DENISON, Circuit Judge.
A freight train became stalled on the east-bound main track of the Pennsylvania Railroad. The automatic block signal system was in use and was in effective operation. One of these signals, which we may call No. 1, was about 300 feet west of the rear of the stalled train. The next one, which we identify as No. 2, was 4,500 feet farther west. According to this system, when an east-bound train passed No. 2, and while it remained in the block between No. 2 and No. 1, No. 2 showed a red light. As the train left the block and entered on the next block east, No. 1 showed a red light, and No. 2 changed from red to yellow, and would continue yellow so long as any part of the train was in the block adjacent on the east *362to No. 1. When that block also became unoccupied, the light at No. 2 would change from yellow to white. It was the duty of the engineer on any following train, when he passed the yellow signal to reduce speed and bring his train under control, and this meant, the witnesses say, to reduce to a speed of from 5 to 10 miles an hour. If, then, this engineer at a subsequent block got the red signal, he must stop until the light again became yellow or white. There was .also a rule and practice which, when a train stopped upon the main track, required the flagman to go back far enough to protect his train and be prepared to give his stop signal to any following engineer. The testimony tended to show that this rule and practice were rightly subject to the interpretation that “far enough to protect his train” meant about 40 car lengths, or 1,600 feet.
The stalled train sent for help, which soon arrived in the shape of a light train, consisting of engine and caboose, coming from the west. It coupled onto the rear of the other train and prepared to help. Pursuant to the rule and practice, the light train sent a flagman back. In his progress west, he had reached a point about 800 feet from his own caboose when he saw another train approaching. It was close to him before it was visible, because of the extreme fogginess prevailing, but with his red and white lights he gave the prescribed stop signal. This train was a special fast freight, running about 35 miles an hour, and it is imdisputed that it passed the yellow block at No. 2, and passed the brakeman’s stop signal, and passed the red block at No. 1, all at full speed and without any attention to the signals. It crashed into the standing train, and the engineer, who thus had not seen or had disregarded all signals, was killed. His representatives brought this action in the court below, under the federal Employers’ Liability Act (Comp. St. §§ 8657-8665). His negligence is not denied, but it is the theory of the action that his personal negligence and the negligence of the railroad in not giving the flagman’s stop signal farther west cooperated, whereby each became a proximate contributing cause, - in which event, as specified in the act, and as pointed out by the Supreme Court in Norfolk Co. v. Earnest, 229 U. S. 114, 33 Sup. Ct. 654, 57 L. Ed. 1096, Ann. Cas. 1914C, 172, and as applied by this court in Pennsylvania Co. v. Cole, 214 Fed. 948, 131 C. C. A. 244, and by the Supreme Court in Union Pacific v. Hadley, 246 U. S. 330, 38 Sup. Ct. 318, 62 L. Ed. 751, his representatives would be entitled to recover a proportionate part of the damages.
The trial court directed a verdict for the defendant. Clearly, the case should have been given to the jury, if there was substantial evidence indicating that there was negligence by the railroad in the matter of the flagman’s signal and that such negligence-was a proximate contributing cause of the disaster. The critical question is, therefore, as ' to the existence of such evidence on each of these points.
[1] We may observe, in a preliminary way, that tire action of the trial judge in directing a verdict and1 in refusing a new trial is not wholly without importance. We have often pointed out (e. g., Louisville Co. v. Lankford, 209 Fed. 321, 325, 126 C. C. A. 247) that the manner of witnesses, their emphasis, etc., may give the case a color *363not perfectly reflected in the printed record, and that, when the trial judge sent the case to the jury and refused a new trial after a verdict for the plaintiff, the reviewing court must be especially cautious in concluding that there was nothing sufficient to justify that course. The converse must be true. Where the conclusions of the trial judge have been what they are upon this record, the reviewing court may safely assume that the plaintiff’s case was not helped out by an impressive atmosphere not shown in the transcript. Patton v. Railway Co., 179 U. S. 658, 660, 21 Sup. Ct. 275, 45 L. Ed. 361.
