SELYA, Circuit Judge.
This appeal represents the final voyage of a case that has persisted in the federal courts for nearly a decade. Finding, as we do, that the judgment below is fully supportable, we affirm.
I. BACKGROUND
Because these waters have been so thoroughly charted, we merely sketch the background insofar as is necessary to set this appeal and the underlying litigation into perspective,
A
Plaintiff-appellant Dedham-Westwood Water District (which, together with its predecessor in interest, Dedham Water Company, we shall call “Dedham”) is a regulated public utility. It supplies drinking water to some 40,000 persons who reside in the Massachusetts towns of Ded-ham and Westwood. One source of this water is the White Lodge Well Field. The well field lies in an industrial park along the west bank of the Neponset River. Defendant-appellee Cumberland Farms Dairy, Inc. (“Cumberland”) operates a truck maintenance facility on the river’s east bank.
In March 1979, Dedham discovered that two wells were contaminated with volatile organic chemicals (VOCs). It removed the wells from service, contacted the Massachusetts Department of Environmental Quality Engineering (DEQE), sought alternative water supplies, and began a somewhat haphazard investigation aimed at fixing responsibility for the pollution. This investigation included a series of surface-water tests.
In June 1979, Dedham began to pump the contaminated wells to waste. Shortly thereafter, it hired two consultants, Calgon Corporation and Metcalf & Eddy, to assess treatment alternatives. In early 1980, Ded-ham’s governing board budgeted a sum of money to pay Metcalf & Eddy for designing a two-stage treatment plant that would deal with both the VOC problem and a separate water-quality issue. In July 1981, Dedham submitted plans for the treatment plant to DEQE. Before year’s end, Ded-ham publicly announced that it would build the treatment plant at White Lodge.
A few months prior to this announcement, a Dedham employee discovered YOCs in a drainage ditch running from Cumberland’s property. Dedham promptly hired Geraghty & Miller (“G & M”), a firm specializing in hydrogeology, to investigate the source of the contamination and deter
mine the zone of capture. In January-1982, a drawdown test conducted by G & M established that the Neponset River was not a barrier to the flow of contaminants; theoretically, contaminants could flow under the river and into the well field. G & M proceeded to sink test wells in and around the White Lodge field. In July of 1982, it identified Cumberland as a likely source of pollutants. Three months later, G & M reported that, in its opinion, Cumberland was the major cause of the White Lodge contamination.
The G & M report heralded the start of the instant litigation. Invoking the Comprehensive Environmental Response, Compensation, and Liability Act of 1980 (CERC-LA), 42 U.S.C. §§ 9601-9675 (1988), and its state-law counterpart, Mass.Gen.L. Ch. 21E (1990 & Supp.1991), Dedham sued Cumberland in federal district court. It asserted claims for response costs incurred as a result of actual and threatened contamination at the well field.
As the suit plodded toward trial, the planned treatment plant was slowed by zoning questions, neighbors’ objections, dissatisfaction on DEQE’s part, and a gaggle of other roadblocks. After appellant redesigned the plant to feature lower aerating towers and better emission controls, these problems dissipated. DEQE granted a permit. Construction began in 1985 and the treatment plant went on line in early 1987.
Appellant’s court case progressed more deliberately. It took four full years to overcome a jurisdictional obstacle.
When the case was finally tried, Dedham did not succeed in proving that Cumberland was the source of the contamination. The district court (Tauro, J.) entered judgment for the defendant.
Dedham Water Co. v. Cumberland Farms Dairy, Inc.,
689 F.Supp. 1223 (D.Mass.1988)
(“First Trial Op.”).
Dedham appealed on the basis of Judge Tauro's purported failure to examine an alternative theory of liability under CERCLA and Chapter 21E, viz., whether Cumberland, although not guilty of causing the discovered contamination, may nevertheless have posed an actionable threat of future contamination, to which Dedham responded in an objectively reasonable (if costly) manner. We directed the district court to revisit this aspect of the matter.
Dedham Water Co. v. Cumberland Farms Dairy, Inc.,
889 F.2d 1146 (1st Cir.1989)
(“Earlier Merits Appeal
” or
“EMA
”). We also decided that a new trier should preside.
