concurring in part, and dissenting in part.
Following an initial trial, despite an erroneous nominal damage award, Pryer had won something important: He had vindicated his constitutional rights by obtaining the formal determination of a jury that the savage beating administered to him by Appellants was wrongful. Under the majority’s ruling, that determination is taken away, even though no one contends that it was the product of any error. Solely because the guards benefitted from an erroneous instruction that contributed to an inadequate damages determination, the majority now holds that they must be accorded a further windfall by giving them a second bite at the liability issue. I respectfully dissent.1
I.
At the outset, it appears that the majority has failed to apply the appropriate standard of review. In Vizzini v. Ford Motor Co., 569 F.2d 754 (3d Cir.1977), we first enunciated the scope of review to be applied in deter mining the appropriateness of an order limiting a new trial to damages. See id. at 760 (“[T]he standard we apply on review is whether, viewing the circumstances as a whole, it was consistent with the sound exercise of discretion for the trial judge to limit the new trial to the issue of damages.”); see also Stanton v. Astra Pharmaceutical Prods., Inc., 718 F.2d 553, 576 & n. 42 (3d Cir.1983) (observing that “the only question is whether *459the district court abused its discretion in ordering” a partial new trial on damages).2
The majority acknowledges the abuse of discretion standard, but it applies a de novo standard, reversing the trial court because this court is led to “favor a new trial on all issues” due to this court’s inability to understand why the jury awarded only nominal damages. Supra at 456. I recognize that under Gasoline Products it must “clearly appear” that a partial retrial may be had without injustice.3 Nevertheless, I believe that the required clarity is for the trial judge, whose view should prevail even if the issue is less than clear to us as a reviewing court.
There is good reason- to commit such matters to the sound discretion of the trial court. The trial court has seen all of the evidence first-hand.. It has ruled on motions, instructed the jury, and overseen the conduct of trial. In short, the trial court has lived with and thought about the case for a considerable period, and so is in an advantageous position to determine what will or will not overtax the jury’s discernment. We, on the other hand, upon review of a cold record, cannot be expected to achieve the same mastery of every nuance of the case. It is hardly surprising that, where a trial judge is justly assured that issues are separable, the matter may appear less clear to a reviewing court.
The majority seeks to justify its departure from our established standard of review, and its consequent substitution of its discretion for that of the trial judge, by invoking two purported exceptions: First, the majority asserts that no deference is due “where a trial court fails to explain its grounds for exercising discretion, and its reasons for doing so are not otherwise apparent from the record”. Supra at 453 n. 4. However, the District’Court’s reasons for limiting the second trial to damages are made abundantly clear in its opinion: As the Court explained, “a new trial is appropriate on the issue of damages because the court1 failed to properly instruct the jury on damages” (by, inter alia, giving an unwarranted nominal damages instruction), and because the resulting nominal damages award was against the weight of the evidence. Implicit in the Court’s recital of errors in the damages instruction and award was its recognition that there was no error in the liability instructions or verdict.4 Ordinarily, the *460presence of error limited to a single issue should be considered reason enough to limit retrial to that issue. Although the majority may consider the stated reasons insufficient, it cannot fairly by said that the trial Court “articulated no rationale”.5
Second, the majority asserts that the District Court’s grant of a partial retrial “turns on the application of a legal precept to the evidence”. Supra at 453 n. 4. The majority does not identify what “legal precept” it deems controlling. The due process and fairness concerns set forth in Gasoline Products clearly speaks to discretion, and we (and other courts) have always considered its application to be an exercise of discretion, rather than a legal determination. The majority’s cryptic invocation of the “legal precept” exception, without identifying a controlling question of law, threatens to eviscerate the abuse of discretion standard: potentially, every exercise of discretion may be subjected to plenary review at the whim of the reviewing court by pointing out that some “legal precept” is somehow involved. By substituting its judgment for that of the District Court (in which the discretion is intended to be reposed) on the basis of such ill-considered, amorphous exceptions, the majority has fundamentally altered our standard of review, effectively overruling sub silentio our decision in Vizzini and Stanton.
