SUMMARY: Appeal on fact – insufficient basis for interfering with trial court’s findings of credibility – insufficiency of evidence of negligence.
 On the night of 26 October 1997 an Isuzu bakkie, driven by one Flanagan, was being used to tow a Mazda 323 motor-car on the national road between Queenstown and King William’s Town. The appellant was at the wheel of the Mazda. While they were negotiating a pass between Cathcart and Stutterheim the Mazda came adrift from the Isuzu. It collided into the rock face on the right hand side of the road. The appellant sustained bodily injuries in the course of the collision and in consequence claimed compensation from the Road Accident Fund (the respondent) in terms of the provisions of Act No 56 of 1996. The trial came before Chetty J who ordered absolution from the instance on the strength of strong credibility findings in favour of the Fund’s witness and against the appellant and his witness. The appellant appealed against that decision. The issue before us is whether or not there are grounds for interfering with the learned trial judge’s credibility findings on appeal.
 Three witnesses testified before Chetty J: the appellant, who was the ‘driver’ of the Mazda and the plaintiff below; one Cumming, the owner of the Mazda, (both on behalf of the appellant); and the driver of the Isuzu, Flanagan, on behalf of the Fund. The common ground in their versions was that Cumming, whose Mazda had broken down and was stranded in Queenstown, had asked the appellant and Flanagan to help him to tow it back to King William’s Town. They agreed. Flanagan was to use his Isuzu bakkie as the tow vehicle, and, because Cumming had no driver’s licence, the appellant was to steer the Mazda. The three of them, and a man called Arends, left King William’s Town in the late afternoon of Saturday, 26 October 1997 in the Isuzu. They arrived at Queenstown, made the Mazda ready for towing, attached it to the Isuzu with a rigid metal tow bar, and set off back to King William’s Town. Flanagan drove the Isuzu, with Arends as a passenger. The appellant steered the Mazda and Cumming kept him company.
 The statement that they had made the Mazda ready for towing is a euphemism. They saw to the tyres, and Flanagan ‘bled’ the brakes. But the Mazda had no battery. This meant that it was without headlights, it had no indicators or emergency flashlights, and its driver could not attract attention by flashing his bright lights or hooting. There was nothing they could do about it. The party, apparently aware that a vehicle in this condition should not be towed on a public road especially after dark, debated delaying their return to King William’s Town until the following morning. They decided against it.
 They left Queenstown at about 21h00 or 22h00. They stopped briefly at Cathcart, checked the tow bar, and then proceeded towards Stutterheim. The road was wide with a tarred surface. It was properly marked with centre white lines and yellow shoulder lines. On that section there were a number of passes with inclines and declines. There was some other traffic that night, but apparently not much.
 At a point about 25 kilometres from Stutterheim the tow combination went out of control. The Mazda began to sway from side to side and, according to the appellant and Cumming, one of its wheels went over the tow bar. This caused the tow bar to become disconnected. The Mazda’s brakes stopped working, and it ran out of control over a ditch and into the rock face on the right side of the road. The Isuzu came to a stop just forward of it on the incorrect side of the road. Flanagan was unable to pull it off the road to a place of safety because one of his back wheels had gone over the tow bar and was suspended in the air. The tow bar was still attached to the tow hitch at the front end and was still fastened into the eyelet in the body work of the Mazda designed for connecting a tow bar or chain. The eyelet and a portion of the Mazda’s bodywork adjacent to it had pulled away from the rest of the Mazda. Flanagan disconnected the tow bar, got the Isuzu back on to its wheels, and moved it to a place of safety off the surface of the road. He then attended to the appellant. He drove him to hospital.
