This view is further confirmed, and the predominance of this notion of likeness in primitive times further illustrated, by the fact that our system of presenting ideas to the eye originated after the same fashion. Writing and printing have descended from picture-language. The earliest mode of permanently registering a fact was by depicting it on a wall; that is--by exhibiting something as _like_ to the thing to be remembered as it could be made. Gradually as the practice grew habitual and extensive, the most frequently repeated forms became fixed, and presently abbreviated; and, passing through the hieroglyphic and ideographic phases, the symbols lost all apparent relations to the things signified: just as the majority of our spoken words have done. Observe again, that the same thing is true respecting the genesis of reasoning. The _likeness_ that is perceived to exist between cases, is the essence of all early reasoning and of much of our present reasoning. The savage, having by experience discovered a relation between a certain object and a certain act, infers that the _like_ relation will be found in future cases. And the expressions we constantly use in our arguments--"_analogy_ implies," "the cases are not _parallel_," "by _parity_ of reasoning," "there is no _similarity_,"--show how constantly the idea of likeness underlies our ratiocinative processes. Still more clearly will this be seen on recognising the fact that there is a certain parallelism between reasoning and classification; that the two have a common root; and that neither can go on without the other. For on the one hand, it is a familiar truth that the attributing to a body in consequence of some of its properties, all those other properties in virtue of which it is referred to a particular class, is an act of inference. And, on the other hand, the forming of a generalization is the putting together in one class, all those cases which present like relations; while the drawing a deduction is essentially the perception that a particular case belongs to a certain class of cases previously generalized. So that as classification is a grouping together of _like things_; reasoning is a grouping together of _like relations_ among things. Add to which, that while the perfection gradually achieved in classification consists in the formation of groups of _objects_ which are _completely alike_; the perfection gradually achieved in reasoning consists in the formation of groups of _cases_ which are _completely alike_.