which contradicted circuit and Supreme Court precedent. Its dictum would have been benign, though, had a 1992 panel not converted it to a holding. See Board of County Comm'rs of San Juan County v. Liberty Group, 965 F.2d 879, 884 (10th Cir.1992) [San Juan ] (applying rule that beneficiary of instruction error "has the burden of showing that the error 'almost surely did not affect the outcome' ") (quoting id.). San Juan relied solely on Lusby; it did not cite any valid authority or address the contrary precedent of this circuit and the Supreme Court. This court has not applied the rule since.

It is well-settled that " 'when faced with an intra-circuit conflict, a panel should follow earlier, settled precedent over a subsequent deviation therefrom.' " Clymore v. United States, 164 F.3d 569, 573 n. 5 (10th Cir.1999) (quoting Haynes v. Williams, 88 F.3d 898, 900 n. 4 (10th Cir.1996)). Lusby 's and San Juan 's comments assigning appellees the burden of establishing harmlessness are thus not good law. They had not, as of 1992, displaced the rule of Scritchfield.

This court need not answer the distinct question whether the 1995 O'Neal opinion did overrule Scritchfield, and thus relieve appellants of the burden of proving prejudice. Even assuming that O'Neal governs, we must affirm. O'Neal eliminates the use of a "burden of proof" to review prejudice, but it does not eliminate an appellant's duty to provide an appendix, or the parties' shared duty to make arguments citing the record.

In rejecting the use of burdens or presumptions, the O'Neal Court stressed the need for thorough, conscientious record review, and suggested that invoking a "burden of proof" may short-cut that review. See id. at 435-37, 115 S.Ct. 992. It emphasized that cases of "virtual equipoise," in which a court may properly resort to a default rule, are rare. Id. at 435, 115 S.Ct. 992. It also quoted Justice Traynor's rejection of decisionmaking "aids":

"Whether or not counsel are helpful, it is still the responsibility of the ... court, once it concludes there was error, to determine whether the error affected the judgment. It must do so without benefit of such aids as presumptions or allocated burdens of proof that expedite fact-finding at the trial."

Id. at 437, 115 S.Ct. 992 (quoting Traynor, supra, at 26). O'Neal requires courts to review the record and reach a decision on prejudice whenever possible, instead of simply identifying some appeals as "close cases" and invoking a "burden of proof" to resolve them.

O'Neal thus makes it even more crucial that appellants provide sufficient appendices, and both parties make arguments supported by sufficient citations to enable meaningful review. This court has repeatedly declined to "sift through" entire records to find support for arguments devoid of citations. See, e.g., United States v. Rodriguez-Aguirre, 108 F.3d 1228, 1237-38 n. 8 (10th Cir.1997); Harolds Stores, Inc. v. Dillard Dep't Stores, Inc., 82 F.3d 1533, 1540 n. 3 (10th Cir.1996). By eschewing the shortcut of a "burden of proof," O'Neal only increases this court's need for sufficient guidance from parties to enable us to review voluminous records intelligently, rather than sift through them aimlessly.22

If the appendices and arguments leave this court unable to determine whether an error was prejudicial, we must resolve the appeal by invoking a default rule imposing the risk of equipoise on one party. To do so is not to impose a "burden of proof" on that party, and is fully consistent with O'Neal. That opinion governs cases in which a court truly cannot decide if an error was harmless and must resort to a default rule. See id. at 435, 437, 115 S.Ct. 992. While the Court intended to narrow that class of cases by rejecting the use of "burdens of proof," it recognized that such cases will still arise. When they arise in the habeas context, O'Neal holds, the error must be deemed prejudicial. Id. at 436, 115 S.Ct. 992. While eliminating any "burden of proof," O'Neal thus did not shrink from "placing the risk of doubt" on one party. Id. at 439, 115 S.Ct. 992.

To do so was not optional. If a court must resolve each appeal, and cannot do so by flipping coins, then it must place the "risk of doubt" on one party or the other. That risk becomes relevant if the court cannot ascertain whether an error was harmless. It may be unable to do so because, as envisioned in O'Neal, the parties provide an adequate record and arguments, but the court's review leaves it in "virtual equipoise." Or the court may be unable to do so because, as here, the parties' failure to provide such material prevents it from commencing any review.

