A single language is called a stock or family when it is not found to be cognate with any other language.  Languages are said to be cognate when such relations between them are found that they are supposed to have descended from a common ancestral speech.  The evidence of cognation is derived exclusively from the vocabulary.  Grammatic similarities are not supposed to furnish evidence of cognation, but to be phenomena, in part relating to stage of culture and in part adventitious.  It must be remembered that extreme peculiarities of grammar, like the vocal mutations of the Hebrew or the monosyllabic separation of the Chinese, have not been discovered among Indian tongues.  It therefore becomes necessary in the classification of Indian languages into families to neglect grammatic structure, and to consider lexical elements only.  But this statement must be clearly understood.  It is postulated that in the growth of languages new words are formed by combination, and that these new words change by attrition to secure economy of utterance, and also by assimilation (analogy) for economy of thought.  In the comparison of languages for the purposes of systematic philology it often becomes necessary to dismember compounded words for the purpose of comparing the more primitive forms thus obtained.  The paradigmatic words considered in grammatic treatises may often be the very words which should be dissected to discover in their elements primary affinities.  But the comparison is still lexic, not grammatic.

A lexic comparison is between vocal elements; a grammatic comparison is between grammatic methods, such, for example, as gender systems.  The classes into which things are relegated by distinction of gender may be animate and inanimate, and the animate may subsequently be divided into male and female, and these two classes may ultimately absorb, in part at least, inanimate things.  The growth of a system of genders may take another course.  The animate and inanimate may be subdivided into the standing, the sitting, and the lying, or into the moving, the erect and the reclined; or, still further, the superposed classification may be based upon the supposed constitution of things, as the fleshy, the woody, the rocky, the earthy, the watery.  Thus the number of genders may increase, while further on in the history of a language the genders may decrease so as almost to disappear.  All of these characteristics are in part adventitious, but to a large extent the gender is a phenomenon of growth, indicating the stage to which the language has attained.  A proper case system may not have been established in a language by the fixing of case particles, or, having been established, it may change by the increase or diminution of the number of cases.  A tense system also has a beginning, a growth, and a decadence.  A mode system is variable in the various stages of the history of a language.  In like manner a pronominal system undergoes changes.  Particles may be prefixed, infixed, or affixed in compounded words, and which one of these methods will finally prevail can be determined only in the later stage of growth.  All of these things are held to belong to the grammar of a language and to be grammatic methods, distinct from lexical elements.