truth of a proposition. Its value is not a constant quantity. It may be strong or weak, according to circumstances. From weak evidence it is impossible, of course, to build up a satisfactory proof. To know whether a given bit of evidence is weak or strong, is, therefore, the business of every one who would hope to argue successfully.

A thoroughly satisfactory classification of the kinds of evidence is not easy to find. The most natural classification, perhaps, as it certainly is the most familiar, is that which divides all evidence into two kinds, testimonial evidence, and circumstantial evidence. By testimonial evidence is meant evidence given by a human being,—that is, evidence which consists of statements of fact or of opinion based upon some person's observation or experience and bearing directly upon the point at issue; by circumstantial evidence is meant any kind of evidence that is not testimonial,-in other words, any kind of evidence that owes its force wholly to something inherent in itself, and not to the fact that it is asserted by any particular person. Circumstantial evidence is unlike testimonial evidence in that it does not bear directly upon the point in dispute, but must be made to apply to it by inference.

It is a popular supposition that testimonial evidence is, from its very nature, much more convincing than circumstantial evidence. As Huxley points out, however, this is an unwarranted assumption:

Suppose that a man tells you that he saw a person strike another and kill him; that is testimonial evidence

of the fact of murder. But it is possible to have circumstantial evidence of the fact of murder; that is to say, you may find a man dying with a wound upon his head having exactly the form and character of the wound which is made by an ax, and, with due care in taking surrounding circumstances into account, you may conclude with the utmost certainty that the man has been murdered; that his death is the consequence of a blow inflicted by another man with that implement. We are very much in the habit of considering circumstantial evidence as of less value than testimonial evidence, and it may be that, where the circumstances are not perfectly clear and intelligible, it is a dangerous and unsafe kind of evidence; but it must not be forgotten that, in many cases, circumstantial is quite as conclusive as testimonial evidence, and that, not unfrequently, it is a great deal weightier than testimonial evidence. For example, take the case to which I referred just now. The circumstantial evidence may be better and more convincing than the testimonial evidence; for it may be impossible, under the conditions that I have defined, to suppose that the man met his death from any cause but the violent blow of an ax wielded by another man. The circumstantial evidence in favor of a murder having been committed, in that case, is as complete and as convincing as evidence can be. It is evidence which is open to no doubt and to no falsification. But the testimony of a witness is open to multitudinous doubts. He may have been mistaken. He may have been actuated by malice. It has constantly happened that even an accurate man has declared that a thing has happened in this, that, or the other way, when a careful analysis of the circumstantial evidence has shown that it did not happen in that way, but in some other way." 1

1See Huxley's Three Lectures on Evolution.

A distinction must be made between ordinary testimony and the testimony of an expert or authority. The latter, if it can be employed, is much more weighty, of course, than the former. Care must be taken, however, to see that it really is authoritative. Hence before using the argument from authorityunder which term is included not only the testimony of so-called experts, but the statements or assertions made in such compilations as dictionaries, encyclopedias, and the like-one should ask, To what extent is the person who gives the testimony, or the work which is cited, as the case may be, accepted as an authority? The force of the argument will depend solely, of course, upon the willingness of those to whom it is addressed to accept anything coming from such a source as authoritative. With regard to ordinary testimony, the tests to apply are such as will determine whether or not the person giving it possesses average intelligence and has a reputation for truthfulness.

Circumstantial evidence is of various kinds. These kinds may conveniently be grouped under three heads: (1) evidence which, from a known cause, points to a probable effect; (2) evidence which, from a known effect, points to a probable cause or necessary condition of that effect; (3) evidence which is founded on the belief that things which are alike in one or more essential particulars are apt to be alike in others. Arguments based on these three kinds of evidence are commonly known, respectively, as the arguments from antecedent probability, from sign, and from resemblance.

The argument from antecedent probability tries to account for the fact or matter in dispute by bringing forward some known fact or circumstances preceding it, and saying that the former might be expected as a result of the latter. It infers what is likely to happen from what has happened, what is likely to be true from what is admittedly true. In general, it is a strong argument; but its value depends largely upon the clearness with which the cause and effect relation is brought out. If that relation is made unmistakable, the force of the argument is very great, since it establishes a strong presumption in favor of the proposition to be proved. For the argument to become conclusive, however, it is necessary to show not only that the known cause is adequate to produce the alleged effect, but that nothing has interfered with its producing that effect. That a thing is likely to happen is not, of course, proof conclusive that it has actually happened. The argument from antecedent probability, therefore, though a strong argument, must usually be supported by arguments of another kind.

The argument from sign is the exact opposite of the argument from antecedent probability. It tries to account for the fact or matter in dispute by bringing forward other facts or circumstances which are indisputable, but which themselves can be satisfactorily accounted for only on the supposition that the fact in dispute is true. From an effect, it infers a cause or necessary condition of that effect. Thus, the existence of ice in a pool of water on a spring morning is an indication of a freezing temperature the night before; the fact that the clothes of a man accused

of murder are found to be bloody, is an indication that he is guilty of the crime charged against him. The force of this argument varies with circumstances. A single argument from sign may be so weak as to be almost valueless as proof; again, it may be absolutely conclusive. For example, the fact that a man died on the twentieth of the month proves conclusively that he was alive on the nineteenth. In general, the force of the argument comes from the cumulative effect of a number of signs all pointing in the same direction.

The argument from resemblance differs from both the argument from antecedent probability and the argument from sign in that it does not infer directly an effect from a cause, or a cause from an effect, but infers rather that like causes will be followed by the same or similar effects, or that like effects must result from the same or similar causes. Its force depends wholly upon the kind and degree of resemblance existing between the things compared. If the resemblance is superficial or fanciful, the argument has no weight whatever. For the argument to have probative value, the resemblance must be real,—that is, it "must hold in all particulars essentially connected with the point under discussion."

What has sometimes been called the argument from analogy" is simply a particular form of the argument from resemblance. Here the resemblance is not so much in the things themselves as in the relations in which they stand to other things. This

1 Baker and Huntington, Principles of Argumentation, p. 107.