Rather than reversing whenever we cannot be sur e that the issues are separable, I would uphold the trial court’s exercise of discretion unless we could say with some assurance, based upon clearly identified factors in the record, that the issues are not separable.6 Because the majority’s opinion in this case does not identify a record basis for overriding the District Court’s discretion, it effectively leaves no discretion. This result is contrary to the purpose of the Gasoline Products rule in that it will preclude even justly limited retrials, thereby thwarting judicial economy and unduly burdening plaintiffs who have already fairly obtained a favorable liability verdict. See Wheatley v. Beetar, 637 F.2d 863, 867-68 (2d Cir.1980) (noting that where defendant has had a fair trial on the issue of liability it would be “grossly unfair to plaintiff, as well as contrary to the spirit of F.R. 59, to require a retrial of the question of defendant’s culpability” solely because the jury did not properly consider damages) (internal quotation marks omitted).7
II.
In deciding whether a new trial on damages alone would be unjust, it is not enough simply to observe that issues con-*461eerning damages and liability are interwoven. Of necessity they are always interrelated, because only those damages proximately caused by conduct giving rise to liability are recoverable. The universal requirement of a causal link between liability and damages means that the issues can never be completely unlinked.8 •
Nevertheless, it is clear that in many cases damages may justly be tried apart from liability. Such separate trials are a common occurrence, both ab initio and on retrial.9 See Rosa v. City of Chester, 278 F.2d 876, 882 (3d Cir.1960) (observing that “[f]ederal appellate and district courts have time and again ordered new trials as to damages only”). See generally Propriety of limiting to issue of damages alone new trial granted on ground of inadequacy of damages — modem cases, 5 A.L.R. 5th 875 at § 4 (1993) (providing extensive citations to cases where “new trial was required upon damages only, since defendant’s liability was supported or established by the evidence and there was no necessity to retry the liability issue”).
Assessment of the degree to which facts concerning liability and damages are interrelated seems a quixotic venture with little direct bearing on the justice or injustice of separate trials; and it does not lead to any workable standard. A more serviceable approach is to consider whether there is something about how the interrelationship unfolded at trial that would make separation of the issues on retrial unjust. Courts have focused on two aspects in particular: whether the jury’s erroneous treatment of damages infected its determination of liability, and whether it appears that the verdict on liability and damages together represents a compromise. See, e.g., Olsen v. Correiro, 1995 WL 62101 *4 (D.Mass., Feb.3, 1995) (observing that in determining “interwoven” test “some courts ask whether the invalidity of the jury’s reasoning as to one part of the verdict could have ‘infected’ their reasoning as to the other part” and that “[a] variation on the ‘infection’ inquiry is the ‘compromise verdict’ inquiry, under which a court asks ... whether ... an inadequate damages award reflected the jury’s ambivalence about its finding of liability”).
The first test, then, is whether the error which undermined the damages verdict also corrupted the liability verdict. Apropos of this standard, we have held that “if the error compelling the reversal relates solely to the damage question, the new trial will be restricted to that question.” Rosa, 278 F.2d at 883. Moreover, we have consistently recognized the propriety of partial new trials “in those cases where it is plain that the error which has crept into one element of the verdict did not in any way affect the determination of any other issue.” Vizzini, 569 F.2d at 760 (quoting Romer v. Baldwin, 317 F.2d 919, 922-23 (3d Cir.1963)) (additional citations omitted).
Here, there is no suggestion that the erroneous instructions or reasoning which necessitated retrial as to damages had any bearing on the liability verdict. The finding of liability was amply supported by the evidence, and Appellants have not assigned any error with respect thereto. I do not *462read the majority opinion as indicating that the liability verdict was erroneous.
It is with respect to the second test for unfairness — the “compromise verdict” inquiry — that I must part ways with the majority. Appellants never took the position, either before the District Court or in briefing before this Court, that the liability verdict represented a compromise.10 Ordinarily an appellate court will not reverse a trial court on grounds not asserted by the parties. Nevertheless, the majority opines that “the verdict most likely represented a compromise among jurors with differing views on whether the correctional officers were liable”, because the majority is “unable to understand” why the jury would otherwise return an award of $1.00 despite Pryer’s uncontested evidence of injuries. Supra at 457.
