STURGTS, J.
We all agree that this case should be reversed and remanded, but for somewhat different reasons. The case was first assigned to Robertson, P. J., and the following statement of facts prepared by him, with some slight modifications, will be adopted.
Plaintiffs’ sawmill, near defendants’ railroad, was destroyed by fire. They sued defendants and obtained a judgment from which defendants have appealed. The trial was to a jury and the defendants submit here that the verdict should have been directed for them and that an instruction given in behalf of plaintiffs which authorized a verdict in their favor, if the jury believed “from all the evidence in the case that it was more likely that plaintiffs’ mill was set on fire by an engine being operated on defendants’ track than from any other cause” was erroneous. The plaintiffs’ mill was situated on the west side of defendants’ railroad, parallel therewith and about ninety feet from the center of the track. At that point the railroad runs nearly southwest and northeast and curves to the east. On the night of April 24, 1915, at about eleven o’clock, a freight train passed on defendants ’ railroad going north and about half an hour later, the south end of the mill was discovered in flames. The mill was a frame building, two stories high, and with an old plank roof which was partly open (forty feet downstairs and ten or twelve feet upstairs on the south end) on the side toward the railroad. On the night of the fire and for some days prior thereto there was a high wind blowing from the direction of the railroad toward the mill and the weather was dry. There was testimony that the *135track was about level at the place where the defendants’ railroad passed the mill and the train “slowed up and then speeded up.” One witness who resided a short distance north of the mill, less than one thousand feet, testified that she observed the train when it passed her but did not notice it throwing any sparks. Some other witnesses testified that they had seen engines at that point throw sparks forty feet high at night, as high as the trees, and one witness testified that three days before the mill burned she put out two or three small fires during the daytime between the mill and the railroad, started within about five feet of the mill, and that these were started by a passing train. Other witnesses testified that they had seen live sparks going beyond the right of way. The mill was not in use at the time it burned and had not been operated for over a year. No other origin of the fire is suggested. Defendants offered no testimony. Some of the witnesses stated that the mill was forty feet from the right of way and all who spoke of the distance from the track placed it at about ninety feet. The witness who put the fire out a few days before the mill burned said the mill was thirty feet from the right of way fence and that the fires she extinguished were twenty-five feet from the fence. On the day on' which this witness discovered the fire near the mill there must have been a high wind and the weather was evidently dry, as the witness testified that these conditions had existed some days prior to this fire. Assuming that it is found that the engine was emitting sparks and that the conditions w'ere as testified to, then we would, in order to uphold this verdict, have to reason about as follows: At one time a few days before the’ mill burned a live spark was carried to within five feet of the mill and there ignited the dry vegetation. Therefore at this time, practically the same weather conditions prevailing, it would be carried five feet further and ignite the mill. The fire was discovered about half an hour after the train passed and the whole mill was then on fire, showing that it had been burning approximately 'that long. For *136this reason it cannot be told just where the fire first started.
Under these facts Robertson, P. J., is of the opinion that the testimony is not sufficient to fix liability upon defendants for this fire under the rulings in Gibbs v. Railroad, 104 Mo. App. 276, 282, 78 S. W. 835; Manning v. Railroad, 137 Mo. App. 630, 635, 119 S. W. 464; Fritz v. Railroad, 243 Mo. 62, 148 S. W. 74.
Farrington, J., and the writer are of opinion that plaintiff made a case for the jury under proper instructions but that the instruction -complained of is erroneous, for these reasons: The evidence in the case is purely circumstantial but facts may be proven by circumstantial evidence as well as by direct evidence. We all agree that the instruction is not erroneous, however, because not requiring the facts proven to be such as to exclude every reasonable possibility of the fire having some other origin. The statement in the opinion (Sheldon v. Railroad, 29 Barb. 226) quoted from in Peck v. Railroad, 31 Mo. App. 123, 128, to the effect that the proof in this kind of cases must be such as to leave no reasonable doubt as to the origin of the fire, has not received the sanction of our Supreme Court in subsequent cases which have come to our attention. [Kelley v. Railroad, 151 Mo. App. 307, 310, 311, and cases there cited.] In the case of Big River Lead Co. v. Railroad, 123 Mo. App. 394, 400, 101 S. W. 636, the statement of the law quoted from in the Peck case was condemned and the decisions to the contrary reviewed. We think our Supreme Court has, in effect, repudiated that doctrine, and the New York Court of Appeals, in the same case 14 N. Y. 223, repudiated the utterances of said quotation. In Campbell v. Railway, 121 Mo. 340, 349, it is said: “The evidence was all circumstantial. It is important, then, to show that there was a possibility that sparks may have been thrown a distance sufficient to reach the building in which the fire originated and that they contained heat enough to set it on fire. The fact that live sparks were thrown from engines and did ignite grass and other combustible materials would tend to prove the probability that the fire was *137communicated from an engine.” And again, “We think the evidence tends to prove the possibility and consequent probability that the fire - was communicated to plaintiff’s property from one of defendant’s engines and that the evidence was admissible and its probative force was for the determination of the jury.” See also the Manning case, supra. The mere use of the word “likely” instead of “probable” in the instruction is of little consequence, as we think they carry much the same meaning, though the word “probable” should be used to avoid criticism and as being more accurate and less likely to be misunderstood in this connection.