O'Neal assigned the risk of doubt in habeas appeals to the State, noting the writ's purpose and the uniquely uneven stakes in those nominally civil cases. See id. at 442-43, 115 S.Ct. 992. If O'Neal applies to non-habeas cases, this court concludes, the risk of doubt must rest with the appellant. That rule accords with the appellant's responsibility for the appendix, and its status as the party seeking to change the status quo. More importantly, it preserves that part of our longstanding Scritchfield rule which is not inconsistent with O'Neal. The O'Neal Court rejected the use of decisionmaking shortcuts like a burden of proof, but acknowledged the need to place the risk of doubt on one party. Our Scritchfield rule, meanwhile, has two components: it uses a burden of proof, and it puts the risk of doubt on appellants. Only the first is inconsistent with O'Neal.

Accordingly, even if the instructions were erroneous, MK's failure to provide a sufficient appendix, and the parties' failure to make arguments affording any guidance to the voluminous record, would require us to affirm.

MK challenges the sufficiency of the evidence of GIT's claimed damages in general, and of three specific categories of damages: attorney's fees, equitable adjustments, and claims on behalf of subs. It also argues that the contract barred GIT as a matter of law from asserting the latter claims without having first settled with and paid its subs.

MK makes two overlapping challenges to GIT's damage evidence. The first is that GIT ignored the contract provision detailing the damages to which it was entitled, and that the court allowed it instead to submit evidence supporting "a common-law measure of damages." The second is that GIT did not offer sufficient evidence of a "causal connection" between MK's acts and the majority of the damages GIT claimed in the exhibits submitted to the jury.

As noted above, the contract required the court to determine all substantive legal issues in accord with the body of government-procurement law. In addition, GP 55H, part of the contract's termination-for-convenience clause, provides that "[t]he cost principles and procedures of FAR, Part 31 ... shall govern all costs claimed, agreed to, or determined under this article." Under those principles, as interpreted in the decisions of the federal courts and boards of contract appeals which handle most federal-contracting disputes, GIT bore the burden of proving damages in accord with the contract " 'with sufficient certainty so that the determination of the amount of damages will be more than mere speculation.' " Lisbon Contractors, Inc. v. United States, 828 F.2d 759, 767 (Fed.Cir.1987) (quoting Willems Indus. v. United States, 155 Ct.Cl. 360, 295 F.2d 822, 831 (1961)). MK was not obliged to present evidence attacking items of damages if GIT had not made a prima facie showing that the items were properly included in its claim. See id. In addition, FAR part 31 required GIT to prove that incurred costs were "reasonable." See 48 C.F.R. §§ 31.105(a)-(b), .201-3, .204(a).

For purposes of analysis, it is appropriate to divide GIT's total claim of roughly $11.35 million into equitable adjustments ($3 million), lower-tier subcontractor claims ($3.7 million), attorney's fees ($1.35 million), unpaid contract work and post-termination equipment expenses and work ($1.9 million),23 and other ($1.4 million). (All figures include overhead, profit, and bond.) As detailed below, GIT offered no admissible evidence whatsoever to support its claim of roughly $1.35 million for attorney's fees.24 GIT also did not offer sufficient evidence to support its claims for roughly $3 million in equitable adjustments to the contract. Nor did it offer sufficient evidence of most if not all of the roughly $3.7 million in damages it claimed on behalf of its subs.25

a. GIT's evidence was not based only on a common-law damages theory

MK's broader argument is that GIT premised all of its damage evidence on a common-law theory of damages fundamentally inconsistent with the contract and the FAR. MK acknowledges that the district court rejected GIT's common-law theory and properly instructed the jury to award damages in accord with the contract. MK argues, however, that the court nonetheless admitted evidence inconsistent with the contract and did not require evidence consistent with it. MK is incorrect. GIT did seek a common-law measure of damages rather than the damages provided by the contract.26 It did so, however, as part of its claim that the termination had been not merely wrongful but in bad faith. If GIT had proved bad faith, it could have recovered full common-law damages for breach of contract. See Apex Int'l Mgmt. Servs., Inc., ASBCA No. 38087, 94-2 B.C.A. (CCH) p 26,842, at 133,550, 1994 WL 117148 (1994); see also Section of Public Contract Law, American Bar Association, Government Contract Law: The Deskbook for Procurement Professionals 325 (1995) [hereinafter Gov't Contract Deskbook ]. The court, however, declined to submit the bad-faith claim to the jury. It ultimately required GIT to delete anticipated profits from its claim, and it instructed the jury on damages recoverable under the termination-for-convenience clause. MK identifies several other allegedly improper damage items, but, with one minor, arguable exception,27 no actually improper item went to the jury.

b. GIT did not present sufficient evidence of the causation and reasonableness of at least two of its categories of claimed damages

MK frames its second sufficiency-of-the-evidence argument in terms of the common-law requirement of "causation for breach of contract damages." Under the contract, however, and the federal contracting law that it incorporates, GIT did not have to prove an entitlement to common-law breach-of-contract damages. It had to prove an entitlement to equitable adjustments to the contract price.28 See generally Cibinic & Nash, supra, at 430, 671, 692 (explaining primacy of equitable-adjustment analysis (citing, e.g., Johnson & Son Erectors, ASBCA No. 24564, 81-1 B.C.A. (CCH) p 15,082, at 74,599, 1981 WL 7060 (1981), aff'd, 231 Ct.Cl. 753, 1982 WL 1441 (1982))).

