The only place where our problems permit of the use of analogy is in the making of so-called constructions, i.e., when we aim to clarify or to begin the explanation of a case which is at present unintelligible, by making some assumption. The construction then proceeds in analogy to some already well known earlier case. We say: “Suppose the case to have been so and so,” and then we begin to test the assumption by applying it to the material before us, eliminating and constructing progressively until we get a consistent result. There is no doubt that success is frequently attained in this way and that it is often the only way in which a work may be begun. At the same time, it must be recognized how dangerous this is, for in the eagerness of the work it is easy to forget that so far, one is working only according to analogy by means of an assumption still to be proved. This assumption is in such cases suddenly considered as something already proved and is counted as such with the consequence that the result must be false. If you add the variability in value of analogy, a variability not often immediately recognized, the case becomes still worse. We have never been on the moon, have therefore apparently no right to judge the conditions there--and still we know--only by way of analogy--that if we jumped into the air there we should fall back to the ground. But still further: we conclude again, by analogy, that there are intelligent beings on Mars; if, however, we were to say how these people might look, whether like us or like cubes or like threads, whether they are as large as bees or ten elephants, we should have to give up because we have not the slightest basis for analogy.
In the last analysis, analogy depends upon the recurrence of similar conditions. Therefore we tacitly assume when we judge by analogy that the similarity of conditions contains an equivalence of ultimately valid circumstance. The certainty of analogy is as great as the certainty of this postulate, and its right as great as the right of this postulate.
If, then, the postulate is little certain, we have gained nothing and reach out into the dark; if its certainty is great we no longer have an analogy, we have a natural law. Hence, Whately uses the term analogy as an expression for the similarity of relation, and in this regard the use of analogy for our real work has no special significance. Concerning so-called false analogies and their importance cf. J. Schiel’s Die Methode der induktiven Forschung (Braunschweig 1865]).