In this paper, I claim that there are two kinds of elements which are involved in grammaticalization. One is a contributor which triggers a given change or contributes to the change, while the others are “free riders”, which are grammaticalized thanks to the contributor. I take up the issue of the emergence of a determiner system, i.e. a DP in English. The articles, definite the and indefinite a/an, are believed to have developed from the Old English demonstratives se/seo and the numeral an ‘one’ (cf. Sommerer 2011). This is an instantiation of grammaticalization (Hopper and Traugott 2003). I assume that a syntactic determiner system, a DP (Abney 1987), was absent in Old English. More specifically, I make two proposals: ① grammaticalization should be defined within the framework of functional category emergence described below, suggesting a unidirectionality of language change; ② grammaticalization progresses in two stages: a primary grammaticalization and a secondary one.

The items involved in secondary grammaticalization are “free riders”, which are grammaticalized due to contributors. Free riders are grammaticalized by making use of the space created by contributors. The lexical elements that contribute to creating a new functional space in the structure are contributors. That is, contributors cause the primary grammaticalization. I argue that the contributors in the DP emergence grammaticalization are the demonstratives se/seo and the free rider is the numeral an ‘one’ in Old English. It is well known that the indefinite article appeared later than the definite article in English. I argue that this later emergence can be accounted for if we assume that grammaticalization progresses in two stages. The emergence of the indefinite article a/an is an instance of secondary grammaticalization. Hence, the ancestor of the indefinite article, i.e. Old English numeral an ‘one’, is a free rider that was grammaticalized in the space created by the contributor se/seo. The emergence of the definite article the is an example of primary grammaticalization. Hence there is a time difference in appearance between the two determiners. This hypothesis also leads to a whole review of existing grammaticalization theory. I will propose a new theory of grammaticalization. Assuming that grammaticalization is a structural change as argued by Roberts and Roussou (2003) and Gelderen (2004), I claim that grammaticalization means creating a functional space in a given structure.