Of this last influence a few illustrations must be given. In chemistry it is seen in the facts, that the dividing of matter into the four elements was ostensibly based upon the single property of weight; that the first truly chemical division into acid and alkaline bodies, grouped together bodies which had not simply one property in common, but in which one property was constantly related to many others; and that the classification now current, places together in groups _supporters of combustion_, _metallic and non-metallic bases_, _acids_, _salts_, &c., bodies which are often quite unlike in sensible qualities, but which are like in the majority of their _relations_ to other bodies. In mineralogy again, the first classifications were based upon differences in aspect, texture, and other physical attributes. Berzelius made two attempts at a classification based solely on chemical constitution. That now current, recognises as far as possible the _relations_ between physical and chemical characters. In botany the earliest classes formed were _trees_, _shrubs_, and _herbs_: magnitude being the basis of distinction. Dioscorides divided vegetables into _aromatic_, _alimentary_, _medicinal_, and _vinous_: a division of chemical character. Caesalpinus classified them by the seeds, and seed-vessels, which he preferred because of the _relations_ found to subsist between the character of the fructification and the general character of the other parts. While the "natural system" since developed, carrying out the doctrine of Linnaeus, that "natural orders must be formed by attention not to one or two, but to _all_ the parts of plants," bases its divisions on like peculiarities which are found to be _constantly related_ to the greatest number of other like peculiarities. And similarly in zoology, the successive classifications, from having been originally determined by external and often subordinate characters not indicative of the essential nature, have been gradually more and more determined by those internal and fundamental differences, which have uniform _relations_ to the greatest number of other differences. Nor shall we be surprised at this analogy between the modes of progress of positive science and classification, when we bear in mind that both proceed by making generalizations; that both enable us to make previsions differing only in their precision; and that while the one deals with equal properties and relations, the other deals with properties and relations that approximate towards equality in variable degrees.