must have begun. If the world ever was made and wound up like a watch, whether five thousand years ago, as theologians used to suppose, or twenty-four hours ago, we can never see anything on the face of it to indicate when that took place. All we can say is that if this world-watch really was going six thousand years ago as it is going now, then at precisely that time the hands were in such and such a position.

It seems much easier to believe that the world started five thousand or five billion years ago than twenty-four hours ago, merely because our whole conception of so distant a past is vaguer. But if it started five thousand years ago, it started with fossils in the rocks and all the other absolutely definite conditions that must have preceded the present and that lead scientists to trace its history beyond the five thousand years of its actual existence. And if it started five billion years ago, every detail must have been just as definite; however hazy our idea of it may be. It might just as well have been created twenty-four hours ago with mines half empty and cities all built and ships in the harbors, and adult human beings busy at their work, with brains so fashioned that they look back towards an imaginary past and believe they remember it. Thus, starting from the present as science does and must, and assuming as it does that things acted in the past and will act in the future according to the same laws that we find now, it is quite impossible for us to find any point at which the world-process must have begun or must come to an end, even if there really was such a beginning and really will be such an end.

One might think that the uncertainty due to the fragmentary nature of our knowledge and this possible accumulation of errors should affect our knowledge of the • Monupast quite as much as our knowledge of the future; ments.' but as a matter of fact it does not. We know a great deal more about the past; and the reason is that many conditions, when they are once produced, remain practically unaltered for a great many years. No geologist can tell whether the

place where I am now writing will ever be covered with water or not; but from the layer of gravel beneath the surface of the ground I can infer with reasonable certainty that once it was. No one can tell whether Vesuvius will ever destroy a city again, but from the buried remains of Pompeii and from written records we can infer that once it did. This is merely because we have reason to believe that such things as beds of gravel, the stones of buried cities and forgotten documents remain comparatively unchanged for years or centuries. Thus when we find such things they take us at one leap beyond all the intervening years, and each of them shows us one fragment of the past as it existed at the time.

Accordingly, if we are right in assuming that such relics of the past have remained unaltered in any given respect, and if we have any means of finding out their age, we can use each one of them as a starting-point in the construction of the past, or at least as something fixed by which to test our inferences about it. In such cases we might almost say that we really do start with the past. This is why we can tell more about some things in the past than in the future; why there is such a thing as serious history where there is no corresponding prophecy.

It is because there are such absolutely or relatively permanent monuments' of the past as these that it is possible for us to distinguish between relatively direct evidence concerning past events and indirect. The most direct evidence concerning any event is, of course, the present and personal perception of the event itself. This is something independ ent of 'monuments' of any sort, but if the event is past, this evidence is unattainable. The most direct evidence then possible is the perception of something which we can assume to be a direct effect of the event, and after that the supposed effects of effects of the event or the supposed effects of something that would have caused the event; until the chain is as long or crooked as you please. The most direct evidence of the destruction of Pompeii by a volcanic eruption would be

the perception of the event itself. The most direct evidence possible for us is the personal perception first of the unbroken lava and then of the houses being dug out from beneath it. Photographs of such lava-beds and houses give less direct evidence, and the tales of persons who claim to have seen such photographs less direct still.

It is hardly necessary to add that the most reliable evidence about past events is afforded by monuments which we know to have undergone the least possible change, and which are connected with the events under investigation by the shortest and surest series of causal relations. The value of written records as monuments of the past will be spoken of in the next chapter.

stantial

One hears a great deal in discussions of criminal trials about circumstantial evidence. Evidence is called circumstantial when the witnesses have not observed, Circumand therefore cannot tell about, the very fact at evidence. issue (e.g., A murdering B), but can and do tell about various other facts so connected with the fact at issue by the law of causation that from them the jury can infer what that fact really was. Thus circumstantial evidence' is only the common name for indirect evidence of the sort that may be given in a court of law. If A and B were heard talking loudly as though they were in a violent dispute and if A was afterwards seen leaving the house covered with blood, while B was found stabbed to the heart with A's knife, there would be good circumstantial evidence that A had killed him, and he might very well be convicted and hanged for a crime that no one had seen him commit.

Circumstantial evidence is conclusive only if a supposed state of affairs is the only one that will fit in with the ascertained facts according to the general laws of nature. In a simple case of astronomy or chemistry a conclusion of this sort can be drawn with practical certainty; but with human affairs the case is somewhat different; for human life and the conditions that determine it are so exceedingly complex that it

is very rarely possible to say with anything like certainty: This supposed state of affairs and this alone is consistent with all the established facts. In the case given, for example, it is quite possible that B was seized with a sudden mania during which he first picked a quarrel with A and then seized A's knife and stabbed himself, while the blood on A was due to his efforts to prevent B from doing the mad deed. Or it might be that A and B were not quarrelling with each other but with some third person, who had afterwards picked up the knife, committed the murder and then made his escape. Or the real truth of the matter might be something entirely different that nobody happens to think of at all. A person convicted of some crime on circumstantial evidence is thus often convicted on a degree of probability which falls conconsiderably short of practical certainty; which means, of course, that amongst a large number of such cases there are a few in which the person convicted and punished is really innocent. This accounts for the general feeling of uneasiness or dissatisfaction when a trial of vital importance is decided altogether upon circumstantal evidence. But we must remember that the uncertainty of a conclusion based upon such evidence is due to the complexity of the situation dealt with and our ignorance of many of its details, not to a defect in the general principle; for the principle is that involved in all indirect evidence, and it is the only one in virtue of which we can gain any knowledge whatever of either the past or the future.

*

* See p. 334.

CHAPTER XXXV.

TESTIMONY.

tance.

Yet the im

AMONGST the most important monuments of past events are the impressions made by them on the minds of persons who were present at the time. From these impres- Its imporsions, if we can find out what they are, we can often judge the nature of the events that made them. pressions in the minds of others can never be observed directly, but only inferred from what they say and do. Hence there is always a double inference: from what a witness says to his thoughts, and from his thoughts to the events that caused them. The necessity for this double inference often makes the correct estimation of testimony very difficult; and yet in all matters of history—whether they be the events of last week told in a police court or the events of two hundred years ago set forth in a formal treatise-its proper estimation is most important. If we estimate it wrongly we are bound to reach false conclusions; and if we ignore it altogether we can hardly reach any conclusions at all, for apart from the memories of men and what they have written we have very few unaltered monuments of the past life of individuals, and our human environment is so complex and our knowledge of human nature so slight that an investigator who tried to find out about the past deeds of anybody else merely by going back step by step from his own personal experiences would hardly get started before he had to stop.

Even if there were plenty of other evidence, reliable testi