STALEY, Circuit Judge.

Libellants, the owners of the tug Minerva, appeal from a decree in admiralty dismissing their libel, after the court had field the Minerva solely at fault 'for a collision between it and the tanker Merry ■Queen. They would have us decide that all the fault should be on the Merry Queen or, at least, that it was a mutual-fault collision.

At albout three o’clock on the afternoon of April 22, 1948, the Minerva, with the barge Interstate No. 4 made up to starboard, stern first, was making about four miles per hour almost due east down the Christiana River in ¡Wilmington, Delaware. She was just above the bend which is upriver from the confluence of the Christi-ana and Brandywine rivers. This bend requires almost a horseshoe turn to starboard for a descending vessel. ■ The Minerva was 55 feet long and had a beam of 15 feet. The Interstate No. 4 was 208 feet long, with a beam of about 39 feet. The two vessels were so lashed that approximately 150 feet of the barge extended beyond the stem oif the tug. The barge was light and had a 3-foot house on her stern. She had 9 feet of 'freeboard, which increased to 12 feet at the stern. The deck of the Minerva’s pilot house was 5 feet above the water. Therefore, made up as the tow was on this occasion, the view to starboard, the direction from which the Merry Queen came around the bend, was partially obstructed.

■ The Merry Queen was 155 feet long, with a beam of 33 feet. She was ascending the river, well to her right, at a/bout 3% knots. Within a quarter of a mile of the bend, at a point just above where the Brandywine meets the Christiana, she gave one long blast to indicate her approach to descending vessels.

Because of trees on t'he south shore, the view around the bend was obstructed. The day was clear, with an east wind] of about 15 miles an hour and an ebb tide of about D/2 knots. The channel at the point of collision lies toward the north bank and is about 300 feet wide.

The version of the collision, according to those on the Minerva, was that she was in mid-channel when the Merry Queen was sighted rounding the bend. Each vessel then signaled for a port-to-port passage. The Merry Queen, however, rounded in front of the barge, swerved to port, and rammed the Minerva, just aft of the latter’s port bow, causing quite extensive damage. The collision' occurred about mid-channel. At that time, the Minerva had turned about 45° to starboard from her straight course. Her engines were full speed astern, and she had lost almost all headway. She blew no whistle other than the passing signal, and those aboard heard no bend signal from the Merry Queen.

The trial judge, however, found the facts substantially as related below by the Merry Queen’s master, stating that his story was credible and was supported by the probabilities of the situation. The master testified that. he got no response to his bend signal, so he went on at half speed. *577 After proceeding about a thousand feet, the lookout shouted that a barge was coming down the river. At this time, the vessels were about fifteen hundred feet apart. Because of the way in which the tug and tow were made up, the tug was not visible, and those on the Merry Queen thought • that the Interstate No. 4 was a runaway barge. To avoid a collision, the Merry Queen pulled over to starboard, and, after some backing and filling, came to a stop parallel to and within about six inches of the American Car & Foundry Company’s concrete pier on the north bank. The master expected the barge to pass to his port. At this time, the Minerva was first sighted behind the barge. She was at a 45° angle, the barge’s stern (the front, as she was going) being on the right, and the tug and barge’s bow on the left, side of the river. The Merry Queen blew one blast for a port-to-port passage but heard no response from the Minerva. The latter was unable to control her tow around the bend and collided with the Merry Queen’s port bow about 20 feet aft. At the time of impact, the .Merry Queen’s engines were stopped, and she was dead in the water. The collision occurred approximately 30 feet from the north bank and in the upper part of the bend. The Merry Queen was forced backwards but not into the pier by the collision, and she suffered negligible damage.

We think the findings are amply supported. The district court said that "The Minerva was a little tug in charge of a big barge” and that, in attempting to make the bend to starboard, the ebb tide would have a tendency to set her stern to port, causing her to “skid” into the Merry Queen. In any event, the evidence supports the finding that the Minerva was in the middle or slightly on the left of the channel, certainly not on the right as required by the Narrow Channel Rule. 1 There can be no doubt that this channel is within that rule. Furthermore, it is not disputed that the Minerva did not sound a bend signal as required. 2 The district court concluded, and properly so, that this bend was one to which Rule V applied; consequently, the Minerva’s fault is clear.