See In re Dedham Water Co.,
901 F.2d 3, 4-5 (1st Cir.1990) (per curiam) (interpreting D.Mass.Loc.R. 8(i)).
B
Before setting sail, we deem it imperative (a) to delineate the narrow perimeters within which the further proceedings on remand were to be conducted and (b) to report the outcome of those proceedings. In the first trial, appellant was unable to “prove[ ] that the contaminants released by Cumberland Farms had migrated to the [White Lodge] wells.”
EMA,
889 F.2d at 1149. Dedham did not challenge this ruling on appeal. Hence, the issue could not be resurrected during the second trial. From that point forward, it had to be taken as gospel that Cumberland “was not liable for the expenses incurred by Dedham ... in investigating the cause of the pollution to its wells and rectifying [that pollution].”
Id.
The only issue that remained open at the second trial was whether Cumberland’s “releases (or threatened releases) might nonetheless have caused the plaintiff to incur ‘response costs’ even though those releases did not
in fact
contaminate the wells.”
Id.
at 1157 (emphasis in original).
Phrased another way, the pre-remand proceedings conclusively determined that Cumberland was not legally responsible for contaminating the White Lodge well field. Therefore, to the extent that appellant’s activities — e.g., retaining consultants, performing scientific studies, building the
treatment plant-were in response to actual contamination, Cumberland was home free. But, polluting substances that stopped short of the well field's boundary could conceivably have caused the appellant to incur expenses compensable under CERCLA. Id. at 1151-54, 1157-58. If, and to the extent that, Cumberland posed such a threat, and Dedham acted in response thereto, Cumberland might be held liable. Since the district court had neglected to make a finding on this point, further proceedings were required.
The second trial was perforce limited to a resolution of this issue and, if the district court found in appellant’s favor, an assessment of damages. The case was redrawn to Judge Skinner. After taking additional testimony, he determined that appellant had not incurred any response costs as a result of releases, or the threat of releases, attributable to appellee.
Dedham Water Co. v. Cumberland Farms Dairy, Inc.,
770 F.Supp. 41 (D.Mass.1991) (“Second
Trial Op.”).
Dedham appeals. We affirm.
II. APPLICABLE STANDARDS OF REVIEW
When a district court makes findings of fact in a bench trial, the clear-error standard pertains.
See
Fed.R.Civ.P. 52(a);
see also Gopher Oil Co. v. Union Oil Co.,
955 F.2d 519, 526 (8th Cir.1992) (clear-error standard applies to district court’s findings of fact in CERCLA cases). Consequently, the court of appeals must defer in considerable measure to the nisi prius court. Findings of fact will be given effect unless, after reading the record with care and making due allowance for the trier’s superior ability to gauge credibility, the reviewing court “form[s] a strong, unyielding belief that a mistake has been made.”
Cumpiano v. Banco Santander Puerto Rico,
902 F.2d 148, 152 (1st Cir.1990). In this process, equal respect is afforded to the district court’s evaluation of documentary and testimonial evidence,
Reliance Steel Prods. Co. v. National Fire Ins. Co.,
880 F.2d 575, 576 (1st Cir.1989); and, moreover, the same high level of respect attaches whether “the conclusions of the trial court depend on its election among conflicting facts or its choice of which competing inferences to draw from undisputed basic facts.”
Irons v. FBI,
811 F.2d 681, 684 (1st Cir.1987).
As a general rule, causation questions are grist for the factfinder’s mill,
see, e.g., Peckham v. Continental Cas. Ins. Co.,
895 F.2d 830, 837 (1st Cir.1990), and thus, are within the traditional scope of clear-error review. Where causation is in issue, “[n]ot only ordinary fact questions, but also ‘evaluative applications of legal standards ... to the facts’ are properly [for the factfinder].”
Springer v. Seaman,
821 F.2d 871, 876 (1st Cir.1987) (citations and footnote omitted). We see nothing about the issue of causation in a CERC-LA case that would serve either to sidetrack these principles or to reconfigure their application.
