This is an appeal from an interlocutory decree of the District Court for the Southern District of New York, Archie O. Dawson, J., sitting in admiralty, by which the respondent, American Tankers Corporation (American), was adjudged liable for damage resulting from the contamination of fluid cargo owned by libellant, Esso Standard Oil Company (Esso). The district court found that the contamination resulted from the negligence of American’s employees on board the respondent tanker, the SS. Kaposia, in the course of dis*487charging the cargo to Esso’s terminal at Harkness Point, Philadelphia on July 25, 1951. The only questions presented for our decision are whether certain findings of fact, which are basic to the decision and which rest largely on the district court’s evaluation of expert testimony, are clearly erroneous within the meaning of Rule 52(a) of the Federal Rules of Civil Procedure, 28 U.S.C.A. We find that none of the contested findings is clearly erroneous and consequently we affirm the decision below.
It was not disputed here or in the district court that the contamination occurred through the intermixture of Esso Gasoline with Essoheat Medium, a heating oil, in the heating oil lines and shore tanks in the course of the discharge of both cargoes from the Ka-posia. Nor is it disputed that the in-termixture could only have occurred through the fault of either libellant’s or respondent’s employees. Under the charter the liability of respondent is limited to injury to cargo occurring through its own fault or the fault of its servants before the cargo passed the vessel’s permanent hose connections in the course of discharge.1 The district court found that the contamination did occur through the fault of respondent’s employees on the Kaposia as a result of the improper opening of four valves to permit the cross-connection of the cargo lines servicing the separate ship’s tanks containing heating oil and gasoline. Before proceeding to examine respondent’s contentions we must state very briefly the relevant physical structure of the Kaposia and the receiving structure at the Harkness Point terminal.
The Kaposia was a modified T-2 tanker, with three cargo systems known as the port, center and starboard systems. Each system was served by a separate pump, and each pump fed a separate permanent discharge hose on deck. Although these systems were separate, they could be cross-connected in two ways: by cross-over valves in the pump room; and by cross-overs at tank 5 from the starboard to the center system, and at tank 7 from the center to the port system. The district court apparently found that the record would only support the possibility of intermix-ture on the ship resulting from the opening of the cross-overs at tanks 5 and 7, and consequently we will not discuss further the possibility of contamination as a result of a cross-over at the pump room.
On shore, to accommodate the three-cargo tanker, there were three independent cargo lines. The receiving end of each of these lines was composed of flexible hose which was connected to the ship’s lines. Each of these hoses led directly from the deck towards a dock of similar elevation and connected with a pipeline. Each pipeline on the dock proceeded inshore for a few feet and then made a right-angle turn to proceed parallel to the shore line. Immediately after this turn each line contained a swing-type check-valve, whose purpose was to retard any backflow from the shore lines to the vessel. Inshore of each check-valve a small take-off or sampling line ran off the main line to permit sampling of the product in the main lines. Each main line then ran several feet further parallel to the shore line, and ended. This piece of straight pipe from a point after the sampling line to the closed end was referred to at trial as a “header.” Discharge of the product entering each of the headers was achieved by nine branch lines from each of the three headers running vertically downward from the elevated dock and connecting to a single set of nine lines which once again proceeded inshore at right angles to the shore line and terminated in field storage tanks. In the headers at the top of each of the nine branch lines there was a plug valve, the function of which is to control the flow into the branches. By opening only one of these valves on a header, cargo could be directed to the appropriate shore line *488and tank. It should be noted that although the three header systems were basically independent, they could be cross-connected, since each header was joined to each shore line by vertical branch pipes. Thus, for example, if on both the number 1 header and the number 2 header the branch lines to the first shore line and tank were open, the two headers would be cross-connected through that shore line. The district court found that the contamination in suit could have occurred on shore only by cross-connection of two otherwise independent headers as a result of the improper opening of a plug valve by libellant’s employees.