Libellants recognize the respect we accord the findings of the trial judge in' admiralty, even though we sit de novo, 3 but they assert that the finding here is based on a theory not supported by the testimony of either side. It is said that the version given by the master of the Merry Queen is physically impossible, and that the trial judge rewrote the testimony to resolve the conflicts.

The first objection is that, according to libellants, the master testified that the Minerva was at an angle of 45° to the Merry Queen when she struck. Libellants then show that this could not be, if the Merry Queen was within six inches of the pier when the collision occurred. The one 45° position would put the Minerva’s port quarter up on the pier, and the other would put the barge completely through the Merry Queen. The district court, however, from the master’s testimony as a whole, was of the opinion' that he meant that the Minerva was at an angle of 45° with the upper reach of the channel. This view of the testimony would put the vessels in a nearly parallel position when they met and would avoid the impossibility pointed out by libellants. This finding would also explain why the Merry Queen was not forced against the pier. The impact would drive her backwards, not to starboard.

Our own examination of the testimony convinces us that the district court’s view is correct. The master had been questioned previously about the Minerva’s angle, and he referred to the Minerva as being at a 45 or 50° angle while still above the bend. As the district court said, if the Minerva were at a 45° angle to the channel, while still above the Merry Queen, unless she changed course quickly, she would be nearly parallel with the Merry Queen when she *578 struck, down where the bend becomes more acute.

Libellants point out what they consider to be further inconsistencies and impossibilities in the Merry Queen's story of the collision, but we think these arguments are directed to the wrong forum. They show only that the trier could well have found the facts favorably to libellants in the first place, not that a contrary finding is clearly erroneous, which it must be in order to be upset here. 4 This rule is particularly stringent when the findings are reached upon conflicting testimony and are based upon the credibility of witnesses whom the district court saw and heard. 5 Respondents 6 called only one witness, the Merry Queen’s master. They did not produce the helmsman' or the lookout, who, presumably, saw what happened. Citing Coyle Lines v. United States, 5 Cir., 195 F.2d 737, modified on other grounds, 1952, 198 F.2d 195 and 887, libellants say that we must presume that these men, if called, would, have testified unfavorably to respondents. The latter did show, however, that the helmsman was in some institution, following a nervous breakdown', and that the lookout was no longer in their employ, that he had been searched for but could not be found. This is enough to rebut the inference sought to be invoked. 7

We think, however, that libellants’ final point is sound and requires a division' of damages here. The testimony of the master of the Merry Queen supports the district court’s conclusion that she did not sound the bend signal within the distance required by the rule. 8 The district court found that she signaled when a quarter of a mile from the bend. This established the violation of the statutory rule. It was then up to respondents, under the doctrine of The Pennsylvania, 1873, 19 Wall. 125, 22 L.Ed. 148, to show not only that the violation might not have been a contributory cause of the collision, or that it probably was not, but that it could not have been. 9 The district court said that the signal was given “at a point which should have given the Minerva plenty of warning that a vessel was approaching.” We think that is not enough to sustain the rather heavy burden of avoidance placed upon respondents by The Pennsylvania. Each vessel was making roughly 350 feet per minute. Had the Merry Queen complied with the rule and signaled a quarter of a mile *579 sooner than she did, the Minerva would have had a little less than four additional minutes to take the necessary precautionary steps. Who can say that, given that additional time, those measures would not have been successful? The very fact that that question may properly be asked seems to us to show that the Merry Queen has not sustained her burden. 10

It could be argued that, since those on the Minerva did not hear the Merry Queen’s bend signal even when given a quarter of a mile too late, obviously they would not have heard it if given a quarter of a mile sooner. But it is not certain that they would not have heard it. “Would not” must go a long way before it becomes “could not,” in satisfaction of the rule of The Pennsylvania. On this horseshoe bend, had the signal been given when the Merry Queen was a quarter of a mile farther downstream, the Minerva would have been about a quarter of a mile farther upstream, but not much farther from the Merry Queen as the crow flies, and at a somewhat better point from which to receive the signal on the east wind.

