WISDOM, Circuit Judge.

This appeal is one of three appeals, argued the same day, in which this Court is asked to reverse judgment for the Panama Canal Company in an action by a shipowner for damages to a vessel caused by its striking a bank of the Canal while it was under the compulsory pilotage of pilots employed by the Canal Company.

In the instant case the S./T. Andros Venture, a Canadian flag vessel, struck the banks of the Canal in the narrow Gaillard Cut damaging its hull and cargo. Andros Shipping Co., Ltd., owner of the vessel, filed a libel against the Canal Company for hull damage. The Empresa Nacional Del Petroleo, the Chilean cargo owner, sued both Andros Shipping and The Canal Company for cargo damage. The suits were consolidated for trial below and on appeal. The district judge held that uncontrollable hydrodynamic forces, bank suction, caused the accident, and rejected the claims against the Canal Company based on the pilot's alleged negligence. As for the cargo damage, the district court applied the Pennsylvania doctrine in holding the Andros Shipping Company liable. 184 F. Supp. 246. We affirm.

I.

The Andros Venture arrived at the Atlantic entrance to the Panama Canal at noon December 22, 1955, to make a southbound transit of the Canal. Esso Export Corporation had chartered the vessel to carry a cargo of crude oil from Puerto La Cruz, Venezuela to Valparaiso, Chile. The vessel carried 27,794 tons of oil and had an observed mean draft of 33 feet 11½ inches, although the calculated draft based on the weight of the cargo was 34 feet 5½ inches. The ship is a single-screw, steam turbine tanker built in Quebec in 1953. Its overall length is 624.8 feet, its beam 84.2 feet, and its mean authorized salt water draft 33 feet 11 7/8 inches. It weighs 17,845 gross tons and 13,280 net tons.

Pilots Hector Grant and William Calcutt boarded the Andros Venture in Limon Bay at five in the morning to take the ship through the Canal. Grant had been a Canal pilot for fifteen years, Calcutt for twenty-two years. The master did not inform either pilot that the vessel was overdraft, had a sag, and had sustained hull damage some time before arriving at the Canal. Calcutt took the conn of the vessel for the first leg of the trip from Limon Bay to Gamboa. He encountered some difficulty controlling the vessel, and on its approach to the north end of Gatun Locks the engines were stopped and a tug placed alongside the port bow. Traveling at four knots, the vessel took several sheers when Calcutt attempted to reduce speed. By use of the engine and rudder, and with the help of the tug, he was able to obtain control in the channel, which at that point is 500 feet wide. During this part of the transit it appeared that the ship was acting sluggishly. Calcutt testified that when he cut off the engines in Gamboa Reach "she just wouldn't answer the rudder" although she was still moving forward. In the Gatun Locks the Canal Company personnel reported an overdraft of several inches, and Pilot Grant requested the ship's master to give the vessel a six-inch drag. At Gamboa Reach Grant assumed the conn. He placed the tug San Pablo ahead of the Andros Venture on a 300-foot howser to assist it through Gaillard Cut. The vessel passed La Pita Signal Station at 1310 hours traveling at slightly under six knots toward the 300-foot channel of Cucaracha Reach, where serious trouble was lying in wait. The vessel sheered and struck the east bank then the west bank in the narrow Gaillard Cut, causing extensive damage to its hull and cargo. There is no satisfactory explanation for the vessel's sheers. The Court put them in the category of "those unexplainable incidents which frequently occur in the navigation of vessels". Grant died before the trial. At the hearing conducted by the Board of Local Inspectors the day after the accident, Grant testified that the cause of the accident was insufficient rudder power and the unusually slow response of the vessel's engines. The district court found:

"There are, of course, many possible explanations for a sheer. The irregularities of the Canal result in unbalanced forces requiring continuous corrective rudder action. Should the rudder, in the course of taking corrective action, happen to be tending to direct the vessel's head in the same direction as the bottom and bank suction forces created by such irregularities, a moderate sheer could develop. In this case, the possibly poor trim, the vessel's indented hull, her overdraft, would constitute aggravating and complicating factors. In brief, sheers will occur — the real problem of shiphandlers is to determine the appropriate corrective action, and then obtain its prompt and proper execution by ship personnel."