A compromise verdict results when jur ors resolve their inability to make a determination with unanimity as to liability by finding inadequate damages. As the majority correctly observes, we have previously held" that a limited new trial is unjust where the verdict is “obviously the result of compromise” on questions of liability and damages. Id. (citing Rosa, 278 F.2d 876).11 To avoid undue interference with the trial court’s discretion to limit a retrial, we have required trial on all issues only where the compromise is “obvious”. See Rosa, 278 F.2d at 883.12 This standard is in keeping with the practices of our sister Courts of Appeals.13 It is not enough that the damage award is inadequate, or nominal.14 Rather, the compromise must be evident from other factors of record.15 Where the presence of compromise is doubtful, a new trial on liability is not *463required.16
Here the presence of a compromise is certainly not obvious. That the jury was able to find some of the initial defendants hable and others not clearly demonstrates its ability to assess and distinguish conduct, and to adjust its liability findings accordingly. And unlike some of the cases relied upon by the majority, the jury in this case did not simply “split the difference” between the parties’ positions in arriving at a damages figure. Measured against the extent of the beating and injuries inflicted on Pryer, an award of $1 hardly seems a “compromise”.17
The majority does not point to any factors suggestive of compromise, other than the purportedly inexplicable deficiency in damages. But that deficiency is readily explained by the District Court’s erroneous instructions. One need not reach out in search of speculative alternative explanations where the very reason for the new trial was the court’s acknowledgment of its erroneous instructions on damages. To all appearances, the reason for the jury’s nominal damages award was the court’s erroneous instruction on that very issue, and not that the jury did not mean what it said as to liability.18 In Rosa, we found that an inadequate damages verdict was not indicative of compromise where the inadequacy was apparently due to erroneous evidence and instructions.19 Numerous cases are in accord.20 In the present case, we should similarly decline to presume that the liability verdict was a product of compromise.21
*464III.
Appellants have raised the possible apportionment of Pryer’s injuries in a separate assignment of error, contending that the trial court improperly prevented the jury in the second trial from considering whether some or all of the injuries resulted from the conduct of guards who were absolved of liability in the first trial. If Appellants’ contention were correct, they would be entitled to a new trial on damages, in which they could attempt to assign responsibility for particular injuries to the other guards. They would not, however, be entitled to revisit the question of their own liability.
Appellants’ contention on apportionment turns on whether the evidence in the second trial would allow a non-speculative attribution of particular injuries to particular causes. The majority addresses this issue only in passing. While it does opine that “it is far from clear that Pryer’s injuries cannot be apportioned”, this comment appears to be part of its discussion of the first trial. See supra at 456.22 Whether such an apportionment was supported by the evidence in the first trial is of no moment, since the majority has agreed with the District Court that the damages verdict was tainted by erroneous instructions that necessitated a new trial.
The issue properly before us concerns whether apportionment was proper on the record before the jury in the second trial (not whether the first jury could or did apportion damages). As to that issue, Appellants had the burden of providing a record basis for non-speculative attribution of particular injuries to particular causes.23 Appellants do not attempt to parse the record to show how such a determination could be made; nor does the majority. Appellants merely point out (and the majority repeats) that one non-liable guard admitted to beating Pryer on the leg which was broken. But this was only one beating in a long-running melee that included other assaults to the same leg. It appears that any conclusion that one guard’s conduct was the sole cause of that injury would be speculative.
As the District Court held,
The evidence in this case clearly established that plaintiff suffered a single indivisible injury caused by multiple blows inflicted by'multiple guards. There is no logical, reasonable or practical basis to apportion his pain, suffering ... between the blows inflicted by the four *465defendants and the blows inflicted by the other guards, or to attribute such injuries to the blows inflicted by any particular guard. Any attempt to do so would be purely arbitrary.
Mem., March 3, 2000, at 6. Because nothing in Appellants’ submission or the majority’s opinion convinces me that the court’s analysis was incorrect, I would uphold its refusal to allow the jury to speculate as to apportionment of Pryer’s damages.
. I join, however, in Part 11(A) of the majority’s opinion, which upholds the District Court’s decision setting aside the nominal damages verdict.
. Our application of an abuse of discretion standard is in keeping with the practice of other courts. See, e.g. Sprague v. Boston & Maine Corp., 769 F.2d 26, 28 (1st Cir.1985) (allowing separate retrial on damages and noting that "[t]he trial judge is afforded broad latitude ... to limit the new trial to the issue of damages if liability has, in his opinion, been competently determined by the jury”) (citing cases); Young v. International Paper Co., 322 F.2d 820 (4th Cir.1963) (affirming trial court's limitation to issue of damages and holding that trial judge, in reviewing record, could conclude that jury properly found defendants liable but verdict too low to be sustained).
