The resemblance of C to S may, however, vary in degree from the faintest analogy to community of species. The difference between them may far outweigh their resemblance. There may even be no material likeness, but only a similarity of function, or position, or of any the most trivial attribute. Only it is to be observed that the kind and degree of resemblance between S and C determine the kind and degree of resemblance between A and _I_. We must not infer specifically unless the case is specifically like the precedent. In all other instances we can only infer proportionally or by transfusion. (2) None of the antecedents must be a verbal or identical proposition, that is, a proposition which merely substitutes one name or nominal phrase for another; nor must the case be merely the precedent expressed in other words, or the precedent a paraphrase of the case. In any of these circumstances one of the elements of the argument is wanting; we have two names for one thing or two propositions giving the same information. (3) The precedent may (as has been already remarked) be a general idea, or may be an individual idea or object. If S A has occurred frequently it is certain to be generalised, and so may form a maxim, a law, a rule, an induction, &c. But one well-observed precedent is enough to suggest a conclusion, if there has been no experience to the contrary. There is therefore no dialectical difference between arguing from a general idea (class notion) to an individual or subordinate idea, and arguing from one individual to another. Comparison and inference occur in both. (4) After separating A from S care should be taken that it is A and not S that is used to generate _I_. Examples are plentiful of theorems in which S and A change parts, which invalidates the conclusion. Other errors in stating theorems intended to be arguments will be noticed under the head of 'Fallacies.' The following is an argument conformable to the above rules. Tyrants | deserve death --------------------+----------------------------- Caesar was a tyrant | _no doubt he deserved death_ This square mode of stating the argument is adapted from the general type, and brings out the mutual relations of the compared parts better than the three-lined arrangement. The word 'therefore,' which usually introduces a logical conclusion, is ambiguous. It may mean that the antecedents are the causes of the fact mentioned in the conclusion, or merely that the antecedents are the reasons why we believe the conclusion. The former is the scientific 'therefore,' the latter is the purely dialectical. I shall generally omit the illative word, and print conclusions in italics, besides entering them invariably in the fourth compartment of the parallel when this arrangement is adopted.