An equitable adjustment is a change in contract price. It compensates a contractor for increased costs reasonably incurred because the government (or, in this case, MK) increased the amount or difficulty of work required by the contract, or delayed or accelerated that work. See id. at 669-73 (citing, e.g., Modern Foods, Inc., ASBCA No. 2090, 57-1 B.C.A. (CCH) p 1229, at 3544, 1957 WL 4960 (1957); Celesco Indus., Inc., ASBCA No. 22251, 79-1 B.C.A. (CCH) p 13,604, at 66,683, 1978 WL 2148 (1978) (describing equitable adjustment as "the difference between the reasonable cost of performing without the change ... and the reasonable cost of performing with the change")). Some equitable adjustments are for work added by formal change orders. The contract included a standard clause allowing such orders.29 Other equitable adjustments result from "constructive changes," which occur when the government does something to increase a contractor's costs without issuing a formal change order. See id. at 381, 399, 408-09, 429-35; see also 48 C.F.R. § 52.243-4(b), (d).

If a contractor thinks that the government has caused a constructive change, it must ask the government to equitably adjust the contract. The contractor must show either its estimated costs, if it requests the adjustment before doing the additional work, or its incurred costs, if it requests the adjustment after having completed the work. See Cibinic & Nash, supra, at 669-71. The contractor may convince the officer administering the contract to adjust the contract during performance. If not, it must convince an adjudicator to award an adjustment after the contract is completed or terminated. See id.

To prove that it is entitled to an equitable adjustment, a contractor must show liability, causation, and injury. See Wunderlich Contracting Co. v. United States, 173 Ct.Cl. 180, 351 F.2d 956, 968 (1965). It must prove that the government somehow delayed, accelerated, augmented, or complicated the work, and thereby caused the contractor to incur specific additional costs. See id. at 969. See generally Cibinic & Nash, supra, at 429-35, 697-702. The contractor must not only prove that the government specifically caused its increased costs, but must prove that those costs were reasonable, allowable, and allocable to the contract. See, e.g., McDonnell Douglas Corp. v. United States, 40 Fed.Cl. 529, 536 (1998) (citing 48 C.F.R. (FAR) §§ 31.201-1 to 201-4 (requiring and defining allowability, reasonableness, and allocability)). See generally Cibinic & Nash, supra, at 697-702. The contractor bears the burden of proof on all of those factors. See McDonnell Douglas, 40 Fed.Cl. at 536; see also Lisbon Contractors, 828 F.2d at 767.30

GIT presented very little or no evidence of how MK's actions specifically caused GIT to incur the costs claimed in its damage exhibits. Nor did it present evidence that those costs were reasonable.31

i. Attorney's fees

The court erred in submitting GIT's $1.35 million claim for outside attorneys' and consultants' fees ("attorney's fees") to the jury. GIT presented no admissible evidence whatsoever of such fees. The court apparently made a simple error of omission at the end of this long, complex trial. During the trial, as detailed below, the court properly ordered GIT to remove the claim for attorney's fees from its damage exhibit. Despite MK's prompting, however, the court ultimately failed to require GIT to do so. On appeal, MK has demonstrated GIT's complete failure to document its attorney's fees. GIT has not meaningfully contested that failure.

GIT's final damage exhibit claimed roughly $1.55 million for "equitable adjustment costs to date." Kip Cooper, GIT's damage witness, testified that roughly $450,000 thereof was for time spent by GIT employees to prepare GIT's claim. The rest was for the fees of "outside consultants and attorneys." With overhead, profit, and bond, that amount totaled roughly $1.35 million.

MK objected that GIT had introduced no evidence of the fees, i.e., invoices from the outside attorneys. Both Cooper and GIT's counsel promptly conceded that the invoices were not in evidence. The court declined to let Cooper testify as to the amount incurred, observing that attorney's fees "have to be proved like any other item of damage. It requires some more proof than this man saying that's what was paid." The court further noted that if, indeed, none of GIT's evidence substantiated its claimed fees, then the fees "will be deleted from that exhibit and from your claim.... I'm not going to let you claim specific expenditures that were not supported in the documents." GIT accepted this ruling without protest.