During the discharge in question the Kaposia’s tanks containing gasoline were serviced by the ship’s starboard pump, and the cargo was discharged into the number 3 header on shore. The tanks containing heating oil were serviced by the port pump, and the heating oil was discharged into the number 1 header. The gasoline and heating oil were distributed to their respective shore lines and tanks by the opening of the appropriate plug valve on each header. In the course of discharge, Esso’s employees took samples from the respective sampling lines every five minutes. The contamination of the heating oil was first discovered by making one of these tests while both products were flowing. At 6:50 P.M. on July 25, 1951, after both products had been flowing for some hours, the test taken at the number 1 header (heating oil) disclosed that the heating oil was contaminated with gasoline, and very shortly thereafter operations were closed down on shore and on the ship. It is not disputed that subsequent tests established that before and after the contamination the cargo still in the Kaposia’s tanks was not contaminated. With this summary we can proceed to the examination of respondent’s contentions.
At trial there was no direct testimony to the effect that improper valves were actually open either on the Kaposia or on shore. The district court reached its decision by finding first the locations at which and the conditions under which contamination was discovered. Then, by resolving a conflict in expert engineering testimony, it determined that from these preliminary facts it could be deduced as a necessary inference that the contamination must have occurred on board the Kaposia.
We shall consider first respondent’s attack upon the validity of this inference, insofar as the inference was based on the undisputed finding of a contaminated sample at the sampling line on the number 1 header. Respondent advances several arguments. The first is that the district court was clearly erroneous in finding that the physical location of the sampling line was at a point between the ship and the header before the line entered the header. If there were error in this determination it would necessarily infect the derivative inference; but the assertion is without support in the record or in the opinion below. That the court clearly understood the physical location of the line is made amply clear by its opinion:
“Upon reaching the No. 1 header the gasoline, it is alleged, would proceed back down the No. 1 line * * and pass through the sample house and by the check valve and on to the ship’s flange.” 148 F.Supp. at page 902.
Indeed, at trial the court noted the possibility of this confusion, and expressly clarified its use of the word “header.” The argument is without merit.
Respondent next insists that the district court erred in its finding that from the discovery of contamination at the sampling line on the number 1 header contamination on board ship could properly be inferred. But respondent mistakes the finality which attaches to the finding that the inference was justified. It is elementary that if the trial court in deciding a contested issue of fact is confronted with a substantial issue of the credibility and the weight to be attached to conflicting oral testimony, its findings are not to be reversed un*489less they are clearly erroneous. Federal Rules of Civil Procedure 52(a); see, e. g., Iravani Mottaghi v. Barkey Importing Co., 2 Cir., 1957, 244 F.2d 238, 248, certiorari denied 1957, 354 U.S. 939, 77 S.Ct. 1402, 1 L.Ed.2d 1538; Lee Dong Sep v. Dulles, 2 Cir., 1955, 220 F.2d 264; Orvis v. Higgins, 2 Cir., 1950, 180 F.2d 537, certiorari denied, 1950, 340 U.S. 810, 71 S.Ct. 37, 95 L.Ed. 595; Broadcast Music, Inc. v. Havana Madrid Restaurant Corp., 2 Cir., 1949, 175 F.2d 77. This is not the kind of case in which the contested finding is based
“ * * * not on facts to which a witness testified orally, but only on secondary or derivative inferences from the facts which the trial judge directly inferred from such testimony.” American Tobacco Co. v. The Katingo Hadjipatera, 2 Cir., 1951, 194 F.2d 449, 451, certiorari denied, 1952, 343 U.S. 978, 72 S.Ct. 1076, 96 L.Ed. 1370.
and we may therefore not disregard such a finding merely because other rational inferences are available from the undisputed facts. See E. F. Drew & Co. v. Reinhard, 2 Cir., 1948, 170 F.2d 679, 684. However positive the laws of physics may otherwise appear, in this case the physical behavior of particular fluids in particular pipes was a contested issue of fact as to which competent and conflicting expert oral testimony was heard and evaluated by the district court. We are not aware of any principle which requires such testimony to be treated differently in this court from any oral testimony on an issue of fact, and such a finding upon expert testimony will not be reversed unless this court “is left with the definite and firm conviction that a mistake has been committed.” United States v. United States Gypsum Co., 1948, 333 U.S. 364, 395, 68 S.Ct. 525, 542, 92 L.Ed. 746; see Gindorff v. Prince, 2 Cir., 1951, 189 F.2d 897.