The court ruled that the actions of the pilot before and after the striking were proper.

II.

The appellant's principal contention is that the pilot, Grant, took radical and unnecessary action to break the initial sheer to starboard and failed to act quickly enough to prevent the second sheer to port that was the foreseeable consequence of his efforts to break the starboard sheer. Ensnarled in this question is the hydrodynamic phenomenon of bank suction and its tendency to throw a ship off course by causing sheer. Bank suction occurs when a large ship passing through a narrow channel with vertical sides veers from the centerline. The asymmetrical flow of water past the ship builds up the water level between the bow and the near bank with the result that the bow is forced away from that bank. At the same time the water level between the stern and the near bank falls below the normal water level, sucking the stern toward the bank. The net result is to cause the ship to sheer toward the opposite bank. The phenomenon is well known and is discussed in "The Panama Canal Pilots' Handbook". To counteract bank suction it is necessary to use a rudder setting that tends to turn the ship toward the near bank; this will tend to turn the bow in and hold the stern out from the bank. The correction requires a delicate hand, since overcorrection may head the ship toward the near bank, while undercorrection may cause the ship to swing rapidly toward the far bank.

The initial sheer was to the right. Grant ordered a hard, left rudder. The appellant argues that this order was wrong; the rudder should have been set to the right. Its expert witness, however, Anthony Suarez, testified that the proper corrective rudder setting would be to the left. A right rudder would have accentuated the sheer and driven the ship into the west bank. Andros Shipping also asserts that it was negligent for the pilot to hold the hard left rudder for two minutes while increasing the engine speed to half speed and while the tug was also pulling the bow away from the west bank, since he should have known that this would bring on the secondary sheer toward the east bank. The appellant has not shown, however, that these measures were not reasonable measures to prevent a bank striking after the first sheer. Grant had learned earlier in the trip that the ship maneuvered sluggishly. Guiding an eighty-four-foot-wide ship through a three hundred-foot-wide channel, Grant's judgment was that a more moderate correction would not suffice. He stated that even at the hard-left rudder setting the ship did not respond until he stepped up the speed. Pilot Calcutt corroborated Grant, testifying that holding the hard-left rudder setting for two minutes was "perfectly normal; nothing the matter with that."

When the ship broke out of the sheer to the right Pilot Grant set the rudder amidships and went back to a dead slow speed. As to this, Pilot Calcutt said: "Now, you can have a full left rudder on the ship and the ship starts to respond and he puts his wheel amidships, that doesn't mean the ship is going to steady out and go down the channel. She is going to carry that swing. I was, however, satisfied at that time. I never would have left the bridge if there was any doubt in my mind that she was not going to straighten out." He returned to the bridge ninety seconds later. At that time the ship was sheering to the left, Grant had changed the rudder to full right, the stern of the ship was then still to the west of the centerline, and the tug was pulling the bow to the right. This testimony supports the Canal Company's assertions that Grant was not negligent. Calcutt's judgment that the ship was straightening out backs up Grant's decision to put the wheel amidships after the sheer to the right was broken rather than to take immediate steps to prevent a sheer to the left. Calcutt's statement of facts, which is not contradicted by other testimony, indicates that little if any time elapsed after the ship began to sheer to the left before Grant acted to try to break that sheer.

