Again, take the still hardier and more numerous branches of the _Quercus_ or oak family. M. De Candolle has expended a vast deal of ingenuity to show that the various members of this old and ancestrally-knotty family have all descended from two or three of the hardier varieties. He arrives at this conclusion from a geographical survey of what he would call the "whole field of distribution," and "the probable historical connection between these congeneric species." But science should deal with as few probabilities as possible, especially where experience furnishes no guide to certainty, and only the remotest clue to likelihood. We should never predicate probabilities except on some degree of actual evidence, or some likelihood of occurrence, falling within the limits, analogically or otherwise, of human observation and experience. In no other way can we determine whether an event is probable or not. But here we have not so much as a probable experience to guide us. Geographical distribution in the past is hardly a safe criterion to go by, because we can never be absolutely certain that we have the requisite data on which to form a determinate judgment. The _Quercus robur_ may furnish the maximum test to-day, but a few concealed pockets of nature may bring some other variety of the congeneric species to the front to-morrow, requiring M. De Candolle to correct his classification. There are no less than twenty-eight varieties of this one species of oak, all of them conceded to be spontaneous in origin, and it has been on the earth quite as long as the more stately tribe of Sequoias. Besides, not more than one twenty-thousandth part of the earth's surface has been dug over to determine the extent to which any one of its varieties has flourished in the past.