Grammaticalization produces a new association of meaning and form in a grammatical construction (schema) of a language system. There are, however, two degrees of novelty here:

1. An established construction preserves its function while its form is replaced. This is renewal.
2. A construction with a grammatical function not fulfilled before in the language is produced. This is innovation.

A grammaticalization process may enrich a language system by a new grammatical category. Examples from the history of Indo-European languages include the following:

• One example is the introduction of a numeral classifier system in the Persian language.
• The development of an article system, consisting of a definite and an indefinite article, in the Germanic and Romance languages is another, often investigated case of this sort. These languages stem from Proto-Germanic and from Vulgar Latin, resp.; and these latter languages definitely had no articles, working in the relevant respects like many other articleless languages like Chinese or Cabecar. Grammaticalization here created an entirely new grammatical category as far as the languages in question are concerned. It is true that the grammaticalization of the article in Germanic and Romance involved language contact between these two subfamilies, in whichever direction. However, this detracts little from the innovative nature of this particular grammaticalization process since the languages of these two subfamilies had no contact with other article languages.
• The Romance languages grammaticalize demonstrative pronouns not only to articles, but also to personal pronouns which then become pronominal clitics and cross-reference indexes. These latter grammatical categories are a novelty if compared with the grammar of Latin.

While renewal is a conservative change which upholds the traditional system, innovation is revolutionary. The scope of the concept, though, is the individual language system. Consequently, no matter whether the introduction of a new category is triggered by language contact or has a purely intrinsic motivation, it is an innovation in the language concerned.

Moreover, innovation with respect to human language as such would be relevant for the study of the evolution of language. No doubt, at certain points in this long period, one variety was the first to introduce numeral classifiers or case relators or gender. The only thing one can be sure about for all of these innovations is that they are due to grammaticalization. This is on purely theoretical grounds, as there are, of course, no data on such events. Even today, where a considerable portion of the world's languages has been described, the only approach methodologically available to an innovation on a global scale is the discovery of a unique position feature in a language, i.e. a feature that this is the only language known to possess it. A collection of such “nonesuch” properties is in the Rara & Universals Archive. For every feature that is proper to only one language, this language must have innovated it at one point, possibly independently from contact with any other language. However, given our insufficient knowledge on living languages and our deficient or lacking knowledge of extinct languages, such speculation does not lead very far.

What must be retained, at any rate, is the unique role of grammaticalization in grammatical change: by contrast with all other kinds of grammatical change, including reanalysis and analogical change, grammaticalization is the only process which has the power to lead to innovation (Meillet 1912 [1921]: 133, Lehmann 2004: 184-186). A theoretical consequence of this insight is that the development of grammar in the evolution of human language must, to a large extent, be an outcome of grammaticalization.