In an appellate inquiry into a factual question of this nature it is not enough for the appellant to raise a doubt as to the accuracy of the trial judge's findings; the appellant must show that the findings are "clearly erroneous." McAllister v. United States of America, 1954, 348 U.S. 19, 75 S.Ct. 6, 99 L.Ed. 20. The decisions of a pilot in the delicate and hazardous task of navigating a large ship through the Panama Canal involve a matter of judgment. A court must avoid basing its decisions on hindsight, and it must make allowance for the legitimate differences in technique of various pilots. In this case there is little to show that the pilot acted negligently. The appellant relies on the fact that the Panama Canal Company's Pilots' Handbook and the appellant's expert both declare that moderation in efforts to correct a sheer is important; the Pilots' Handbook states that "the thing to do is to apply energy in opposite directions to both ends by keeping the tug at full speed ahead and backing the vessel's engines at full speed astern." It seemed clear to the trial judge, as it does to us, that in an area where innumerable factors may produce small but crucial differences no general principle can be followed blindly in every case. This the manual itself advises. Pilot Calcutt stated that he would never have attempted to break the sheer of a vessel the size of the Andros Venture by the backing operation described by the manual. Calcutt was the only experienced Canal pilot present on the ship at the time of the accident other than Grant, and by his testimony all of Grant's actions were proper. Pilots are members of such a closely bound, exclusive confraternity that in some cases it would not be unreasonable for a trial judge to raise an eyebrow when one pilot backs up the judgment of another. Here, however, none of the other witnesses present at the accident criticized Grant's actions. These witnesses included the regular officers aboard the appellant's ship. Captain Eynon, the master of the Andros Venture, stated specifically that he had no criticism of Grant and that he was satisfied that the pilot used due care. All of those present agreed that the allision was not caused by the fault of Grant. We hold that the trial judge was justified in relying on their testimony and in rejecting the contention that Grant should have avoided the accident by adhering to different general principles.

III.

In its brief the appellant argues strongly that the pilot was negligent in failing to employ a second tug when he knew of the vessel's tendency to sheer when slowing down and of the ineffectiveness of a single tug to control the vessel. Apparently the case was not tried on this theory, for there is little or no evidence to support it in the record. The appellant did show that the decision on employment of a tug or tugs was within the pilot's discretion and that a second tug was available. But no witness testified as to the advisability of employing two tugs to take the Andros Venture through the Gaillard Cut as a dead vessel in the situation that confronted Grant. There was testimony that the ship was sluggish and had a tendency to sheer, but it appears that vessels not infrequently tend to sheer during transit, and there is no indication that on the approach to the Gatun Locks the Andros Venture sheered more seriously than other vessels successfully taken through with a single tug. No attempt was made to show a practice of using two tugs. The only evidence on this point was submitted by the Panama Canal Company: during the period up to March 16, 1959, this was the only allision in 93 transits by 15 vessels of the same hull design as the Andros Venture. All of these transits were made in the same manner with a tug ahead on a hawser. In short, all that appellant offers is the evidence that this ship tended to sheer, that sheering represented a hazard, and that the pilot did not employ a second tug to reduce the danger; and of course, the fact that an accident did occur. But failure to take a precaution does not of itself constitute negligence, even when an accident later happens. It constitutes negligence only if a reasonable man would have taken the precaution in the prevailing circumstances. Prosser on Torts § 30 (2nd Ed. 1955); see United States v. Carroll Towing Co., 2 Cir., 1947, 159 F.2d 169, 173.

The instant case is distinguishable from (The Aurora Borealis) Panama Canal Co. v. Sociedad de Transportes Maritimos, 5 Cir., 1959, 272 F.2d 726 in which we affirmed the district court's holding that the pilot was negligent in not using two tugs. The Aurora Borealis was a Hog Islander, a type of ship noted for its unmaneuverability. We said: "This is not a case where there was one sudden, uncontrollable, unexpected sheer that forced the vessel into the bank. The whole transit of the Aurora through the Cut consisted of sheering, correcting a sheer, and then sheering the other way. While the tug was being made fast to the Aurora, it veered off course. While being towed, the Aurora veered first to port and then to starboard before the final port veer that caused it to strike the east bank. Long before the accident, this erratic course should have made it clear to a reasonably prudent pilot that he could not control the Aurora with a single tug." On the other hand, the Andros Venture was a newly built, "prestige" ship similar to fourteen other ships which transited the Cut without incident. Here, we cannot say that the pilots aboard the Andros Venture knew the vessel would sheer. We cannot say that, after the vessel sheered on its approach to Gatun Locks but responded (if sluggishly) to control, the pilots knew or should have known that such a sheer as occurred in the Gaillard Cut could not have been prevented or corrected with one tug ahead on a hawser.