See, e.g., United States v. R.W. Meyer, Inc.,
889 F.2d 1497, 1507 (6th Cir.1989) (reviewing finding of causation in CERCLA case for clear error),
cert. denied,
494 U.S. 1057, 110 S.Ct. 1527, 108 L.Ed.2d 767 (1990).
In apparent recognition of the high hur-die that the clear-error standard presents, appellant seeks to cast the district court's decision in a more malleable medium. Although appellant reserves the claim that clear error was committed, see infra Part IV, it spends most of its energies attempting to convince us that the court below applied an incorrect legal standard. See infra Part IlL If this position is viable- and that is a large "IF," best written in capital letters-it improves appellant's chances in two ways. First, unlike findings of fact, errors of law are subject to plenary review. Second, to the extent that findings of fact can be shown to have been predicated upon, or induced by, errors of law, they will be accorded diminished respect on appeal. See RCI Northeast Servs. Div. v. Boston Edison Co., 822 F.2d 199, 203 (1st Cir.1987).
III. THE MISTAKE-OF-LAW THESIS
In an effort to put some flesh on the bare bones of its mistake-of-law thesis, appellant develops two criticisms of the judg
ment below. One criticism is relatively general; the other is relatively specific. The gist of the broader criticism is that the district court erroneously factored a subjective component into the calculus of decision by considering what Dedham actually knew (or believed) about a threat of potential harm and when Dedham actually acquired such knowledge. The more particularized criticism suggests that, regardless of whether a subjective component could properly be considered, the trial court’s conception of the legal standard was faulty because it slavishly embraced a temporal standard, requiring that appellant prove it had identified Cumberland as the source of threatened releases before it incurred the costs for which recovery was sought.
We examine these contentions
seriatim.
In concluding this section of our opinion, we also discuss a fallback position that appellant visualizes as an anchor to windward.
We begin with appellant’s insistence that any inquiry into its institutional state of mind was erroneous because the only legitimate question before the district court was whether, objectively viewed, Cumberland posed a threat that warranted responsive action. There is a gaping hole in the fabric of this construct.
Appellant’s theory at the second trial, as expressed in its pleadings, evidence, and trial memoranda, hinged upon precisely the mixed subjective/objective approach it now reviles. Nothing could make this point clearer than a reading of the closing argument given by Dedham’s counsel in the district court-a summation that focused squarely on “what did they know and when did they know it?” Indeed, appellant conceded in closing argument that “[t]o the extent there was no idea of the activities on Cumberland Farms up to April of 1979 ... costs [incurred prior thereto] could be fairly excluded as having been incurred in response to a contamination.”
If the district court’s view of the law was in error-and we do not suggest that it was-Dedham must bear its fair share of responsibility for leading the court down a garden path. A litigant who attempts to sell a particular view of the law to the trial court cannot later appeal on the ground that the court bought the litigant’s wares.
See Austin v. Unarco. Indus., Inc.,
705 F.2d 1, 15 (1st Cir.) (“in general, a party may not appeal from an error to which he contributed, either by failing to object or by affirmatively presenting to the court the wrong law”),
cert. dismissed,
463 U.S. 1247, 104 S.Ct. 34, 77 L.Ed.2d 1454 (1983);
McPhail v. Municipality of Culebra,
598 F.2d 603, 607 (1st Cir.1979) (“A party may not ‘sandbag’ his case by presenting one theory to the trial court and then arguing for another on appeal.”).
Moreover, we think that appellant’s failure to articulate the subjective/objective dichotomy in the district court was particularly egregious under the circumstances of this case. Our opinion in the earlier merits appeal strongly suggested that Dedham’s institutional state of mind was relevant to the issue of whether the costs it incurred were threat-related and/or recoverable against Cumberland on that basis.
See, e.g., EMA,
889 F.2d at 1157 (observing that “[a] plaintiff ... under certain circumstances might reasonably think that a particular release would prove likely to contaminate his wells”);
id.
at 1158 (implying that if, objectively, a defendant’s releases pose a threat, a subjective fear of contamination, reasonably held, can then establish a defendant’s liability for response costs incurred by the plaintiff).
To be sure, these statements are
obiter dictum,
that is, observations relevant, but not essential, to the determination of the legal questions then before the court. Dictum constitutes neither the law of the case nor the stuff of binding precedent.