We are left with no such conviction. Contrary to respondent’s contention, the district court did not base its inference of contamination on board ship upon a finding that had the contamination occurred on shore, no contamination would have been discovered at the sampling line on the number 1 header. The opinion below clearly states its premise that
“The substantial quantity of gasoline detected in this line could not have been a back flow from the No. 1 header.” (148 F.Supp. at page 902.) (Emphasis added.)
This finding is amply supported by the statements of respondent’s expert, Pickering, and libellant’s rebuttal expert witness, Burde. The crucial issue was clearly stated to Pickering. In response to a direct question whether if contamination occurred on shore gasoline would be detected in a heating oil sample at the number 1 header sampling line under flow conditions he replied:
“I don’t believe I could answer that question fully. I could say yes to that part that it would be in the header. But if it were in the header, eventually I think it would show up in a test in the header * * * I don’t know if it would take one minute or five minutes or ten minutes. * * *”
Pickering admitted that he was an expert on tankers, and not as familiar with the design and functioning of shore installations and that “I always seem to feel that the ship was right.” Burde, who was on libellant’s technical operating staff, and who assisted in the design of installations on ships and on. shore, testified affirmatively that no such discovery would be made under conditions of flow. Since Marsden, Esso'S yard supervisor of the Harkness Point terminal, testified and the district court found that the samples taken disclosed 95% gasoline in the heating oil lines, the court’s conclusion that the finding of a substantial sample at the number 1 header sampling line could not have been the result of back flow but must indicate contamination on the ship, on this record, is not only not clearly erroneous, but is amply supported by the evidence.
Respondent similarly argues that no inference of the occurrence of contamination aboard ship can be drawn from. *490the discovery of contamination at the ship’s flanges where the shore lines were attached. The district court based this conclusion on two additional facts: First, that “a back flow of gasoline would have had to overcome the flow of heating oil coming from the ship toward the shore”; second, that “any backflow would have been stopped at the check valve unless it was malfunctioning.” Respondent urges that both reasons are “invalid.”
Respondent points out that at the time the flange sample in question was taken at 10:30 P.M. the flow of heating oil had been closed down for more than three hours. But this is not in conflict with the district court’s findings. It merely emphasizes what is clear enough from the chronology set out in the opinion, that the court believed the testimony of Marsden and Steumpfle that, after the discovery of the contamination, all plug valves in the heating oil system were closed down and thereafter were never inadvertently opened. Therefore, unless the contamination at the ship’s flange occurred before the positive closing of the plug valve on the number 1 header — in other words, while heating oil was flowing — it would have been prevented by both the check valve and the plug valve, and would have been virtually impossible.
Apparently in recognition of this improbability, respondent bases its claim on an asserted change of circumstances resulting in gasoline being forced into the heating oil header and back to the ship by the action of Esso's employees, while the ship’s port pump was shut down. The crux of the argument is the assertion that Marsden testified that after the discovery of gasoline contamination at the sample line on the heating oil header, he deliberately directed the pumping of gasoline through the heating oil header by ordering the ship’s starboard (gasoline) pump to be started, and by closing the heating oil plug valve on the heating oil header, and opening the gasoline plug valve. If this had been done there would of course have been a cross-connection of the heating oil and gasoline headers, and there is some evidence in the record to support the assertion that such a cross-connection while the port, or heating oil pump was shut down could have resulted in contamination at the ship’s flanges.