The burden is on the appellant to come forward with evidence showing the pilot's negligence. This the appellant failed to do. The trial court correctly held that the pilot exercised due care on this matter, and we affirm its decision.

IV

In The Nonsuco and The Aurora Borealis we held that when an allision occurs in the Panama Canal the mere fact that the vessel is under compulsory pilotage is not enough in itself to make the doctrine of res ipsa loquitur applicable. Victorias Milling Co. v. Panama Canal Company, 5 Cir., 1959, 272 F.2d 216; Panama Canal Company v. Sociedad de Transportes Maritimos, S.A., 5 Cir., 1959, 272 F.2d 726. We added the caveat: "We do not say res ipsa loquitur is never applicable to a Canal bank-striking. For example, when an injury occurs while a vessel is in the locks of the Canal or when a vessel in the tow of two tugs strikes the side of the Canal, the vessel is so completely under the control of the tugs that it is reasonable to infer negligence. We simply hold that the doctrine of res ipsa loquitur is not available to the libelant in the general fact situation presented in this case." The appellant seeks to bring the case at bar within the scope of the caveat.

Res ipsa loquitur is a rule of evidence that allows an inference of negligence to be drawn when the circumstances indicate that an accident would not have occurred unless the defendant was negligent. Three conditions are necessary for the principles of res ipsa loquitur to apply: "(1) the accident must be of a kind which ordinarily does not occur in the absence of someone's negligence; (2) it must be caused by an agency or instrumentality within the exclusive control of the defendant; (3) it must not have been due to any voluntary action or contribution on the part of the plaintiff." Prosser on Torts § 42 (2d ed. 1955). The doctrine rests on a process of elimination. When one or more of these conditions does not exist, the rationale underlying the rule (the exclusion of all reasonable explanations for the accident other than the defendant's negligence) is punctured. The presence of the conditions is in each case a question of fact. In this case the necessary factual foundations are lacking.

First, it is by no means clear that this accident would not ordinarily have occurred in the absence of someone's negligence. Bank suction is a phenomenon known to create dangers to ships passing through a narrow reach. The Report of the David W. Taylor Model Basin, often cited in these Panama Canal cases, points out that in "some cases the sheer that results cannot be overcome by the rudder and the vessel strikes one of the banks." The dangers were increased in this case by the fact that the ship was large, heavily laden, and acting sluggishly. And the ship was passing through one of the narrowest cuts in the entire canal. It is undisputed that this transit involved certain potential hazards, creating a substantial danger that an accident might occur in the absence of any negligence. The district judge found that this was the case.

Second, the facts also fail to show the presence of the second condition necessary for the application of res ipsa loquitur: that the agency or instrumentality causing the accident be within the exclusive control of the defendant. To be sure, the Panama Canal Company pilot was at the conn when the Andros Venture struck the bank, but the ship was manned by the appellant's employees. The pilot stated that his orders were promptly carried out and that he had no criticism of the crew; but the regular officers of the Andros Venture also stated that they had no criticism of the actions of the pilot. If one is to disregard these statements and look for negligence, it is entirely possible that the fault lay with the ship's officers and crew. Even after the pilot came aboard, the officers and crew retained responsibility for many important functions in the operation of the ship. Had the pilot's orders been executed improperly, or had a member of the crew been otherwise negligent in the performance of his duties, that negligence surely could have caused the accident.

The division of actual control (as distinguished from the right to control) that exists when a pilot takes a live ship through the Canal and the evidence here disclosing a substantial probability that the allision was caused by other factors that the negligence of the pilots exclude this case from the scope of the caveat stated in Victorias Milling. As the district judge said, "Considering all of the evidence, certainly this Court cannot say that it is more probable than not the casualty was caused by the respondent Panama Canal Company."

V.