See Great Lakes Dredge & Dock Co. v. Tanker Robert Watt Miller,
957 F.2d 1575, 1578 (11th Cir.1992);
Milgard Tempering, Inc. v. Selas Corp.,
902 F.2d 703, 715-16 (9th Cir.1990). In short, dictum contained in an appellate court’s opinion has no preclusive effect in subsequent proceedings in the same, or any other, case.
Be that as it may, courts often, quite properly, give considerable weight to dictum — particularly to dictum that seems considered as opposed to casual.
See, e.g., McCoy v. Massachusetts Inst. of Technology,
950 F.2d 13, 19 (1st Cir.1991) (court of appeals should ordinarily respect considered Supreme Court dicta) (citing authorities), cer
t. denied,
— U.S.-, 112 S.Ct. 1939, 118 L.Ed.2d 545 (1992);
Posadas de Puerto Rico Assocs., Inc. v. Asociacion de Empleados, Etc.,
873 F.2d 479, 482 (1st Cir.1989) (court of appeals should ordinarily respect considered dictum of state’s highest court on points of state law). When, as here, the district court, on remand, proposed to act upon dicta contained in the appeals court’s earlier opinion in the same case, and no one demurred, it is especially important that we toe the mark and hold the parties to the usual consequence of invited error. To do otherwise “would place a premium on agreeable acquiescence to perceivable error as a weapon of appellate advocacy.”
Merchant v. Ruhle,
740 F.2d 86, 92 (1st Cir.1984).
Then, too, because appellant’s neoteric thesis was not asserted below, it is procedurally defaulted. “It is hornbook law that theories not raised squarely in the district court cannot be surfaced for the first time on appeal.”
McCoy,
950 F.2d at 22;
accord, e.g., Boston Celtics Ltd. Partnership v. Shaw,
908 F.2d 1041, 1045 (1st Cir.1990);
Sanchez-Arroyo v. Eastern Airlines, Inc.,
835 F.2d 407, 408 (1st Cir.1987);
Clauson v. Smith,
823 F.2d 660, 666 (1st Cir.1987). There is no basis in the present record to relax the rule. After all, “[l]itigants cannot try their case under one theory and then urge a remand on appeal so it can be tried again under a different theory.”
Johnson v. Allyn & Bacon, Inc.,
731 F.2d 64, 73 (1st Cir.),
cert. denied,
469 U.S. 1018, 105 S.Ct. 433, 83 L.Ed.2d 359 (1984).
For these reasons, appellant’s newly conceived subjective/objective dichotomy does not assist its cause.
We turn now to appellant’s more specific asseveration: its claim that the district court erroneously required proof that it identified Cumberland as the source of the threat prior to incurring any response costs. This asseveration is belied by the record.
The short of it is that the district court simply did not embosom, or operate on, the premise that, in order to prevail, Dedham must have identified Cumberland as the source of threatened contamination before incurring response costs. Rather, the district court found, wholly apart from the question of identification, that the actions which appellant took were not in response to approaching or threatened releases of any description.
See Second Trial Op.,
770 F.Supp. at 42-43. Thus, the court did not decide the point mechanically, based strictly on whether identification preceded expenditure, but decided it on a much different basis: that appellant was not responding to a perceived threat, wheresoever originating, when it set its course, but was responding instead to the need to rid its wells of existing contamination. In other words, appellant’s construction of the treatment plant, its search for the sources of the contamination, and its other response costs were undertaken to define the magnitude of a known problem (actual contamination), to rectify that problem and, relatedly, to fix responsibility for it with an eye toward litigation — not as an environmental rejoinder to threatened releases of any kind.
On this aspect of the case, appellant is tilting at windmills. Its claim of legal error dissolves upon analysis of the decision below and the record upon which that decision was based. The district court did not require appellant, as a condition precedent to recovery, to have identified the party responsible for the threat before beginning to incur response costs. Instead, the court’s focus was exactly where it belonged — not on whether the chicken preceded the egg, or vice versa, but on whether Cumberland somehow, or in some way, caused Dedham to incur response costs.
This was precisely the task we set in remanding the case.