. See Gasoline Products Co. v. Champlin Refining Co., 283 U.S. 494, 500, 51 S.Ct. 513, 75 L.Ed. 1188 (1931) (holding that a partial new trial "may not be properly resorted to unless it clearly appears that the issue to be retried is so distinct and separable from the others that a trial of it alone may be had withoul injustice"). See also, e.g., Williams v. Rene, 72 F.3d 1096, 1101 (3d Cir.1995) (holding that under Gasoline Products a new trial must extend to all issues where “the issue of damages ... is so intertwined with liability that one cannot be submitted to the jury independently of the other without confusion and uncertainty.”).
.That the District Court’s decision was predicated on the absence of - error affecting the liability verdict is further "apparent from the record” through the Court's express reliance on Wheatley v. Beetar, 637 F.2d 863 (2d Cir.1980), in which the Second Circuit upheld a new trial limited to damages because the defendant had received a fair trial on the issue of liability.
. Where a trial court articulates its reasons for an exercise of discretion, it need not anticipate and address every possible concern in order to be entitled to deference. Here, for example, the District Court cannot reasonable by faulted for failing to address tech possibility that the damages verdict was the result of compromise, as that suggestion was never advanced by the Defendants, but instead originated with the majority in this appeal.
. An example of the appropriate level of deference may be found in Stanton, where we reversed the partial retrial only because a "great deal of evidence" in the record "strongly suggested]" a compromise verdict, based on two objective factors: a comparison of the verdict amount to the evidence that provided a "compelling indication" of compromise, and the jury’s answers to "special questions" that "apparently absolved [the defendant] of liability”. Stanton, 718 F.2d at 576-78 & n. 45. As discussed in Part II, below, no such compelling objective factors are present here.
.Moreover, in the present case, the majority has gone beyond simply substituting its judgment for that of the trial court: As will be explained below, the majority has based its decision on considerations that were not placed before the trial court.
. See Olsen v. Corretro, 1995 WL 62101 *3 (D.Mass., Feb.3, 1995) (observing that the "interwoven” test of Gasoline Products is "quite difficult to apply in practice, for damages and liability are in some sense interwoven”).
. The rules expressly authorize such separate trials. See Fed.R.Civ.P. 42(b) (permitting, in furtherance of convenience or when conducive to expedition and economy, separate trial of any separate issue); Fed.R.Civ.P. 59(a) (permitting new trial on "all or part of the issues").
. Indeed, when invited to adopt that position at oral argument, Appellants' counsel effectively declined to do so.
. It is significant that the statement in Rosa cited by the majority is followed immediately by an explicit recognition of the converse rule where there is no such obvious compromise: "On the other hand if the error compelling the reversal relates solely to the damage question, the new trial will be restricted to that question.” Rosa, 278 F.2d at 883.
. Cf. Huddleston v. Crain Brothers, Inc., 183 F.Supp. 874 (W.D.Pa.1960) (concluding new trial on damages alone was appropriate where verdict was not “absolutely the result of a compromise on liability and damages”) (emphasis added) (citing Rosa, 278 F.2d 876).
. See Burger King Corp. v. Mason, 710 F.2d 1480, 1487 (11th Cir.1983) (defining compromise verdict as "one where it is obvious that the jury compromised the issue of liability by awarding inadequate damages.”) (emphasis added); Freight Terminals, Inc. v. Ryder System, Inc., 461 F.2d 1046, 1053 (5th Cir.1972) (same); Great Coastal Express, Inc. v. International Brotherhood of Teamsters, 511 F.2d 839 (4th Cir.1975) (new trial required where inadequate verdict "could only have been a sympathy or compromise verdict”) (emphasis added). See also Spell v. McDaniel, 604 F.Supp. 641, 653 (E.D.N.C.1985) (finding new trial on damages proper where the "totality of the circumstances simply [did! not point unerringly to a compromise verdict”) (emphasis added).