The trial judge found for the cargo owner on the ground that the Andros Venture violated statutory safety regulations and that such violations made the Pennsylvania doctrine applicable. In The Pennsylvania, 1874, 19 Wall. 125, 22 L. Ed. 148, the Supreme Court established the doctrine:

"[W]hen * * * a ship at the time of a collision is in actual violation of a statutory rule intended to prevent collisions, it is no more than a reasonable presumption that the fault, if not the sole cause, was at least a contributory cause of the disaster. In such a case the burden rests upon the ship of showing not merely that her fault might not have been one of the causes, or that it probably was not, but that it could not have been. Such a rule is necessary to enforce obedience to the mandate of the statute."

This Court has observed: "This rule is not a rule of liability. It creates a shifting of the burden of proof as to causation." Green v. Crow, 5 Cir., 1957, 243 F.2d 401. Andros Shipping Company failed to carry its burden of proving that the statutory faults were not causally related to the accident and could not have been related.

The Pennsylvania doctrine acts as a sanction to enforce compliance with regulations designed to promote maritime safety. It penalizes violators of the safety rules by imposing on them a severe burden of proof that may be decisive when the causes of an accident are unknown or in dispute. Frequent application of this rule to ship collisions and other cases make clear that this standard governs suits in admiralty. There has been a recent tendency to modify the rule in collision cases where one ship is manifestly negligent and the other has been guilty of minor infraction. Certain cases have found that the lesser violation could not have been a contributing cause to the accident. Other decisions have adopted a major-minor fault rule that allows the ship with the minor fault to recover notwithstanding its infraction. These cases, however, focus on the relative faults of the parties; they rest on the rationale that the ship guilty of the major violation should not be allowed to escape the full penalty rather than on the rationale that the ship guilty of only a minor fault ought not to be penalized. The accident of the Andros Venture was of the sort that the safety statutes were designed to prevent, and the cargo owners were certainly within the group of persons for whom such statutes were designed to afford protection. We hold that the trial judge was correct in ruling that the Pennsylvania doctrine was applicable to the instant case.

The evidence amply supports the trial judge's finding that the Andros Venture was guilty of statutory faults. Under the applicable Canadian and American law the ship was required to have as engineroom personnel a licensed first-class engineer, a second-class engineer, and at least, a fourth-class engineer in charge of each watch. The trial judge found, on sufficient evidence, that the ship had only two properly licensed engineers on board. The testimony shows that the engine room at one time executed orders which had never been given and at another time failed to execute an order which had been given. The trial judge also found that the appellant's evidence was inconsistent as to who was in charge of the engine room at the time of the accident, and that the vessel's Engine-room Bell Book had been altered in a manner never satisfactorily explained. Another serious statutory violation was the fact that the vessel was overdraft in violation of Canadian law and the International Load Line Convention. There was a maze of conflicting evidence as to the draft of the vessel at different points during the trip, but the evidence most favorable to Andros Shipping is that the recorded draft on arrival at the Canal was three-eighths of an inch over the maximum authorized draft, and it is undisputed that the vessel had a sag in the center which lowered the actual draft to several inches lower than the recorded draft. The trial judge found other faults of the vessel: hull irregularity, poor rudder design, overloading, failure to comply with the legal requirements for manning, and failure to have the propulsion engines operating at maximum ahead power. These deficiencies, which added up to unseaworthiness, in the findings below, provide the factual basis for the application of the legal standard of proof enunciated in The Pennsylvania.

The only remaining question is whether the appellant met the burden of proof imposed on it. We affirm the finding of the district court that it did not. It is significant that uncertainty exists as to the reasons causing the Andros Venture to sheer and to be unable to break the sheer without accident. It is an acceptable hypothesis that the explanation for this accident was the vessel's lack of maneuverability, and that the statutory faults of which the Andros Venture was guilty did represent the missing margin of safety. Unqualified personnel in the engine room may have caused delay in carrying out orders that would explain why the ship did not respond quickly. The overdraft unquestionably contributed to the sluggish behavior of the ship. These and other deficiencies account for the vessel's lack of maneuverability. The trial judge properly concluded that the appellant failed to show that its statutory faults "not only did not contribute to the disaster, but could not have done so."

The Court has considered all of the issues the appellant has raised, whether or not these are discussed in this opinion.

The judgment is

Affirmed.

* * * * *