We consider that it is sufficient to sustain a verdict for plaintiffs in a case of this character based on circumstantial evidence, that the evidence shows that the fire could have been communicated from the engine and that such is the most probable source of its origin. It will be noticed, however, that the instruction first mentioned, given for plaintiffs, directs a finding for them in case the jury found that the origin of the fire was more likely from the engine than from any other cause disclosed by the evidence; in other words, that plaintiff is entitled to a verdict whenever the jury finds that the evidence points to the engine as the most probable source of the fire without requiring a finding that the engine did in fact start such fire. We think the jury should be required, in all cases, and especially so when the evidence is purely circumstantial, to find that defendants’ engine did in fact set fire to the destroyed property and that there is a substantial difference between requiring the jury to find that the engine was the source of the fire and in finding that it was the most probable source. Here the jury were told to base their verdict on a finding “that it was more likely that plaintiffs’ mill was set on fire by an engine operated on defendants’ track than from any other cause,” without requiring a finding that it did do so: A jury might readily find that the engine was the most probable cause but be unwilling to find that defendants’ engine did, in fact, set the mill on fire; and no less a finding should be allowed in determining defendants’ liability.
*138In Fritz v. Railroad, 243 Mo. 62, 78, 148 S. W. 74, speaking of a case similar to this the court said: "The evidence relied on to show the origin of the fire is strictly circumstantial — that is, the main fact, viz., the cause of the fire, stands to he proved by proof of surrounding conditions and circumstances ‘whose existence is a premise from which the existence of the principal fact may be concluded by the necessary laws of reasoning.’ [Bl. L. Diet., under Circumstantial Evidence.] ‘The soundest test of the validity of that sort of evidence is, that no other theory but the hypothesis upon which the conclusion is based, can be formed.’ [Per Hughes, J., in Musselwhite v. Receivers, 4 Hughes (U. S. Cir. Ct. Rep.), l. c. 169.] In circumstantial• evidence the principal and ultimate fact is got at by way of argument and by method of demonstration in the nature of reductio ad absurdwm. [Will’s Cir. Ev., p. 17.] Therefore, in cases turning on circumstantial (or what' is called by the civilians oblique and by the Scotch jurists argumentative) evidence the proof should have a tendency to exclude any other reasonable conclusion than the principal fact.' Says the author just quoted (p. 17): ‘The force and effect of circumstantial evidence depend upon its incompatibility with, and incapability of, explanation or solution upon any other supposition than that of the truth of the fact which it is adduced to prove.’” 1 G-reenleaf on Evidence, sec..13, speaking of proof by circumstantial evidence, says: ‘‘And from these facts, if unexplained by the prisoner, the jury may or may not deduce or infer or presume his guilt according as they are satisfied or not of the natural connection between similar facts and the guilt of the person thus connected with them.” In! speaking of the proof required and what the jury must find in a case similar to this, the court, in Lead Co. v. Railroad, 123 Mo. App. 394, 402, said: ‘‘That evidence showing no more than a probability that the source of the 'fire was a railroad engine, is sufficient to submit to the jury, was decided, in effect, in Campbell v. Railroad, supra; Sheldon v. Railroad, 14 N. Y. 223; Field v. Railroad, *13932 N. Y. 339; Railroad v. Richardson, 91 U. S. 470; Smith v. Railroad, 63 N. H. 25, and numerous other eases. But of course the probative force of the evidence must be strong enough to induce a belief in the minds of the jury that, the fire in fact originated from a locomotive, not merely that it might have done so. The evidence ought to warrant more than a conjecture as to the source of the fire — ought to suffice to induce a conclusion regarding the matter in the minds of reasonable men.” And in the same case the court commented favorably on an instruction to the effect that a verdict “could not be found for plaintiff on a mere conjecture as to the cause of the fire, but the jury must find that it loas communicated by an engine on defendants ’ road while passing Irondale, and the burden of proving this fact by the greater weight of. the evidence was on the plaintiff; that the evidence to prove the origin of the fire was purely circumstantial, and if the jury found the circumstances were more consistent with the theory that it caught from some other source than defendant’s engines than with the theory that it was communicated by an engine, the verdict should be for defendant.” The last instruction mentioned by the court was given for defendants and it would be entirely proper to give on defendants’ behalf a cautionary instruction, as was done in that case, that there should not be a finding for plaintiff unless the evidence showed that the engine was the most probable cause of the fire or to the effect that before plaintiff could recover he must prove that the fire was more probably caused by the engine than by any other cause These, however, are cautionary instructions in defendants’ favor. For the plaintiff, on whom is the burden of proof, the instructions should require a positive finding that the defendants’ engine set on fire the plaintiffs’ mill. A mere finding that such' cause is more probable than any other is not sufficient to fix liability. The objection to and danger in proving a fact by circumstantial evidence is that it is based on deductions from other facts, and that all the facts not being known a- wrong conclusion is possible, if not probable. It *140would render proof by such evidence far more dangerous to not require the jury to make a'finding that the ultimate fact is true, but only that such is more probable than anything else. For the reasons given, the cause is reversed and remanded.
Farrington, J., concurs. Robertson, P. J., is of the opinion that the instruction is not erroneous, but concurs in the result, as stated in the opinion.