Soon thereafter, GIT tried to introduce the invoices into evidence. The court asked why GIT had not provided them until mid-trial. "We had to go out there and just collect them up," GIT's counsel replied. The court ruled that "[i]f it hasn't been disclosed until the middle of trial, it can't be used. I don't hear anything from you that would justify [GIT's] late action." GIT accepted this ruling without protest.

GIT later eliminated its claims for anticipated profits and prejudgment interest from its damage exhibit, pursuant to the court's rulings on those issues. But GIT did not reduce its claim for "equitable adjustment costs" to eliminate the disallowed attorney's fees.

During the jury-instruction conference the next day, MK moved to exclude the fee claim. MK noted that the court had not admitted any evidence of the fees. The court recalled having "excluded something" and asked GIT what it was. This provoked a confused exchange.32 GIT's counsel forthrightly acknowledged that it was "the outside claim preparation costs for attorneys" that the court had excluded. MK, however, then made an additional argument, which apparently distracted the court. The court then misspoke, saying, "I'm not going to exclude the claim for attorneys' fees.... [J]ust the claim for attorneys' fees is supported apparently in admitted documents...."

MK next moved for judgment as a matter of law (JMOL) under Rule 50(a). The motion reminded the court of its refusal to admit the invoices into evidence and asked it to instruct the jury that there was insufficient evidence to support an award of fees. The court orally denied the motion without analysis or comment.

MK has summarized the parts of the record revealing the court's failure to enforce its order to exclude the attorney's-fee claim. GIT never objected to that order. In reply, GIT has not pointed to any contradictory passage in the record.33

GIT's main argument on appeal is that MK waived the issue by failing to raise it in a post-verdict Rule 50(b) motion for JMOL, despite having raised it in a pre-verdict Rule 50(a) motion. GIT is wrong. MK's failure to renew the attorney's-fees issue in its Rule 50(b) motion does preclude this court on appeal from ordering the entry of judgment in MK's favor on that issue. See Cone v. West Va. Pulp & Paper Co., 330 U.S. 212, 217-18, 67 S.Ct. 752, 91 L.Ed. 849 (1947); Fortier v. Dona Anna Plaza Partners, 747 F.2d 1324, 1334 (10th Cir.1984); Hansen v. Vidal, 237 F.2d 453, 454 (10th Cir.1956). Under Tenth Circuit precedent, however, that failure does not bar MK from appealing the issue of the sufficiency of the evidence to support the attorney's-fees claim. See Marshall v. El Paso Natural Gas Co., 874 F.2d 1373, 1385 (10th Cir.1989); see also Orman v. Standard Constr. & Remodeling Co., No. 92-6323, 1993 WL 262599, at * 1 & n. 1 (10th Cir.1993) (unpublished disposition) (specifying that failure to move for JMOL under Rule 50(a) precludes review of sufficiency of evidence, but that only effect of failure to renew motion under Rule 50(b) is to limit relief to new trial).34

ii. Equitable adjustments

This court further concludes that GIT did not present sufficient evidence to support its claims for roughly $3 million in equitable adjustments. MK has demonstrated on appeal that GIT presented virtually no evidence to show how MK's misdeeds or omissions caused the specific costs which GIT claimed as equitable adjustments. Nor did GIT make any showing that those costs were reasonable and allowable.

GIT summarized its damage claims in several tables. It supported these tables with six boxes of invoices, petty cash records, canceled checks, payroll records, and the like. GIT's final master damage summary, which is substantially reproduced in the margin,35 listed eight "equitable adjustment" claims, which totaled roughly $3 million. It also included a ninth claim, for "Extra Compensation: Wrongful Termination Damages," of roughly $4.7 million. GIT submitted a second-level breakdown for each of its nine claims. The breakdowns divided each claim into two to nine items.

GIT's lone damage witness, Kip Cooper, directed GIT's accounting department and oversaw the company's financing. His main function as a witness was to sponsor GIT's damage exhibits. He very briefly explained what each of the nine breakdowns comprised, and how the summary related to the breakdowns. He also summarized the types of supporting documents for each category of damages, and laid a foundation for the documentation's admission into evidence by testifying that GIT had created the records in the ordinary course of business.

Cross-examination established that Cooper had not been directly involved in deciding to allocate to the project the costs assembled in the six boxes of documentation. Nor had Cooper been involved in evaluating how MK had caused GIT to incur those costs, or whether they were reasonable. GIT personnel on site in Colorado, rather than in the Pennsylvania office where Cooper works, had apparently made those decisions. Cooper said at one point, "I cannot tell you what was going through their heads out there at the site, why this was charged to the project." And later, in discussing an item challenged by MK, he said, "I'm here to tell you it's a direct job cost. I don't know the reasoning behind it."