See EMA,
889 F.2d at 1151-54, 1157-58.
C
Dedham’s fallback position is that the district court committed legal error by failing to appreciate that, in this case, the mere presence of
actual
contamination itself constituted a threat causing Dedham to incur response costs for which Cumberland was liable. This argument is entirely specious. Appellant does not elucidate any plausible theorem as to how or why Cumberland should be liable for response costs expended to deal with threats signalled by actual contamination when Cumberland, during the first trial, was fully exonerated from legal responsibility for that contamination.
See EMA,
889 F.2d at 1149;
First Trial Op.,
689 F.Supp. at 1235.
At any rate, appellant’s plaint boils down to little more than thinly veiled dissatisfaction with the district court’s factual findings. It neither demonstrates an error of law (as opposed to a disagreement over the facts) nor raises a question of jurisprudence sufficient to call the legal underpinnings of the court’s factbound findings into legitimate question. That ends the matter. “The ‘clearly erroneous’ rule cannot be evaded by the simple expedient of creative relabelling.”
Reliance Steel,
880 F.2d at 577. Hence, we decline appellant’s invitation that we play what amounts to a game of charades.
IV. THE MERITS
We come, then, to the ultimate question: were Judge Skinner’s findings of fact at the second bench trial clearly erroneous? Appellant urges us to answer this question in the affirmative, asserting that the court erred grievously in certain key respects. We are unconvinced.
On the merits, appellant's foray implicates the core of the district court’s decision. The court found that appellant only became aware that Cumberland may have released VOCs (and, thus, only became aware of the problem posed by a threat, as contrasted to the problem posed by actual contamination) in 1981.
Second Trial Op.,
770 F.Supp. at 42. In consequence, the court found that appellant could not have begun responding to a threat before that time — well after the treatment plant had been designed and appellant’s overall response paradigm had been formulated. Hence, appellant did not incur any costs to
guard against threats of future harm. We believe these findings are supportable.
The district court’s characterization of the big-ticket item — the treatment plant— as a response to actual contamination, rather than to threats, is completely plausible. That this expenditure, and appellant’s other responses, were not undertaken to deal with threats is strikingly apparent from appellant’s actions — or, more appropriately put, its inaction — in and after 1982, when it incorrectly identified Cumberland as the chief culprit in respect to the existing contamination. Despite the identification and the fact that Cumberland’s facility remained in operation, Dedham changed nothing of substance in its response paradigm. To cite but a few examples, there was no evidence that the new information was considered in making the final design decisions for the treatment plant or that the plant design was altered in any material way.
See Second Trial Op.,
770 F.Supp. at 42 (“The two-stage treatment plant recommended by Metcalf & Eddy in 1980 is substantially what was built.”). Appellant’s own documents state that it retained consultants to search for the sources of the pollution, not to assuage a threat of future harm, but exclusively in order to “recover damages from the polluters, if they can be identified.” Such litigation-related expenses are, of course, not compensable as response costs incurred by private parties under CERCLA § 107.
See, e.g., Leonard Partnership v. Town of Chenango,
779 F.Supp. 223, 229-30 (N.D.N.Y. 1991);
Cook v. Rockwell Int’l Corp.,
755 F.Supp. 1468, 1476 (D.Colo.1991);
Regan v. Cherry Corp.,
706 F.Supp. 145, 149 (D.R.I. 1989).
Although we choose to eschew a complete litany of the evidence underbracing the disputed finding, we deem three other points worthy of mention. (1) The record strongly suggests that G & M’s tests and reports, the earliest of which took place in 1982, provided the first credible basis for recognizing that Cumberland-spawned contaminants could pose a threat to Dedham. (2) Insofar as the record reveals, Dedham never acted to follow up its consultants’ recommendations for additional tests that would have served, among other things, to determine an appropriate response to threatened releases and threatened contamination. (3) The district court’s subsidiary findings in connection with the issue of causation, such as its finding that appellant’s decision to build the treatment plant was merely a response to the actual contamination of its well field and not in any way a response to a threat of events yet to come, were based largely upon assessments of witness credibility. Such assessments deserve substantial respect.
See Anthony v. Sundlun,
952 F.2d 603, 606 (1st Cir.1991) (an appellate court “ought not to disturb supportable findings, based on witness credibility, made by a trial judge who has seen and heard the witnesses at first hand”);
accord Anderson v. City of Bessemer City,
470 U.S. 564, 575, 105 S.Ct. 1504, 1512, 84 L.Ed.2d 518 (1985);
Scarpa v. Murphy,
806 F.2d 326, 328 (1st Cir.1986). In this tangled case, we see no valid reason why we should second-guess the trier’s on-the-spot determinations anent the credibility of the parties’ witnesses.
