consequences. If an impartial revision of deduction then leads to the detection of new consequences which agree with the new facts, such added agreement greatly increases the probability of correctness adjudged on the previous agreement. More significant still is it, when certain peculiar or complicated consequences are deduced, for which no corresponding facts had been previously discovered, and when a return to the field of observation discovers facts of the peculiar character and in the significant situation assigned to them by deduction. This gives wonderful strength to the hypothesis from which consequences so prophetic can be derived: indeed, evidence of this is usually regarded as convincing, for the possibility of such a degree of accordance of consequence and fact being the work of chance is practically ruled out. Finally, if in the course of years, many investigators find many complicated facts in many parts of the world, all of which are successfully matched by the elaborate consequences of an hypothesis that was invented long before observation was so widely extended, the probability of correctness rises to so high an order that the truth of the hypothesis may be accepted, and it may be promoted to the rank of an established theory. The unseen facts that such a theory reveals are commonly accepted as of an equal degree of verity with the facts of direct observation.
The will or the wish of the sane investigator has no power to withhold belief, when this stage of theorizing is reached. And yet it can not be too carefully borne in mind that even if all the above requirements are satisfied, the most that can be said for the established theory is that its probability of correctness is so high that its chance of error may be disregarded. The fair-minded Playfair phrased this aspect of our problem admirably a hundred years ago in the case of river valleys:
It is particularly in this matter of the increasing probability of correctness that the nature of geological or geographical proof is so unlike that of geometrical proof. There is never any talk of increasing the probable correctness of a geometrical theorem, when several different demonstrations are given for it. Each demonstration is absolutely correct alone, as far as anything can be absolute in the limited experience of our finite minds. But in our subject, it is always appropriate
- J. Playfair, "Illustrations of the Huttonian Theory of the Earth," Edinburgh, 1802, 102.