GIT thus presented only one witness to support a roughly $3 million claim for eight equitable adjustments, and that witness lacked direct knowledge of the project. He did not know, in particular, how MK's alleged errors and omissions might have caused GIT to incur increased costs not contemplated in the contract. Beyond Cooper's foundational testimony, GIT presented only a damage summary, a two to nine-item breakdown for each equitable adjustment, and six banker's boxes of time sheets, checks, cash logs, invoices, and similar raw material. It simply did not make any showing of how MK, by delaying or complicating or increasing the work required by the contract, caused GIT to incur specific costs. Nor did it show why those costs were reasonable. GIT has not met its burden of presenting "specific, persuasive evidence or analysis demonstrating how any government action ... caused [its] overruns." Hoffman Constr. Co. of Ore. v. United States, 40 Fed.Cl. 184, 201 (Fed.Cl.1998), aff'd in relevant part, No. 98-5075, 1999 WL 37411, at * 1 (Fed.Cir.1999). "A contractor must present more than general, unsubstantiated pronouncements from its own witnesses that various acts of the government caused ... overruns." Id. (citing Luria Bros. & Co. v. United States, 177 Ct.Cl. 676, 369 F.2d 701, 713 (1966)).

GIT's attempt to rebut MK's argument on appeal is inadequate. GIT notes MK's "repeated admissions" of its "failure to pay GIT for the contract work and extra work." GIT cites parts of the record that evince MK's refusal to pay GIT for some completed work under the contract and for some additional work that MK ordered after it had terminated GIT. The evidence to which GIT points, however, establishes at most that MK caused GIT some damages by refusing to pay it for completed contract work and some post-termination work. Unpaid contract work and unpaid post-termination work, however, comprised only about $1.25 million of GIT's total claim of over $11 million. MK's failure to pay GIT for some of its work under the contract has nothing to do with whether MK caused GIT to incur roughly $3 million in additional costs unanticipated by the contract but recoverable as equitable adjustments. MK allegedly caused GIT to incur those costs not by refusing to pay it for work, but by delaying the work or misinterpreting the specifications. Similarly, MK's refusal to pay GIT for completed work has nothing to do with whether GIT proved that the additional costs for which GIT claimed equitable adjustments were reasonable.

GIT repeats its error in attacking another of MK's arguments. MK asserts that some of GIT's increased costs were due to mistakes made by GIT and its subs, for which MK was not responsible. GIT responds that MK "simply ignores the overwhelming evidence demonstrating that MK knowingly failed to pay GIT." But again, this court may assume that GIT suffered some damages because MK refused to pay for completed work. That does not mean that the additional costs for which GIT sought equitable adjustments were not in part caused by GIT itself, or by its subs.36

2. The Contract Did Not Allow GIT to Recover on Behalf of Its Subs Without Having First Settled Their Claims

The contract allowed MK to terminate GIT either for convenience or default. If MK terminated GIT for default and a court found the termination wrongful, then the termination would convert to one for convenience. GP 55, the standard termination-for-convenience clause for fixed-price construction contracts, established MK's and GIT's duties after such a termination. See 48 C.F.R. § 52.249-2. The clause obliged GIT to terminate its subcontracts, entertain termination-settlement proposals from its subs, and settle their claims. GIT would then make its own termination-settlement proposal to MK, and they would negotiate a settlement. GIT's proposal to MK would presumably include the costs it had incurred in settling and paying its subs' claims.

Because MK terminated GIT for default, however, the parties did not follow that procedure. GIT never settled with or paid three of its four subs. GP 55F(1) set forth MK's obligations if it terminated GIT for convenience but then could not reach a settlement with GIT. Among scenarios contemplated by the contract, that is the closest analogue to what occurred. For work performed before termination, GP 55F(1) obliged MK to pay GIT "(a) [t]he cost of this work; [and] (b) [t]he cost of settling and paying termination settlement proposals under terminated subcontracts that are properly chargeable to the terminated portion of the [contract] if not included in subparagraph F(1)(a) above." MK argues that GP 55F(1)(b) bars GIT from recovering on its subs' claims unless it has "settled and paid" those claims.37

