The existence of phenomena of grammaticalization makes us see that grammar is not given a priori; there are processes creating it. This raises the more general question of how grammar comes into existence. Putting it the other way around: is grammaticalization the only process generating grammar, so that grammaticalization is the same as the development or evolution of grammar?

Most of the time, grammatical structure evolves out of grammatical structure. We have seen two major mechanisms of grammatical change which take grammatical constructions as their input: Both analogical change and reanalysis produce a grammatical construction on the model of a grammatical construction or construction schema which exists in the language. Thus, by their very nature, neither of these two processes of grammatical change has the power of creating new grammatical structure; they presuppose it.

The notion of genesis of grammar presupposes that grammar is created on a basis which is something else than grammar. The following schema visualizes the areas in which the language system is embedded.

Possible sources for new components of the language system are therefore encyclopedic knowledge, pragmatics and phonetics. Encyclopedic knowledge feeds the semantics; phonetics feeds the phonology. Neither of these components of the language system is part of the significative system.

The significative system, in turn, can only be fed by components which border on it. These are, first of all, semantics and phonology. Their roles in this respect are different. Semantics feeds grammar when lexical items are recruited for grammaticalization; to this we turn in a moment. Phonology feeds grammar by morphologization.

With the exception of morphologization, structure in the significative system of language has a double source: Constructions are first formed freely in discourse and then transferred into the syntax by conventionalization. Such a construction may contain a pivotal element – a lexical item or a grammatical formative – which is crucial for the construction. Its source is of necessity the lexicon. By its conventionalization, the construction, and with it the pivotal element, are grammaticalized. In this way, entities which come from the discourse are absorbed into the language system. This idea has given rise to the slogan of ‘the discourse basis of grammar’, which has been circulating in Californian functional linguistics since the 1980s.

A possible misunderstanding concerning the relation between grammar and discourse brought about by grammaticalization must be dispelled. Discourse is the source of grammaticalized constructions. It is, of course, not the source of particular grammatical formatives; these are necessarily recruited from the lexicon. However, there has also been talk about “discourse markers”, a concept which has never been defined properly but which apparently comprises expressions which somehow control the flow of discourse. English items of this class may be now, so, well and even more complex expressions like this being said. There has been some research claiming that such discourse markers have been grammaticalized. This would then seem to show a transition of items into discourse. However, if one applies the criteria of grammaticalization to these discourse markers, the only ones that occasionally work are desemanticization and erosion. The evidence of such items being subject to structural constraints is mostly very limited. This is a field that needs more investigation.

The answer to the question whether grammaticalization is the only mechanism creating grammar is therefore the following: If ‘creation of grammar’ is taken in a literal sense, then grammaticalization is, apart from the morphologization of phonological alternations, indeed the only process. The consequences of this conclusion are rather drastic:

• Every single grammatical formative in every language of the world must have a lexical source. At this point, it must be conceded that this has not even been shown for SAE languages with a well-documented history. For instance, no source is known for the negator and for demonstrative roots in Indo-European languages. Of course, these sources may lie back in prehistorical times, and it may just be a methodological problem that we cannot attain them. The backside of this situation is that one does not know what other kind of source these items could possibly have. It seems inconceivable that they started their life as a grammatical formative – a negator and a demonstrative, in these cases.
• All natural languages have grammar. Whether all of them share a monogenesis or language evolved independently at different places of the planet, the first language must have acquired grammar in some way. Those human beings first created signs whose formation was motivated in some way because this was the only basis on which they could form and understand them. Those were, consequently, concrete significata paired with onomatopoetic and sound-symbolic significantia. Given these conditions, newly created signs could not be grammatical formatives. Since no basis is known on which grammatical formatives could be created, the only solution is that they came about by grammaticalization. The same goes for grammatical structure: it was created by grammaticalizing discourse structure.