In the face of this substantial support for the trial court’s findings on the core issue, appellant tells us that the findings are nevertheless undermined because the court was bound to conclude that appellant knew of the “Cumberland problem” in 1979. Appellant’s tale has three foci. We consider them sequentially.
1.
Appellant places great weight on surface-water testing done in 1979. However, the evidence reflects a legitimate question about whether the testing was conducted in
sufficient proximity to Cumberland’s property to alert appellant at that early date to the possibility of a threat emanating from Cumberland’s facility (or, indeed, the possibility of any threat at all, wheresoever originating). Moreover, the district court received evidence, both testimonial and documentary, indicating that, even after conducting the 1979 tests, appellant complained to DEQE about the agency’s failure to locate the source of the contamination. In these, and other, ways, the record supports conflicting inferences about (a) when appellant first became threat-conscious and (b) when its suspicions first focused on Cumberland.
Appellant asserts that any amphiboly in the record was clarified by the testimony of its witnesses, George Johnstone and Gregory Grimes. Johnstone, a member of Ded-ham's senior management team, testified as to Dedham's intent. Contrary to appellant's importunings, we do not believe that the trial court was compelled to accept Johnstone's self-interested testimony. See Anthony, 952 F.2d at 606 ("what an actor says is not conclusive on a state-of-mind issue"). Grimes, Dedham's production superintendent, testified that he found sixteen parts per billion of 1, 1, 1-trichioro-ethane in a surface-water sample taken in 1979 from a stream some distance below Cumberland's property. Leaving to one side the fact that the sample was taken several hundred feet away from Cumberland's property, neither Grimes nor any other witness testified that this discovery led Dedham to believe, in 1979 or 1980, that contaminants had been released in the course of Cumberland's operations.
We will not churn the waters. When the evidence supports conflicting inferences, the district court’s choice from among the several inferences cannot be clearly erroneous.
See Anderson,
470 U.S. at 574, 105 S.Ct. at 1511;
Jackson v. Harvard Univ.,
900 F.2d 464, 466 (1st Cir.),
cert. denied,
- U.S. -, 111 S.Ct. 137, 112 L.Ed.2d 104 (1990);
Keyes v. Secretary of the Navy,
853 F.2d 1016, 1020 (1st Cir.1988). This is such a case.
2.
The next arrow in appellant’s quiver consists of a stipulation entered prior to the first trial. Appellant reads this stipulation as containing an admission by Cumberland that appellant’s testing of runoff from Cumberland’s land showed the presence of VOCs in 1979 or thereabouts. The fly in the ointment is that the district court did not accept either appellant’s interpretation of the stipulation or the inference appellant drew therefrom. We discern no error.
It is apodictic that a stipulation should be read with an eye toward effectuating the parties’ manifested intentions.
See Washington Hosp. v. White,
889 F.2d 1294, 1299 (3d Cir.1989);
Wo Co. v. Benjamin Franklin Corp.,
562 F.2d 1339, 1344 (1st Cir.1977). The stipulation to which appellant alludes, fairly read, represents no more than the parties’ agreement about the provenance and results of certain chemical tests.
It requires a forced reading to interpret the stipulation’s descriptions of the locations from which the test samples were taken as an admission concerning what Dedham knew and when it appreciated the implications of facts within its ken.
Because the stipulation at most faintly suggests, rather than compels, appellant’s interpretation, the district court’s refusal to accord it decretory significance cannot be faulted. After all, “ambiguous provisions in [a] stipulation must be interpreted by the factfinder, here the district court, as an initial matter.”
Washington Hosp.,
889 F.2d at 1299.
3.
Appellant’s third argument seizes upon a sentence in our opinion disposing of the earlier merits appeal. There, Judge
Bownes wrote: Based upon a survey it made of the surrounding surface waters, Dedham Water believed [in 1979] that Cumberland Farms was the source of the contamination of White Lodge Wells # 3 and # 4.”