Two decisions by the Armed Services Board of Contract Appeals hold that contract provisions like that at issue here require settlement. One says that, "[a]bsent either actual payment or binding settlement agreements, the vendors' termination claims are not allowable in the termination settlement." Murdock Mach. & Eng. Co. of Utah, ASBCA No. 42891, 93-1 B.C.A. (CCH) p 25,329, at 126,195, 1992 WL 206519 (1992), aff'd in relevant part sub nom. Dalton v. Murdock Mach. & Eng. Co. of Utah, No. 93-1440, 1994 WL 51573, at * 1 (Fed.Cir.1994) (unpublished disposition). The other quotes a predecessor of the clause at issue here and reads it to require either payment or settlement:

[T]he Termination for Convenience clause ... permits recovery by appellant of "the cost of settling and paying claims arising out of the termination of work under subcontracts or orders...." In the absence of any evidence that appellant either settled or paid the claims of the two subcontractors, we deny the proposed amount representing settlement with the subcontractors in its entirety.

Atlantic, Gulf & Pac. Co. of Manila, ASBCA No. 13533, 72-1 B.C.A. (CCH) p 9415, at 43,745, 1972 WL 1390 (1972).

A recent Court of Federal Claims opinion necessarily implies that a contractor who seeks to recover on its subs' claims must have settled, but need not have paid, those claims. See McDonnell Douglas, 40 Fed.Cl. at 555 & n. 18. The court approvingly quoted part of the Government's brief. The passage described unsettled subcontractor claims as non-incurred, contingent costs, which the FAR disallows:

"The Government challenges plaintiffs' claim for approximately $26 million in contingent costs not yet incurred. These contingent costs are primarily for program security costs, settlement expenses, and settlement with remaining subcontractors. These costs are unallowable. 'Pursuant to FAR [ § ] 52.249-2. Termination for Convenience of the Government (Fixed-Price) the contractor may recover costs incurred in the performance of the work.' Sterling Millwrights, Inc. v. United States, 26 Cl.Ct. 49, 112 (1992) (emphasis added). Moreover, cost principles expressly prohibit, in FAR Part 31, recovery of non-incurred, contingent costs. See FAR [ § ] 31.205-7(a) and 7(b)."

Id. at 555 n. 18 (first emphasis added). This part of the opinion, titled "Yet To Be Incurred Costs," went on to note that "McDonnell Douglas has reached a settlement with [one sub] since the time of trial. Thus, $5 million [the amount of the settlement] is no longer a cost yet to be incurred." Id. at 555. The court then proceeded to disallow a claim "for four unsettled subcontractors." Id.

The court's dispositions, and its quotation of the Government's brief, necessarily imply that pass-through claims under standard termination-for-convenience clauses require settlement, but not payment. The court disallowed claims on behalf of "unsettled subcontractors." It allowed a claim, however, based on the contractor's post-trial "settlement" with a sub. It did not indicate whether the contractor had paid the latter sub, and did not suggest that it mattered.

Such a rule accords with the FAR's definition of a contingency as "a possible future event or condition ..., the outcome of which is indeterminable at the present time." 48 C.F.R. (FAR) § 31.205-7(a). That section of the FAR, which is among those expressly incorporated by the contract, provides that "[c]osts for contingencies are generally unallowable for historical costing purposes because such costing deals with costs incurred and recorded on the contractor's books." Id. at § 31.205-7(b). Once a contractor has entered a binding settlement agreement with a sub, the cost to the contractor of the sub's claim is no longer contingent. It is not necessary to have paid a cost in order to have "incurred and recorded" it. A settlement agreement, moreover, greatly reduces the risk that a contractor will recover from the government on a sub's behalf but then not pay the sub.

A settled-but-not-paid requirement, finally, also accords with the long-established rule that a contractor need not pay its sub before making a claim on its behalf. See George Leary Constr. Co. v. United States, 63 Ct.Cl. 206, 223, 1927 WL 2960 (1927) ("It is easy to forecast the financial ruin of a Government contractor if the rule is to be established that he may not receive amounts due from the Government under his contract until he establishes ... that he has paid his subcontractor all he owes him."); see also Severin v. United States, 99 Ct.Cl. 435, 443, 1943 WL 4198 (Ct.Cl.1943) (holding that, if contractor is liable to sub, "that liability, though not yet satisfied by payment, might well constitute actual damages to [the contractor], and sustain their suit" under rule that contractor may only sue to recover its own damages). Leary and Severin strongly suggest that, while the termination-for-convenience clause in this case may require GIT to have settled its subs' claims, this court should be leery of interpreting that clause to require that GIT have settled and paid those claims.