 The appellant did not argue that driving the Mazda in its unroadworthy condition was, in itself, a negligent act on the part of Flanagan, perhaps because it was their common decision to tow it in its present state and because there was no causal connection between the collision and the lack of lights, indicators or hooter. But he argued that it gave rise to an even greater duty on Flanagan’s part of drive slowly and with caution. He relied on two grounds of negligence: first, that Flanagan was negligent because ‘he drove the insured vehicle [the Isuzu] at a speed which was excessive in the circumstances, and in particular was excessive for the safe towing of disabled vehicles, in particular the Mazda motor vehicle being steered by the plaintiff’; and, second, ‘that he failed to have and to keep the insured vehicle under proper control and, in particular, allowed the insured vehicle to commence movements from side to side without regard to the effect thereof on the Mazda motor vehicle connected to the insured vehicle by means of an iron tow bar’.
 Chetty J rejected the argument that the appellant had discharged the onus of proving either or both of these grounds of negligence. He did so after an evaluation of the evidence in which found that Flanagan was an impressive and credible witness whose evidence he accepted, and that on the other hand the appellant and Cumming were unsatisfactory witnesses who told deliberate untruths and whose evidence he was obliged to reject as false. Their version was that Flanagan drove at the dangerously high speed of 100 kph and that he caused the Mazda to swing from side to side in a manner, which, at that speed, caused it to go out of control. Chetty J accepted Flanagan’s account of how he drove, and found that on that version no fault could be ascribed to him in relation to the collision. Flanagan’s evidence was that he maintained a moderate speed of between 60 to 80 kph, reducing to 40 kph downhill, throughout the trip, which was a safe speed if regard is had to the unroadworthy condition of the Mazda, and that he did not cause the Mazda to sway from side as the other two described. On occasion he was obliged to move to his left on to the shoulder and then back right into his lane to enable following traffic to pass safely, but in doing so he did not cause the Mazda to sway violently or to go out of control. If his evidence is correct, the Mazda must have become separated from its tow vehicle as the result of some cause unconnected with the way in which he drove.
 It was necessary for Mr Cole, for the appellant, to challenge the trial judge’s findings of fact and credibility. This is never an easy task. Mr Cole nevertheless sought to persuade us that a reading of the record leads to the conclusion that the trial judge’s assessment of the oral evidence was contrary to the probabilities and the evidence as a whole, and therefore wrong. He submitted that this conclusion follows from (a) the failure of the respondent to challenge in cross-examination that the Mazda began to sway violently from side to side some time before the tow bar must have become disconnected; if the appellant’s unchallenged version on that point had been accepted, he argued, the appellant’s evidence of speed, which was in any event suspect, was more probable than Flanagan’s; (b) a concession by Flanagan that if he had been going as slowly as 30 kph the collision would not have occurred; (c) the fact that Flanagan’s version is physically impossible and does not explain how the collision occurred; and (d) the fact that the judgment criticizes the evidence of the appellant and Cumming in minor and unimportant respects only.
 This argument cannot succeed. This is a case where the two versions are so conflicting that they are irreconcilable and mutually destructive. For the plaintiff to succeed in these circumstances he must, in the well known words of Eksteen J (as he then was) in National Employer’s General insurance Co Ltd v Jagers,1 satisfy the court ‘on a preponderance of probability that his version is true and accurate and therefore acceptable and that the other version is therefore false or mistaken and falls to be rejected. In deciding whether that evidence is true or not the court will weigh up the plaintiff’s allegations against the general probabilities’. The first three points raised by Mr Cole do little to assist the appellant in discharging this onus. They attack Flanagan’s evidence. But there is no onus on Flanagan or the Fund. Flanagan does not have to give an explanation of the collision, and neither does the Fund, at any rate until such time as the appellant has placed a credible version before the court which is reasonably capable of acceptance. Furthermore, Mr Van der Linde’s careful analysis of the evidence in chief and cross-examination satisfies me that the appellant’s version of the Mazda swaying from side to side was properly challenged in cross-examination, and that Flanagan’s evidence of his speed was consistent. Nothing can be made of the concession that a collision would not have occurred at a speed of 30 kph. This is not a concession that any speed greater than that was a dangerous speed. I am also not impressed with the submission that Flanagan’s version must be rejected because it is physically impossible. It is based on Mr Cole’s proposition that the Mazda must have been going faster than the Isuzu if it was where Flanagan says he saw it at the crucial time, and that this was not possible because the Mazda was attached to the Isuzu with a rigid metal bar. This kind of reasoning does not commend itself to me in the circumstances of this case. Its foundation is a sophisticated and over-intricate analysis of the witness’s attempt years later to describe an event which suddenly and unexpectedly occurred behind him at night. It is artificial to draw a final conclusion of this kind on the strength of it, especially where the facts cannot be ascertained with the precision necessary for this kind of exercise.
