In the first stage of speech, there are no inflections at all, separate words serving instead of them:—just as if, instead of saying fathers, we said father many, or father father; reduplication being one of the make-shifts (so to say) of this period. The languages allied to the Chinese belong to this class.
In the second stage, the separate words coalesce, but not so perfectly as to disfigure their originally separate character. The Hungarian persons have illustrated this. Language now becomes what is called agglutinate. The parts cohere, but the cohesion is imperfect. The majority of languages are agglutinate.
The Latin and Greek tongues illustrate the third stage. The parts originally separate, then agglutinate, now become so modified by contact as to look like secondary parts of a single word; these original separate substantive characters being a matter of inference rather than a patent and transparent fact. The s in fathers (which is also the s in patre-s and [a]πάτερε-ς]) is in this predicament.
Lastly, inflections are replaced by prepositions and auxiliary verbs, as is the case in the Italian and French when compared with the Latin.
Truly, then, may we say that the phenomena of speech are the phenomena of growth, evolution, or development; and as such must they be taught. A cell that grows,—not a crystal that is built up,—such is language.
But these well-devised selections of suggestive examples, whereby the student may rise from particulars to generals, &c., are not to be found in the ordinary grammars. Indeed, it is the very reverse of the present system; where there are twenty appeals to the memory in the shape of what is called a rule, for one appeal to the understanding in the shape of an illustrated process. So much the worse for the existing methods.
Moulds applied to growing trees—cookery-book receipts for making a natural juice—these are the parallels to the artificial systems of grammar in their worst forms. The better can be excused, sometimes recommended; even as the Linnæan system of botanical teaching can, in certain cases, be used with safety, provided always that its artificial character be explained beforehand, and insisted on throughout.
To stand on the level of the Linnæan system, an artificial grammar must come under the following condition:—It must leave the student nothing to unlearn when he comes to a natural one.
How can this be done? It can be done, if the grammarian will be content to teach forms only, leaving processes alone. Let him say (for instance) that the Latin for—