EMA,
889 F.2d at 1148. Appellant asserts that this statement constitutes a finding which the district court was not at liberty to ignore. We disagree. The statement-which appears only in a section of the panel opinion entitled “BACKGROUND”-does not reflect an adjudication of fact by the
EMA
panel.
It is elementary that the “law of the case” doctrine, upon which appellant relies, ordinarily applies to matters of law, not to matters of evidence.
See, e.g., Gospel Army v. Los Angeles,
331 U.S. 543, 548, 67 S.Ct. 1428, 1430, 91 L.Ed. 1162 (1947). The quoted sentence does not fit in that category. Taken in context, the sentence was a passing reference to a matter not then in issue rather than a considered response to a zoetic question of law. Even a cursory reading of the panel’s opinion adequately evinces that its recital of the factual background was undertaken simply to set the stage for understanding the legal issues thereafter discussed and decided in the appeal.
To be sure, in a very select class of cases, the court of appeals may actually find facts-and the facts so found may come under the law-of-the-case umbrella.
See Heathcoat ¶. Potts,
905 F.2d 367, 370 (11th Cir.1990). The statement relied upon by appellant is not within the narrow confines of this select class. Absent special circumstances, not present here, appellate factfinding is permissible only when no other resolution of a factbound question would, on the compiled record, be sustainable.
See, e.g., LaRoche v. United States,
779 F.2d 1372, 1377 (8th Cir.1985);
In re Southern States Motor Inns, Inc.,
709 F.2d 647, 653 n. 11 (11th Cir.1983),
cert. denied,
465 U.S. 1022, 104 S.Ct. 1275, 79 L.Ed.2d 680 (1984);
Patterson v. Greenwood School Dist. 50,
696 F.2d 293, 295 (4th Cir.1982). The question of what Ded-ham knew and believed in 1979 does not come close to meeting so rigorous a criterion.
Because the factual issue to which appellant gestures was merely discussed, not decided, in the earlier appeal, the district court was not bound to accept the proposition.
See Schultz v. Onan Corp.,
737 F.2d 339, 345 (3d Cir.1984) (“law of the case” doctrine only applies to issues actually decided or decided by necessary implication);
Riley v. MEBA Pension Trust,
586 F.2d 968, 970 (2d Cir.1978) (a court need not adopt as “law of the case” a proposition that was merely assumed, rather than decided, in a prior appeal);
cf.
cases
supra
p. 459 (discussing non-binding nature of dictum in judicial opinions).
In a slightly different vein, appellant raises another set of issues which, directly or indirectly, constitute an assault upon the legal standard of causation set out in our opinion resolving the earlier merits appeal. We are not free to reconsider that standard, even if we were inclined to do so.
See Fournier v. Best Western Treasure Island Resort,
962 F.2d 126, 127 (1st Cir.1992) (in a multi-panel circuit, newly constituted panels are generally bound by prior panel decisions);
Metcalf & Eddy, Inc. v. Puerto Rico Aqueduct and Sewer Auth.,
945 F.2d 10, 12 (1st Cir.1991) (same),
cert. granted, - U.S. -, 112 S.Ct. 1290, 117
L.Ed.2d 514 (1992). Those contentions not foreclosed, some of which are cloaked in mistake-of-law raiment, collectively comprise nothing more than a sustained-and ultimately ineffectual-bombardment of the district court’s factual findings.
V. CONCLUSION
We need go no further. The task which we set for the district court after the initial trial was predominantly factbound.
See EMA,
889 F.2d at 1154, 1157. In the course of the proceedings on remand, the court competently performed that task. In so doing, it determined, principally as a matter of fact, that appellant had not proven an indispensable element of its remaining claim: causation. Mindful that a party alleging causation must carry the devoir of persuasion with respect to the issue,
see, e.g., Swift v. United States,
866 F.2d 507, 509 (1st Cir.1989), and cognizant of the deferential standard of review that Rule 52(a) imports, we cannot say that the court was clearly mistaken either in reaching this result or in making the subsidiary findings of fact upon which the result rested.
Because the district court’s findings are satisfactorily supported by the evidence and free from legal error, the judgment below is velivolant. Consequently, our voyage ends here.
Affirmed.