form-feeling, we need not be too much surprised to find that they seek and avoid certain linguistic developments in common.  We are at present very far from able to define just what these fundamental form intuitions are.  We can only feel them rather vaguely at best and must content ourselves for the most part with noting their symptoms.  These symptoms are being garnered in our descriptive and historical grammars of diverse languages.  Some day, it may be, we shall be able to read from them the great underlying ground-plans.

[Footnote 116:  Not Greek specifically, of course, but as a typical representative of Indo-European.]

Such a purely technical classification of languages as the current one into “isolating,” “agglutinative,” and “inflective” (read “fusional”) cannot claim to have great value as an entering wedge into the discovery of the intuitional forms of language.  I do not know whether the suggested classification into four conceptual groups is likely to drive deeper or not.  My own feeling is that it does, but classifications, neat constructions of the speculative mind, are slippery things.  They have to be tested at every possible opportunity before they have the right to cry for acceptance.  Meanwhile we may take some encouragement from the application of a rather curious, yet simple, historical test.  Languages are in constant process of change, but it is only reasonable to suppose that they tend to preserve longest what is most fundamental in their structure.  Now if we take great groups of genetically related languages,[117] we find that as we pass from one to another or trace the course of their development we frequently encounter a gradual change of morphological type.  This is not surprising, for there is no reason why a language should remain permanently true to its original form.  It is interesting, however, to note that of the three intercrossing classifications represented in our table (conceptual type, technique, and degree of synthesis), it is the degree of synthesis that seems to change most readily, that the technique is modifiable but far less readily so, and that the conceptual type tends to persist the longest of all.

[Footnote 117:  Such, in other words, as can be shown by documentary or comparative evidence to have been derived from a common source.  See Chapter VII.]

The illustrative material gathered in the table is far too scanty to serve as a real basis of proof, but it is highly suggestive as far as it goes.  The only changes of conceptual type within groups of related languages that are to be gleaned from the table are of B to A (Shilluk as contrasted with Ewe;[118] Classical Tibetan as contrasted with Modern Tibetan and Chinese) and of D to C (French as contrasted with Latin[119]).  But types A :  B and C :  D are respectively related to each other as a simple and a complex form of a still more fundamental type (pure-relational, mixed-relational).  Of a passage from a pure-relational to a mixed-relational type or vice versa I can give no convincing examples.