The weakness in respondent’s assertion is that the record discloses no such testimony by Marsden. While Marsden did testify to the opening of the gasoline plug valve on the heating oil header and the closing of the heating oil plug valve in order to clear the heating oil header of contaminated cargo, it is plain that he testified that the ship’s port (heating oil) pump was used. On the very page of the record referred to by counsel for respondent, Marsden testified that shortly after pumping was started the heating oil header again contained good heating oil, and that as a result he once again switched the plug valves so as to send the heating oil to the appropriate heating oil shore line; that very shortly thereafter the product in the heating oil header was again contaminated, and that he then shut down and awaited the arrival of Carlyle, Esso’s chemist. It is obvious that this sequence could not have occurred through the use of the ship’s starboard (gasoline) pump, since it was agreed that that pump and the gasoline itself were not contaminated. Therefore respondent’s assertion that while the port pump was shut down Marsden deliberately caused the flooding of the heating oil header with gasoline under sufficient pressure to account for contamination at the ship’s flanges is entirely without support.
Furthermore, respondent’s key witness, Van Ostrom, Chief Mate of the Kaposia, who was on watch when the contamination occurred, appears to have contradicted respondent’s assertion. Van Ostrom testified: (1) that although all pumping was shut down when the contamination was found, the starboard pump was started some thirty minutes after the discovery and pumped continuously thereafter; (2) that about an hour and a half after the discovery of contamination he agreed with Marsden *491to assist him in the “displacing” of the port or heating oil line. Had Marsden’s efforts depended on flooding the heating oil header with gasoline from the starboard pump he would not have needed Van Ostrom’s assistance, since the starboard pump was already in operation. Indeed, both Marsden and Van Ostrom testified that the operation carried out was the “displacing” of the port line, an operation which, as described by respondent’s witness, Vega, the Kaposia’s second mate, would have required the use of the port pump. The court’s ultimate finding was supported by the testimony of respondent’s own expert that under such conditions of flow, contamination originating on shore could not account for the discovery of a contaminated sample at the ship’s flange. Thus, the inference of contamination on the ship arising from the fact of substantial contamination at the ship’s flanges is on the evidence more obviously justified than the inference arising from the discovery of similar contamination at the sampling line.
Respondent’s final attack on the inference that the contamination occurred on the ship is based upon the condition of the check valve on the number 1 header line. Although there was conflicting testimony at the trial as to the purpose of swing type check valves, respondent’s witnesses contending that a properly functioning valve is apt to leak or “weep” and that such valves are properly employed only to retard and not to stop backflow, the issue is not significant here for several reasons. First, as we have said, the evidence supports the finding that at all relevant times when the gasoline plug valve on the number one header was either concededly or arguably not closed, the port pump of the Kaposia was in operation, so that the significance of the check value was greatly minimized. Second, there was direct and not unreliable evidence to support the propriety of the use of the check valve to stop backflow. Third, the district court’s finding of an inference of shipboard contamination from “substantial contamination” at the ship’s flange is not in conflict with respondent’s characterization of these valves as “weepers.” For all these reasons respondent’s contention that there was a failure of proof that the contested valve in fact was functioning properly is insubstantial. It is further weakened by the testimony of Marsden that the valve functioned properly and the testimony of Steumpfle, Esso’s gauger and checker at Harkness Point, that although the valve was not replaced it was maintained by greasing and packing.
Respondent also contends that the district court’s finding that there was a contaminated sample taken from the Ka-posia’s port pump was clearly erroneous. It apparently does not challenge the accompanying inference that the contamination resulted from negligence on board the Kaposia if the sample was found. Ordinarily we would not find it necessary to decide this issue, since we have found that the inference of the occurrence of contamination on board the Kaposia was justified from the finding of the contaminated header and flange samples, and since the absence of such a sample would be of doubtful significance, conditions in the Kaposia’s pump room having been at all times in respondent’s control. However, we are aware from its own statement in the record that the’ district court attached crucial significance to the finding of such a sample, and it is possible that if it had decided that this finding was not justified, the weight it attached to other evidence might have been significantly altered.
