29 CALIF. L. REV. 689 (1941)

Dean Wigmore, recognizing this priority of logic, discusses the form of argument involved in the use of circumstantial evidence....

Wigmore does not deny that in every instance proof must be based upon a generalization connecting the evidentiary proposition with the proposition to be proved. Conceding this, he argues that the generalization may as well be tacitly understood as expressed, that "the transmutation [from the inductive to the deductive form] is useless, because the Court's attention is merely transferred from the syllogism as a whole to the validity of the inference contained in the major premise.'' Yet it is precisely in this transfer of attention that the value of the transmutation lies. The author's own examples illustrate the point. In the case of the repaired machinery we are told: " 'People who make such repairs [after an accident] show a consciousness of negligence; A made such repairs; therefore, A was conscious of negligence.' '' Before this deductive proof can be evaluated, ambiguity must be eliminated from the major premise. By "people'' shall we understand "some people'' or "all people''? If the argument is intended to read, "Some people who make such repairs show consciousness of negligence; A made such repairs; therefore, A was conscious of negligence,'' it contains an obvious logical fallacy. If intended to read, "All people who make such repairs show consciousness of negligence; A made such repairs; therefore, A was conscious of negligence,'' it is logically valid. However, few could be found to accept the premise that all persons who repair machinery after an accident show consciousness of guilt; that is, that no single case could be found of one who, confident of his care in the past, nevertheless made repairs to guard against repetition of an unforeseeable casualty or to preserve future fools against the consequence of their future folly. Here the result of transmuting a proposed direct inference into deductive form is discovery that it is invalidat least in the terms suggested.

The other proposed argument is equally interesting: " 'Men's fixed designs are probably carried out; A had a fixed design to kill B; therefore, A probably did kill B.' '' Once one attempts to deal, in a quasi-syllogistic form, not with certainties but with probabilities, additional opportunities for fallacy are presented. Suppose that it is argued: "Most As are X, B is an A, therefore B is probably X''; or "Nine-tenths of all As are X, B is an A, therefore the chances are nine to one that B is X.'' Neither of these arguments is logically valid except upon the assumption that As may be treated as a uniform class with respect to the probability of their being X. This can be because there really is no way of subdividing the class, finding more Xs in one sub-class than in another, or because no subdivision can be made in terms of available data. Suppose that nine-tenths of all people in the world have dark eyes. If absolutely all one knew about B was that he was a person, it would be an apparent nine-to-one chance that B had dark eyes. But if one knew B to be a Swede, the percentage of dark eyes in the total population of the world would no longer be important. One would want to know about the proportion of dark-eyed Swedes, which might differ from the ratio among humans generally. Similarly in Wigmore's example. We know that we are interested in the probability of execution of a fixed design of a particular kind: to commit murder. There may be variation in the probability of execution of fixed designs on various subjects. As an initial criticism, therefore, the primary generalization should be "Men's fixed designs to kill are probably carried out.'' In this form we have a valid, quasi-syllogistic argument based upon the limited data available. Still, is the premise sound?

"Men's fixed designs to kill are probably carried out,'' as a major premise in this argument, must mean that they are carried out more often than not. While the word "probable'' can be used in other senses, its meaning here is clear. Hence one would conclude from the single datum that A had a fixed design to kill B, no other evidence being offered, that more likely than not A actually did kill B. But when this argument was presented to a group of law students and teachers, only one was willing to accept the indicated conclusion. Several would accept it if supported by adequate evidence that B had been intentionally killed by some one. Others refused to accept it without still further evidence connecting A with B's death, or at the very least evidence that B had no other enemies. Moreover, there was less hesitancy in accepting the argument in its "inductive'' form. Once the generalization was made explicit, and particularly after discussion of the meaning of "probably'' as there used, doubts as to the propriety of the inference arose or sharpened. The demonstration, however "valid,'' is no better than its major premise, and the more one considers this premise the less reliable it looks. Certainly a permitted inference should rest upon some more easily acceptable law.

Of course, it does not follow that a proposed inference is improper because it can be shown not to follow on the basis of one possible generalization, or because anotherwhich by the rules of logic would validate the inferenceis unacceptable. There may be a third law, as yet unexpressed, which would justify the inference and at the same time be commonly accepted as true. And it may be very important to find the valid and accurate link, since the form of the link will control the form of the conclusion.

Persons who are unwilling to agree that men's fixed designs (at least in case of murder) are "probably'' carried out - or, even conceding the fact of murder, that proof of A's fixed design to kill B establishes A, more likely than not, as B's killer - still agree that somehow this bit of evidence does have some tendency to indicate A's guilt. What form of general statement can reconcile these views? Perhaps something like this: "Men having such a fixed design are more likely to kill than are men not having such a fixed design.'' Those who contend that even fixed designs to kill are more often abandoned or thwarted than carried out can and doubtless will still concede that enough such designs are carried to execution so that the percentage of murderers is higher among persons entertaining such a fixed design than among the general public. Obviously this proposed generalization does not lead us from A's fixed design to kill B to the conclusion that A probably did kill B. There is nothing disturbing in this. This conclusion simply does not follow from the evidence of design. The error was in the original "direct induction.'' In fact, no useful conclusion about A's guilt can be drawn from design or intent alone. On the basis of an acceptable generalization we are able only to place A in a class of persons in which the incidence of murder is greater than among the general public. We cannot now say that A is probably guilty, but we can say that the apparent probability of his guilt is now greater than before the evidence of design was received. This is logical relevancy - the only logical relevancy we can expect in dealing with practical affairs where strict demonstration is never possible. The advantage of the transmutation into deductive (though not strictly syllogistic) form is that we know to what degree of proof we have attained, and do not overstate our results.

Which kind of reasoning is more powerful - deductive or inductive? Do trials lend themselves to one kind of proof more than the other? Deductive reasoning, if valid, is extremely powerful - in fact, it is airtight. Inductive reasoning always requires an inductive leap. Yet as the excerpts above point out, all deductive reasoning requires induction in formulating the major premise, so all legal reasoning is inductively based at some point in time.

Note in The Unopened Drum of Paint problem how changing the substantive rule also changes the mode of reasoning from inductive to deductive. If the issue is whether the second drum is defective, evidence as to the quality of the first drum can only be used inductively. But if the issue is whether P purchased a second drum of paint prior to notice that a drum of the same kind of paint caused leaks in her roof, then evidence that the first drum of paint was defective supports a deductive argument. When does the inductive reasoning take place in this situation? With which part of the deductive process is the law of evidence primarily concerned?

div1.gif (1531 bytes)
Home | Contents | Topical Index | Syllabi | Search | Contact Us | Professors' Pages
Cases | Problems | Rules | Statutes | Articles | Commentary