[2] We assume upon this record that due diligence required the flagman to go back 1,600 feet, if he could, and there establish his signal post, and that his failure to do so, if he had been able, might rightly have been considered, even under the other circumstances of this case, an effective contributing cause; but those assumptions do not reach the present facts. There is no claim that the flagman unduly delayed in getting his signals and starting back down the track, or that he failed to give his signal when the engine, approached. In fact, there is no suggestion that he could have done anything which he did not do, excepting that he should have walked faster. The only testimony on that subject is his, speaking as a witness for plaintiff. He says: “I walked very slowly; it was very foggy.” To walk slowly on a, dark and very foggy night would hardly be criticized, and so the essential inference of negligence rests on his use of the word “very.”' In deciding, not how far, but how fast, to go, he had a right to assume that the yellow block was showing as it was, and that the engineer of any following train would be especially careful in the thick fog and would come on under control. He knew, also, that the standing train was protected by a red block 300 feet in its rear, and that his own signal, given at the point which he could and did reach, even when walking slowly, in connection with the red block, was fairly sufficient to protect his train against any following train that should be reasonably expected. Even from the point he did reach, a signal which was observed would have enabled the following train to slow down, so as to minimize the danger, if not entirely to avoid it.
In addition, no attention was paid to any one of the three signals which the engineer did receive. It is the merest surmise that he would have paid any attention to the flagman’s signal if it had been given a little further west. There is no substantial reason to think that this would have made any difference. A signal which, at 800 feet, was neither observed nor obeyed, would in all probability have been equally inefficacious a few car lengths farther back. We find, then, that on the subject of negligence there is at best only a dubióus inference, and, on the subject of causal relation, not even that. A mere conjecture, standing upon a basis of uncertain inference, does not make substantial evidence. Such a case lacks both the quantitative and the qualitative essential miriimum (Patton v. Ry., 179 U. S. 658, 665, 21 Sup. Ct. 275, 45 L. Ed. 361; Virginia Ry. v. Hawk [C. C. A. 6] 160 Fed. 348, 352, 87 C. C. A. 300; Richards v. Mulford [C. C. A. 6] 236 Fed. 677, and cases cited on page 681, 150 C. C. A. 9) and for the same conclusion on somewhat analogous circumstances, see Great Northern v. *364Wiles, 240 U. S. 444, 448, 36 Sup. Ct. 406, 60 L. Ed. 732; Virginian Ry. v. Linkous (C. C. A. 4) 230 Fed. 88, 144 C. C. A. 386. Though the general rule is that the jury should say what cause is proximate (Milwaukee v. Kellogg, 94 U. S. 469, 474, 24 E. Ed. 256), it is no more invariable than is the rule as to what is negligence, ánd the two cases last cited are illustrative exceptions. We have no difficulty in distinguishing the facts of this case from those of the Hadley and Cole Cases, supra; indeed, as to the particular matter we are discussing, there is no resemblance. In those cases, the negligence of the flagman was conceded,' and the issue which was considered proper for the jury was as to’the negligence of the following engineer. In the present case, the engineer’s negligence is admitted, and it is the conduct of the flagman which is criticized.
It is also suggested that the flagman failed in his duty because he did not use a fusee. Ordinarily, a fusee may be left by the flagman, burning on the track, when he leaves his station and is called back to his train; possibly there is evidence tending to show that it would be prudent, in such a fog as this, for the flagman to light a fusee and let' it burn while he was keeping his protecting station; but there is nothing to show that it was either customary, or was a reasonable precaution, for the flagman, going back to his intended station, to carry a burning fusee as he walked; nothing short of this would have been of the slightest importance on this occasion.
[3] It is also urged that the company was negligent, in that the train dispatcher did not give notice to the fast freight of the obstruction at this point. This is in effect to say that it was negligent to permit one train to follow so close after another that there would not be time for the leading train to stop and send a flagman back the necessary difetance. It is not made to appear that any such system of notice would be practicable, nor is it claimed that there' was any rule or practice requiring the same. The automatic block system was to meet this very situation. Through that system, the company did, on this occasion, give notice to this engineer that the track was obstructed a short distance ahead of him; but he disregarded the warning.
It should be added that the rules required the engineer, where for any reason he could not see the block signals, to stop or proceed with great caution, and that the crew of the preceding light train and others of the fast freight saw and observed these block signals. The fireman of the fast freight was not sworn; we infer that he also was killed.
The judgment is affirmed.