But we need not depend on purely logical arguments, when we have historical evidence to appeal to. As far as we know the history of language, we see it everywhere confined within those three great strata or zones which we have just described. There are inflectional changes, no doubt, which cannot as yet be explained, such as the _m_ in the accusative singular of masculine, feminine, and in the nominative and accusative of neuter nouns; or the change of vowels between the Hebrew _Piel_ and _Pual_, _Hiphil_ and _Hophal_, where we might feel tempted to admit formative agencies different from juxtaposition and combination. But if we consider how in Sanskrit the Vedic instrumental plural, +aśvebhis+ (Lat. _equobus_), becomes before our very eyes +aśvais+ (Lat. _equis_), and how such changes as _Bruder_, brother, and _Brüder_, brethren, _Ich weiss_, I know, A.S. _wât_, and _Wir wissen_, we know, A.S. _wit-on_, have been explained as the results of purely mechanical, _i.e._, combinatory proceedings, we need not despair of further progress in the same direction. One thing is certain, that, wherever inflection has yielded to a rational analysis, it has invariably been recognized as the result of a previous combination, and wherever combination has been traced back to an earlier stage, that earlier stage has been simple juxtaposition. The primitive blocks of Chinese and the most perplexing agglomerates of Greek can be explained as the result of one continuous formative process, whatever the material elements may be on which it was exercised; nor is it possible even to imagine in the formation of language more than these three strata through which hitherto all human speech has passed. All we can do is to subdivide each stratum, and thus, for instance, distinguish in the second stratum the suffixing (R + ρ) from the prefixing (ρ + R), and from the affixing (ρ + R + ρ) languages. A fourth class, the infixing or incapsulating languages, are but a variety of the affixing class, for what in Bask or in the polysynthetic dialects of America has the appearance of actual insertion of formative elements into the body of a base can be explained more rationally by the former existence of simpler bases to which modifying suffixes or prefixes have once been added, but not so firmly as to exclude the addition of new suffixes at the end of the base, instead of, as with us, at the end of the compound. If we could say in Greek δείκ-μι-νυ, instead of δείκ-νυ-μι, or in Sanskrit +yu-mi-na-j+, instead of +yu-na-j-mi+, we should have a real beginning of so-called incapsulating formations.[9]