It is not contested that at about 10:55 P.M. on the night of the contamination, an uncontaminated sample was taken at the port pump. The contest centers on the finding that contamination was discovered in an earlier sample taken at 9:05 P.M. by Carlyle, Esso’s chemist. The oral evidence as to the taking of the earlier sample is sharply in conflict. Van Ostrom, the Kaposia’s mate, testified that Marsden, in his presence, took such a sample, which proved to be un*492contaminated. Marsden denied that he took such a sample, or even entered the Kaposia’s pump room, but asserted that Carlyle did. Carlyle claimed that he took such a sample in Van Ostrom’s presence, and that the sample was contaminated, all of which Van Ostrom denied. Dievler, Esso’s terminal superintendent, testified that Carlyle told him of the contaminated sample shortly after it was taken, and showed it to him. It is obvious that if this were all, the finding that Carlyle took such a sample could not be upset. But respondent urges that the oral testimony of Carlyle at the trial five years after the occurrence of the event is rendered inherently unreliable by the contents of a written report prepared by him within two days of the occurrence of the contamination at the request of Heussner, the terminal manager.
Respondent argues that when oral testimony is rendered extremely doubtful by conflicting documentary evidence, this court may disregard the district court’s finding based on the oral testimony and substitute its own. United States v. United States Gypsum Co., supra, 333 U.S. at page 396, 68 S.Ct. at page 542. However, we do not find that Carlyle’s report as explained at the trial so conflicted with his oral testimony as to leave us with a definite and firm conviction that a mistake was made below. First, Dievler’s testimony and the practically undisputed significance of the flange sample under the conditions the district court found to exist lend independent support to Carlyle’s testimony. Second, although the written report stated only that “A sample was taken at the port pump in the pump room and the product was good * * * ” Carlyle explained that he had delivered a detailed report, including an analysis of the contaminated port pump sample, to Heus-sner on the night the contamination occurred. Although no such report was produced, Heussner’s death prior to trial made further inspection of the fate of the earlier report impossible. This explanation is on the one hand given credence by the summary character of the final report, and on the other weakened by the mention in the final report of a “sample * * * taken * * in your presence,” which stimulus to Heussner’s recollection seems inconsistent with a prior detailed report by Carlyle. That the making of the earlier report would have been an appropriate procedure under the circumstances seems undisputed, since elaborate numerical an-alyses were required by Heussner in order for him to decide, on the evening of the contamination, what disposition was to be made of contaminated cargo in the shore tanks and lines. On this evidence we do not feel that Carlyle’s oral testimony is rendered so doubtful that the district court’s belief of it should be disregarded.2
Respondent urges for his last proposition that the libellant failed to sustain its burden of proof. In the light of what we have already said, it is enough to point out that respondent’s reliance on the case of Pennsylvania R. R. Co. v. Chamberlain, 1933, 288 U.S. 333, 53 S.Ct. 391, 77 L.Ed. 819 is misplaced. The question there was whether it was appropriate for the district court to direct a verdict against a plaintiff whose only witness testified as to facts from which equal inferences of the defendant’s and the plaintiff’s decedent’s responsibility could be drawn. The question was one upon which no expert testimony could be offered. It was held that since even on the plaintiff’s version of the facts neither inference could be said to be rationally preferable, and since the plaintiff’s version was itself directly contradicted by every eyewitness, no case had been made out for the jury.
In summary, the propriety of the inference of shipboard contamination was an appropriate subject of expert testimony, and that testimony amply supports the finding that the facts do *493not support an equal inference that contamination occurred on shore. Similarly, the finding upon expert testimony, that the absence of contamination in the ship’s heating oil tanks did not prove that the contamination occurred elsewhere than on the ship, is not clearly erroneous.
Affirmed.
. The relevant portions of the charter party under which the cargo was carried are set forth in the opinion below, 148 F.Supp. at page 900.
. The same considerations dispose of the appellants’ claim that Esso’s letter to American, of June 18, 1952, describing an uncontaminated sample, constituted an extra-judicial admission.