There is no lack of definitions of grammaticalization in the literature. Concepts are not pre-given and found in the world surrounding us. Just like anybody, scientists are free to form concepts they find useful. Apart from scientific requirements on the formation of a useful concept, the creation of a concept is based on intuition. In the case of an empirical science such as linguistics, the intuition often emerges from experience and familiarity with a set of data.

By its terminological structure, grammaticalization is a process by which something becomes grammatical. So the question is: what is it that distinguishes the grammatical from phenomena outside grammar. These latter are phenomena of the areas surrounding grammar in the diagrams visualizing the language system. Of special importance are two areas: lexicon and discourse.

• The grammatical differs from the lexical in allowing an analytic access to its units, while access to lexical units is holistic. So if something that has its source in the lexicon is grammaticalized, it becomes manageable by rules concerning its selection and combination.
• The grammatical differs from the textual in being subject to constraints which have a structural basis that may be arbitrary from a semantic point of view, while the selection and combination of units at discourse level is motivated semantically. So if something that has its source in discourse is grammaticalized, its selection and combination (Jakobson 1956:242f) becomes subject to purely structural and increasingly arbitrary constraints.

If one accepts these properties of the concept of grammaticalization, one may then define it as has been done in the introductory chapter:

Grammaticalization is the subjection of a linguistic construction or construction schema or of a linguistic operation to rules of grammar.

Grammaticalization is a gradual process, so its manifestations are even harder to compare across languages than instances of grammatical categories or of lexical meanings. The concept of grammaticalization must be operationalized so that one working with it is in a position to decide whether a phenomenon at hand fulls under the concept or not. This involves the formulation of criteria by which we recognize grammaticalization in progress and may even be able to measure the degree to which something has been grammaticalized – the degree of its grammaticality.

The voluminous literature on grammaticalization contains a set of phenomena which commonly accompany processes of grammaticalization, may contribute to a better understanding of the concept of grammaticalization and may even be ingredients of its definition. This chapter provides a discussion of such subordinate concepts. It distinguishes between a set of parameters which define grammaticalization and a set of what may be called accompanying symptoms of grammaticalization which either follow from the definition or are not definitory. Among these latter is also the phenomenon of persistence, which receives separate treatment in ch. 9.