MK has never argued that GIT had to have settled or paid its subs' claims. GIT's brief thus responds only to the argument that it must have "settled and paid" the claims. GIT makes two counterarguments. One is that the Severin doctrine allows a prime contractor to sue on its sub's behalf so long as the prime "has reimbursed ... or remains liable " to the sub. J.L. Simmons Co. v. United States, 158 Ct.Cl. 393, 304 F.2d 886, 888-89 (1962) (discussing Severin ) (emphasis added). The second is that, even if GIT cannot recover on the subs' unsettled claims under subparagraph F(1)(b), it can recover those claims as part of "the cost of this work" under subparagraph F(1)(a).

The Severin doctrine is, strictly speaking, irrelevant to whether the contract allows GIT to assert its subs' claims. Severin is a federal-common-law doctrine that prevents a prime contractor from making claims against the government on its sub's behalf if the sub has released it from liability. See Severin, 99 Ct.Cl. at 443-44; see also George Hyman Constr. Co. v. United States, 30 Fed.Cl. 170, 173-74, 177 n. 14 (1993) (discussing and affirming continued validity of doctrine). Severin only sets a limit, based on privity of contract and sovereign immunity, on contractors' ability to press subs' claims against the government. See George Hyman, 30 Fed. Cl. at 173. The Severin doctrine does not bar GIT's claim, but it can not affirmatively enable that claim if the contract itself bars it.

Severin is nonetheless part of the body of law applicable to government procurement, which the contract incorporates. It may be relevant in determining how to interpret the contract. That includes how to interpret the "cost of this work" clause, on which GIT's second argument depends.

Courts have strictly limited the Severin doctrine, out of reluctance to leave subs with valid claims out in the cold. Courts have applied Severin only if the government proves that a sub has executed an ironclad, unconditional release of a prime. See, e.g., Simmons, 304 F.2d at 888-89; Folk Constr. Co. v. United States, 2 Cl.Ct. 681, 685 (1983) (collecting cases). See generally Henry R. Kates, Facilitating Subcontractors' Claims Against the Government Through the Prime Contractor as the Real Party in Interest, 52 Geo. Wash. L.Rev. 146, 152-55 (1983). Post-Severin opinions often interpret releases and contracts very generously in order to let primes press their subs' claims. See Kates, supra, at 151-59. Those opinions also allow primes to rely on settlements with subs in which the prime's liability is conditional. In such agreements, the sub releases the prime from any liability, and the prime promises only that it will press the sub's claim against the government and pay the sub whatever it recovers. See Simmons, 304 F.2d at 890; Kates, supra, at 155 n. 80 (collecting similar cases). These opinions suggest that the courts which resolve most federal-contracting cases might broadly interpret the "cost of this work" clause if necessary to avoid leaving the subs unable to recover their costs.

GIT, however, has not cited any Severin case addressing whether the "cost of this work" clause, or federal-contracting law in general, allows a prime to recover its sub's costs without having settled with the sub. Courts applying Severin have neither expressly required a "settlement" nor specified that one is unnecessary.38 That leaves this court free to interpret this contract to require a settlement. And while post-Severin courts have generously interpreted contracts in order to avoid leaving subs unable to recover, GIT's subs are in no such danger. If this court reads the termination-for-convenience clause to require a settlement, then GIT and its subs can negotiate settlements, and GIT can then press the subs' claims. GIT's Severin argument thus cannot itself justify recovery on the subs' claims, and cannot aid GIT's "cost of this work" argument.

GIT's brief focuses on Severin. It does not discuss why this court should read the "cost of this work" clause to permit GIT to recover its subs' unsettled claims. Nor does it cite authority for such a reading. It simply notes that the subs' claims are for costs incurred in the work. MK replies by invoking the interpretive principle that courts should harmonize and give effect to every provision of a contract. For GIT to press its subs' claims under paragraph F(1)(b), it must have settled those claims. If this court nonetheless reads paragraph F(1)(a) to allow GIT to press its subs' unsettled claims, MK argues, it would rob paragraph F(1)(b) of any effect.

MK is not entirely correct. It is easier for a prime to settle with a sub and then recover the amount of the settlement under the "settling and paying" clause than it would be to pursue recovery under the "cost of this work" clause. If a prime has settled with its sub, then the government must accept the settlement's amount, within a broad range of reasonableness.39 A termination settlement need not strictly adhere to FAR cost principles. See 48 C.F.R. §§ 49.201 & .303-5(d);40 see also id. at § 49.113; 48 C.F.R. pt. 31. A prime seeking to recover the costs of settling with a sub under the "settled and paid" clause thus need not strictly prove the allowability, reasonableness, and allocability of all of the sub's costs. But if a prime sought to recover its sub's costs directly as a "cost of this work," without a settlement, then the cost-accounting principles of part 31 would apply to the claimed costs without any relaxation.