Respondents counter with the so-called major-minor fault rule, to the effect that where the fault of one vessel is overwhelming and that of the other trivial, the latter’s fault will be excused. 11 That rule is inapplicable here, however, because, as we have said above, the Merry Queen’s fault was far from trivial. It follows that the Minerya’s damages must be divided.

Therefore, the judgment of the district court will be vacated and the cause remanded with directions to determine the amount of the Minerva’s damages and to enter judgment in favor of libellants for one-half that amount.

. Art. 18, Rule V, Inland Rules of Navigation, 30 Stat. 100, 1897, 33 U.S.C.A. § 203, 1928.

. Id.

. The Mamei, 3 Cir., 1945, 152 F.2d 924, certiorari denied Eastern Transport Co. v. Walling, 1946, 328 U.S. 836, 66 S.Ct. 981, 90 L.Ed. 1611.

. They are really respondents and claimants since there was a libel in personam and in rem. We shall use “respondents” merely for convenience.

. 2 Wigmore, Evidence, §§ 285 and 286 (3d ed. 1940). Libellants quoted from the Coyle Lines case, but omitted the following significant sentence: “The Government did not' produce as witnesses the lookout on the foreeastlehead, nor any of the other seamen located there who might have seen the barge, nor did the Government make any showing in the testimony that .these missing witnesses were not in the employ of the Government or that they were searched for or that they could not be located.” 195 F.2d 737, 741.

. “Rule V. Whenever a steam vessel is nearing a short bend or curve, in the channel, where, from the height of the banks or other cause, a steam vessel approaching from the opposite direction can not be seen for a distance of half a mile, such steam vessel, when she shall have arrived within half a mile of such curve, or bend, shall give a signal by one long blast of the steam whistle, which signal shall be answered by a similar blast, given by any approaching steam vessel that may be within hearing. Should such signal be so answered by a steam vessel upon the farther side of such bend, then the usual signals for meeting and passing shall immediately be given and answered; but, if the first alarm signal of such vessel be not answered, she is to consider the channel clear and govern herself accordingly.” Art. 18, Rule V, 30 Stat. 100, 1897, 33 U.S.C.A. § 203, 1928.

. Read v. United States, 3 Cir., 1953, 20.1 F.2d 758; 4 Benedict, Admiralty, § 573 (6th ed., 1940).

. In The Martello, 1894, 153 U.S. 64, 75, 14 S.Ct. 723, 727, 38 L.Ed. 637, the Supreme Court, applying the doctrine of The Pennsylvania, used the following very strict language: “Can it be said in this case that the absence of a mechanical fog-horn could not by any possibility haye contributed to the collision?” (Emphasis supplied.)

. In Wood Towing Corp. v. Paco Tankers, 4 Cir., 1945, 152 F.2d 258, the tanker Chilbar blew her bend signal more than half a mile from the bend. The grounding of the tanker was caused primarily by the tug’s failure to keep her tow in line and allowing it to drift to port and to obstruct the tanker’s passage. Nev-theless, the court divided the damages, saying, 152 F.2d at page 261: “While the causative effect of the tanker’s dereliction is not obvious, the rule is that when a vessel is violating a statutory rule intended to prevent collisions and a disaster occurs, she must show not only ‘that her fault might not have been one of the causes, or that it probably was not, but that it could not have been.’ ”

. This rule is of little help in deciding cases. It has been observed that it is artificial and misleading unless very carefully applied. General Seafoods Corp. v. J. S. Packard Dredging Co., 1 Cir., 1941, 120 F.2d 117. See also The Admiral Schley, 1 Cir., 1904, 131 F. 433, affirmed on rehearing 1 Cir., 1905, 142 F. 64, cer-tiorari denied, 1906, 201 U.S. 648, 26 S. Ct. 762, 50 L.Ed. 905, where it is pointed out that in many cases the rule would operate in reverse directions depending upon which vessel’s faults were first considered in determining the cause of the collision.

Griffin states that “The principle of major and minor faults cannot be carried too far. Even if one vessel’s wrongdoing has been the chief cause of the collision, the other will not be excused if she has been guilty of a substantial contributory fault, especially if that fault was statutory." (Emphasis supplied.) Griffin, Collision, § 226 (1949).