 Mr Cole criticized the learned trial judge for rejecting the evidence of the appellant and Cumming because of contradictions, inconsistencies and improbabilities which were trivial and did not relate to the real issues. A reading of the record does not support his argument. It reveals that these two witnesses contradicted themselves and each other in respects which are not trivial, and that their evidence was shown to be improbable and, in some instances, patently false. They contradicted Flanagan on almost every point of significance: how many tyres were pumped up; whether brake fluid supplied by Flanagan was used to repair the brakes, or cooking oil from the kitchen; whether the tow bar was properly attached to the Mazda with a U-bolt supplied by Flanagan, or whether they used pieces of wire and old seat belt material found lying around in the yard; whether it was Flanagan, or the appellant and Cumming, who did not want to wait until morning to do the towing; whether they stopped before getting to Cathcart, when the appellant and Cumming asked Flanagan to slow down; whether they again asked him to slow down when they stopped at Cathcart. In measuring the effect of these differences Chetty J considered that the probabilities favoured Flanagan’s version. While some of them do not bear directly on the issue of the collision, Chetty J in my respectful opinion correctly held that in this case they are germane to the whole question of credibility. He considered that they ‘impact deleteriously on the credibility of the plaintiff and Cumming’ and that their false evidence was tendered ‘not merely as part of the plaintiff’s narrative but deliberately in order to establish Flanagan’s total disregard for the safety of the plaintiff and Cumming’. It is against this background that Chetty J made credibility findings in favour of Flanagan and against the appellant and Cumming. We cannot lose sight of the rule that an appeal court's power to interfere with the trial court’s assessment of oral evidence and its findings of fact and credibility is limited, and that in the absence of a material misdirection (which was not established) the trial court's conclusion is presumed to be correct. In order to succeed on appeal the appellant must convince the appeal court on good grounds that the trial court was wrong.Bearing in mind the advantage which a trial court has of seeing and hearing the witnesses, it is only in exceptional cases that an appeal court will be entitled to interfere with the trial court’s evaluation of oral testimony. These principles are well known. See R v Dhlumayo2;Taljaard v Sentrale Raad vir Koöperatiewe Assuransie Bpk3; S v Robinson and Other4; S v Francis5. In this case I am of the view that the presumption that Chetty J’s conclusion is correct remains undisturbed.
 Mr Cole argued in the alternative that on Flanagan’s own evidence he was negligent in driving at the speed he said he drove at, bearing in mind the unroadworthy condition of the Mazda. The difficulty with this argument is that there is insufficient evidence to show that Flanagan’s speed was excessive in the circumstances, even when regard is had to the condition of the Mazda. Flanagan said that he drove at between 60 kph and 80 kph, and slowed down to about 40 kph while going downhill to ensure that he maintained control. He considered that this speed was a proper and safe speed in the circumstances, bearing in mind the lack of lights, indicators and a hooter. There was no question of his losing control until the last moment, when, for some reason which he cannot explain, the tow bar became disconnected. One cannot argue that his speed must have been excessive solely because the collision occurred at that speed. That begs the question. What objective criteria are to be used to determine a safe speed in the circumstances? What facts show that his speed was a negligent speed? There is insufficient information, on Flanagan’s version, to come to a conclusion that he drove too quickly in the circumstances.
 In the result the appeal is dismissed with costs.