To allow a prime to recover its subs' costs under the "cost of this work" clause would thus not deny all effect to the "settling and paying" clause. Such an interpretation, however, would weaken the incentive for a prime and sub to settle before the prime makes a claim against the government. It would thus undermine the prime's duty to conclude such settlements.

MK bolsters its argument by asserting that the "cost of this work" cannot include unpaid, contingent costs. FAR cost principles support this argument. As noted above, those principles define a contingency as a possible future event or condition arising from causes whose outcome is not now determinable. See 48 C.F.R. § 31.205-7(a). "Costs for contingencies," the FAR provides, "are generally unallowable for historical costing purposes." Id. at § 31.205-7(b).

Ultimately, whether the subs' costs are unallowable "costs for contingencies" depends on the perspective from which they are viewed. From GIT's perspective, its liability to the subs is indeed contingent. GIT's liability depends on the currently indeterminable outcome of its negotiations and litigation with the subs. MK stresses this perspective. But from the subs' perspective, their costs are not contingent: they have already incurred those costs. GIT stresses that perspective. The subs' claims, in short, are contingent, but the costs included in the claims are not.

Focusing only on the language of GP 55F(1)(a), GIT's position seems stronger. The paragraph refers to "cost," not "claims"; the subs incurred "costs" for the "work." But considering GP 55F as a whole and, more broadly, the structure of termination settlements, MK's position is stronger.

GP 55F provides in relevant part that,

If [GIT] and [MK] fail to agree on the whole amount to be paid [GIT] because of the termination of work, [MK] shall pay [GIT] ...:

(1) For Subcontract work performed before ... termination ...:

(a) The cost of this work;

(b) The cost of settling and paying termination settlement proposals under terminated subcontracts ...; [and]

(c) A sum, as profit on subparagraph F(1)(a) above, determined ... to be fair and reasonable.

GP 55F thus focuses on which of GIT's costs MK must pay. The "cost of settling and paying" under F(1)(b) can of necessity refer only to GIT's cost, not a sub's. The "profit" allowed by F(1)(c) is GIT's profit. The reasonable profit earned by GIT's subs should be part of their settlements with GIT.41 It would be inconsistent to read "cost" in F(1)(a) to mean "cost incurred by GIT or by its subs" when "cost" in F(1)(b) necessarily means "cost incurred by GIT," and "profit" in F(1)(c) means "profit earned by GIT."

More broadly, the standard federal termination-for-convenience clause, which is reproduced in GP 55, requires terminated contractors to settle with their subs. See 48 C.F.R. § 52.249-2(b)(5). A terminated contractor will thus liquidate its subs' claims and consolidate those claims with its own into a single settlement proposal to the government. See id. at § 52.249-2(e).42 This approach reflects the general policy of federal-contracting law, which favors settlement over unilateral determination of termination damages. See 48 C.F.R. § 49.103.

The FAR contains not only standard termination clauses for federal contracts, but principles to govern terminations. See 48 C.F.R. pt. 49. The principles governing settlements with subcontractors begin by stressing the well-settled rule that a sub cannot sue the government directly. See 48 C.F.R. § 49.108-1; see generally Kates, supra at 146 & n. 4 (collecting cases). The same section then notes that a prime is "responsible for the prompt settlement of the settlement proposals of [its subs]." 48 C.F.R. § 49.108-1. The statements' conjunction emphasizes that a sub must recover its costs by settling with the prime, not by directly suing the government. In this case, while the subs have not directly sued MK, it would nonetheless significantly undermine the FAR's settlement structure to accept GIT's theory. That theory suggests that, while a prime must settle with its subs, and while the subs may not sue the government to recover their costs, the prime may nonetheless sue the government to recover the subs' costs without first having settled with the subs.

FAR § 49.108-6 provides further evidence of the FAR's presumption that a contractor will settle its subs' claims before it settles its own claim with the Government. The subsection provides that, if a prime's inability to settle with a sub delays the settlement of the prime contract, the Government may settle with the prime. In so doing, however, the Government "shall except the subcontractor settlement proposal from the settlement in whole or part and reserve the rights of the Government and the prime contractor with respect to the subcontractor proposal." 48 C.F.R. § 49.108-6. If a prime could directly recover its sub's costs under the "costs of this work" clause, there would be no need for a procedure to address situations in which a prime's inability to settle with a sub delayed its settlement with the government.43

The more natural reading of the two disputed paragraphs is that F(1)(b), the one that expressly refers to subs' claims, encom