. See Rosa, 278 F.2d 876 (inadequate verdict in amount of $1 held not indicative of compromise on issues of liability and damages). See also De Luca v. Wells, 58 Misc.2d 878, 297 N.Y.S.2d 35 (1968) (noting that not every inadequate verdict is ipso facto a compromise verdict); Spell, 604 F.Supp. at 653 (holding that an award of minimal damages, in itself, is insufficient to prove a compromise) (citing Burger King, 710 F.2d at 1487). Cf. Edwards v. Sears Roebuck & Co., 512 F.2d 276, 282 (5th Cir.1975) ("A jury's finding as to liability can be binding even though its monetary award is found to be [inadequate] or even improperly influenced — our deference to and faith in the jury system demands at least this much.”).
. See Burger King, 710 F.2d at 1487; Mekdeci v. Merrell National Laboratories, 711 F.2d 1510, 1513-14 (11th Cir.1983).
. Cf. Lockhart v. Westinghouse Credit Corp., 879 F.2d 43, 54 (3d Cir.1989), (rejecting claim for new trial although it ''[could] not be said with fair assurance that [a compromise verdict] did not occur”, and noting that "[o]ur task on review is to ascertain only whether the jury’s verdict is reasonable in light of the evidence presented, and not to indulge in unsubstantiated and speculative assertions”), effectively overruled on other grounds, Hazen Paper Co. v. Biggins, 507 U.S. 604, 113 S.Ct. 1701, 123 L.Ed.2d 338 (1993).
. Compare Stanton, 718 F.2d at 576-77 (finding "compelling indication” of compromise in damages award of about one third of the amount supported by the evidence).
. Any doubt on this score should be resolved by according the jury's liability verdict the usual presumption of propriety and regularity. See Spell, 604 F.Supp. at 651 (in conducting compromise inquiry, court started from proposition that jury's verdict is to be afforded every inference that it is proper and honest).
. See Rosa, 278 F.2d at 882.
. See, e.g., Saide v. Stanton, 135 Ariz. 76, 659 P.2d 35 (1983) (concluding new trial could be limited to damages alone, though liability issue was strongly contested, where inadequacy in damage award resulted from trial court’s erroneous instructions to jury regarding amounts includable in assessing damages); Capelouto v. Kaiser Foundation Hospitals, 7 Cal.3d 889, 103 Cal.Rptr. 856, 500 P.2d 880 (1972) (new trial properly limited to damages where ample evidence supported liability, and inadequacy of the verdict was probably due to erroneous jury instruction); Cedars of Lebanon Hosp. Corp. v. Silva, 476 So.2d 696 (Fla. 3d DCA 1985) (affirming limitation of new trial to damages where inadequate award was apparently generated by trial court's instruction and case was not one where evidence of defendants' liability was tenuous or insubstantial, or where it appeared that juror who would have found no liability whatsoever compromised with those who would have found defendants liable, and observing that new trial on liability was not required simply because the issue was disputed).
.Some courts have employed an explicit multi-factorial analysis to determine whether a compromise verdict was reached, based on: (1) clarity of the jury instructions and verdict form; (2) length of jury deliberations; (3) strength of the evidence as to liability; (4) questions and notes from the jury during deliberations; and (5) whether the case involved a sympathetic plaintiff and unsympathetic defendant. See, e.g., Mekdeci, 711 F.2d 1510; Spell, 604 F.Supp. at 651. Here, there is no *464indication that the liability instructions were unclear; that the jury deliberations were over-long; that liability evidence was lacking; that the jury asked questions evincing uncertainty; or that Pryer, an incarcerated felon, was a particularly sympathetic plaintiff. Compare Vizzini, 569 F.2d at 761 (requiring full retrial where liability evidence was "very thin" and jury’s failure to agree on damages after two full days of deliberations indicated that deliberations were affected by liability issues).
. The majority appears to intimate that the first jury might have arrived at its $ 1 verdict by determining that all of the substantial elements of damage were caused by guards found not liable. This may be possible in theory, but given the usual relationship between force and injury, it is hardly plausible that a jury would find that guards who inflicted multiple broken bones, lacerations, contusions and head injuries used only reasonable force, while concluding that the force used by those who did only nominal damage was excessive.
. See, e.g., United States v. Alcan Aluminum Corp., 964 F.2d 252, 268-69 (3d Cir.1992) ("[Wjhere joint tortfeasors cause a single and indivisible harm for which there is no reasonable basis for division according to the contribution of each, each tortfeasor is subject to liability for the entire harm.[I]t is the tortfeasor’s burden to establish that the damages are capable of such apportionment.”).